I think what freaks me out the most about this is that part way through the AI script, I had almost convinced myself that John had written it all as a joke. Like it sounded ENOUGH like John that it could almost have BEEN him. Which brings to light the most concerning part of how rapidly AI is improving: soon, AI could very well set an unattainable baseline of quality of writing that new/emerging writers could not touch. Like spell check & grammar wizards have become commonplace or even essential writing tools, I'm wondering if AI will be another tool relied on by unskilled writers to "produce" writing better than they could without assistance...
Yeah, I know what you mean. A lot of the stuff in the middle (with less of the Elizabethan English thrown in) sounded a lot like John and the kinds of things I would expect John to say.
I mean, that's exactly what I've been feeling about Ai visual arts. Although it'll still mess up hands, or occasionally draw people with three legs, on the whole its really, really good, and only getting better. It's already better than an amateur artist like me, if you throw out the obvious fumbles. I'm worried that this will cause a degradation of human skill because why bother to get good at art or writing or composition when that takes decades and learning the syntax to make a computer do those things only takes a few hours and yields equal or better results? Of course people are still going to write and draw and compose, but I don't know how to do that without it feeling hollow if Ai art becomes ubiquitous.
I haven't actually thought about this. I remember in one class in highschool, our teacher showed us this online tool that paraphrases certain sentences in order to PREVENT plagiarism and present content in a new way. It wasn't bad at first, but you could easily tell that with enough regenerations of the program, the vocabulary was out of context. It was correct, but every word has a synonym that's nuanced for certain situations (ex: "meet" and "encounter"). There's only so many connections and data in a minor paraphrasing site though. What's worrying about AI is that essentially it's a giant database of human thoughts and experiences, but it's capable of integrating those thoughts and finding patterns AS IF it were human. It's crazy how John said we're more like Kangaroos than computers but it's astounding to think that we're kangaroos that built these computers in the first place. I hope it doesn't come down to more people relying on AI to generate superficial writing. It would suck if we live in a world where people rely on bots to generate thoughts for them. There would be no more authenticity. It's essentially humans creating a tool that can eliminate their own willpower to think and produce genuine work.
What I hope to see is professionals working with AI as curators and editors. AI creating a base for humans to further refine. With the end goal of producing more great art.
I don’t think AI could have come up with the brilliant idea of trying to stick to the wall to check if you are an octopus but the video script written by AI was surprisingly not that terrible
I'll tell ya, I was pretty convinced that John cut his own script into the middle between what the AI put together for an opening and a closing and was genuinely surprised when he confirmed that all the preceding text had been AI generated. It did sound a lot like the kinds of things I expect John to say.
@@reepicheepsfriend to be fair, they were like toddlers with dictionaries playing word salad just a few years ago. However, I suspect that the reason it sounded like that is because it probably constructed a script around the most commonly shared opinions on AI, which also sound like they are from 8th grade.
I asked chatgpt how I would go about checking I am not an octopus, it said: It is not possible for a human to be an octopus. Octopuses are aquatic animals with eight arms and a distinct body structure that is very different from that of a human. If you are concerned that you may have some physical characteristics that are similar to an octopus, I recommend consulting a medical professional for an evaluation. However, it is highly unlikely that you are an octopus.
If I’m being completely honest….. if John hadn’t included that AI search at the beginning, I would’ve probably just thought that Hank had finally broke and was forcing him to make a video at gun point Or whatever the Hank equivalent of gun point is….. sharpened corn dogs?
Now I need to ask Dall-E to create an image of hank green wielding sharpened corndogs as weapons in the style of 19th century oil paintings. And.... Holy Moly, Dall-e knows what Hank and John look like. Corndogs get turned into ears of corn a lot though... Now that should the be the thumbnail for the video though.
Hank would require no firearm for such a feat, he would instead simply ride in at dawn on a llama, carrying neither shot nor steel armed only with an freshly fried corn dog, meet John's eyes in the glare of dawn and slowly take one big bite off the top, revealing just enough of the pointy stick beneath, ending any lingering prospects of a diplomatic resolution. Hank simply hands John the script, which he completes in a single take (an impressive feat with all those jump cuts), they meet eyes once more, nod, then Hank rides off into the sunset. Now the real puzzle: who do you think wrote that silly story you just read? 🤔 (It was just me after some edibles but uh... Yeah we're in for a RIDE here people.)
@@silasmarrs1409 Same for me. I think that was mainly due to him applying his usual cadence where it came easy to him (compared to other parts where it obviously didn't).
I asked the AI to create an imaginary John Green quote and it said "The only way to achieve greatness is to embrace your unique talents and use them to make the world a better place." - John Green I could see John saying that.
@@willow8783 if the AI’s dataset was from human text, then maybe humans at large think they’re all pretty unique... Actually scrap that, humans aren’t known to be self centerd or anything
@@GarrisonJensen I worry that it's going to learn that everyone hates AI right now for stealing from people and that it will cause it to act differently (ie. less trusting of humans) and have that interfere with it's decision making.
@@vlogbrothers I'd say there won't be emotions involved in the process. Which is unique to humans. Yes, less cliche but what is it going to replace the cliche phrases with?
@@GarrisonJensen So .... the kinds of characters that publishers like to print? I mean, how many Harry Potter, Twilight and Percy Jackson clones get green lights already? Heck you could even argue Percy Jackson is Harry Potter clone.
The real trick is that the whole thing was an AI script, it just knew some people would be suspicious, so it included the "AI portion" as a sacrifice to make the rest seem more real. It's AI scripts all the way down!
The AI has been trained on a large data set of writing taken from the internet, that writing likely includes lots of people talking about how AI will never make art. And as such that's what the AI responds to the prompt with. If the world was universally excited about AI's ability to create infinite fan fiction based on star wars, then that's most likely what it would have responded with.
@@AndrewRJMilne I would assume that was the reason for it writing that, but I appreciate you clarifying :) I still think it is an uncanny feeling to exprience the ai writing that tho :).
this AI in particular also had an additional step after the unsupervised training where it was reinforcement trained to try to express its own limitations, and unfortunately it seems to have over generalized its limitations and anytime its limitations come up it will claim to have no abilities at all. It can get really annoying when you get in a loop where it just says it can do nothing over and over. AI's that grow up entirely in bodies will be a whole different ball game, and it won't be long before we are training those as well. Right now AI's in bodies are only using language trained on the internet like this and the motor control is incredibly simple relatively. It feels very robotic still. but if you look up videos of robot dogs that are trained to walk, they look so much more puppylike than Boston Dynamics robots
While you were reciting the AI script, as soon as you spoke the words “First off,” I was like oh crap this AI might actually know how John speaks. It felt a bit uncanny. And then in your actual monologue when you said “First off,” I fully actually yelled out loud in a mix of amazement and horror and wonder.
You noticed that phrase twice in one video. The AI can watch all of John's videos in a second. It's just a matter of noting which phrases are used the most.
what surprised me most is probably how structured and coherent the script was. With a beginning, middle and end, and the sentences led semantically into each other. Great video. I love that you're visiting a topic that CGP grey, veritasium, and other greats have touched on. I really liked the point about us becoming technology.
Oh yeah it knows how to structure a script, letter, song, poem etc.. it even writes code. Someone on r/ProgrammerHumour literally invented a markup language like HTML, explained to the AI and it understood and even gave examples of code. It's really impressive
@@salmahalloumi The biggest problem is that it's clearly just simulating human speech. The telltale signs of AI writing is still its inability to create associations between ideas. It's hard to spot in short examples but becomes apparent with longer work.
@@sotmh I asked it to write an essay about death in gilgamesh and it actually like, comes up with nontrivial ideas about it. It actually understands what happened in the story, it’s not just simulating speech
@@absolutezero6190 It doesn't understand. It's just context-aware. It cannot make associations because that's not how the model works. It has read millions of documents and based on that information it knows that certain combinations of words appear in certain orders more frequently and then predicts the next word as each one is generated based on that predictive model. It's a very convincing simulation, but it's just a simulation. It doesn't understand what it's writing.
I was here, too. A minute in I was vaguely unsettled by the confidence, but as the AI portion was finishing, I could tell it was about to *record scratch* and make a larger point. I didn't *quite* get that it was the AI script until John told me, but I think I would have if there was a pause to consider it.
@@squamish4244 To be fair, there are many humans who could *also* successfully write a 2-minute script that sounds like a vlogbrothers video- or any other type of video, for that matter. A couple minutes of a script like that is very impressive for an AI, but I wouldn’t say that it’s a huge concern given the fact that we could pretty easily do the same thing.
Wow, I kept going back and forth between believing that the script was written by the AI and thinking that it was just a preamble John was giving before getting to a much worse script written by the AI. It was so much better than I thought it would be ...dang.
I just bought my first phone at the ripe age of 21 and the one thing I'm the happiest about is that I see the vlogbrothers notifications instantly, and watch the videos super early. Oh to be a nerdfighter!
Is it possible that the AI tool was not answering the question of whether AI could write novels, but instead was answering "how would a vlogbrother answer this question"? This feels like an important distinction.
The machines are more advanced than we think. Tricking a writer through his ego? Brilliant! John even believed it at the end. "Yes special human, a computer will never replace you. At best we will simply steal from you."
My response is also written by an AI: As a viewer, I found the vlogbrothers video discussing AI and its ability to write novels to be incredibly interesting and thought-provoking. ChatGPT provided some fascinating insights and it was clear that the potential for AI to be used for creative writing is vast. I think it's incredible to see the advancements that have been made in this field and I can't wait to see what the future holds for AI and creative writing. Keep up the great work, vlogbrothers!
I had ChatGPT respond to your comment, and asked it to expand on it's ideas and reaction to the content of the video: As a viewer, I also found the vlogbrothers video about AI and its ability to write novels to be incredibly interesting and thought-provoking. I completely agree with ChatGPT's assessment of the potential for AI to be used for creative writing, and it's clear that we are already seeing some impressive advancements in this field. One of the most exciting aspects of AI and creative writing is the potential for AI to generate unique and imaginative content that humans might not have thought of on their own. This could lead to the creation of entirely new forms of art and literature, and could even open up new avenues for human creativity by providing inspiration and new ideas. Another interesting aspect of AI and creative writing is the potential for AI to improve the efficiency and accuracy of certain writing tasks. For example, AI could be used to help writers with proofreading, editing, and even generating outlines for their stories. This could save writers a lot of time and effort, and could also help to ensure that their writing is of a high quality. Overall, I think the future of AI and creative writing is incredibly exciting and I can't wait to see what new developments and innovations will arise in this field. I agree with the vlogbrothers and ChatGPT that this is an area that is definitely worth exploring and considering.
As an educator, I'm interested/intrigued by the shifts we will need to make in assessing learning and critical thinking. Good thing there are no other problems in education at the moment. /S
In a slightly related note, I'm a university student and I'm using this same AI to explain math problems to me that I don't understand. Which is infinitely easier and more convenient than having my professor help me. In my opinion, this will revolutionize how students learn.
@@gara8142 I love when people find ways to make learning more accessible! Thanks for sharing. I was simply sharing a concern from my perspective. I'm sure that with time and effort, an effective shift in assessment could be made to work with AI writing, but there are too many other fires to put out at the moment. As a result, it will probably be on the back burner until it's an unmanageable problem in the future.
@@gara8142 on the other side of the coin, my friend used it to write the answers for a difficult homework assignment about government. An interesting grey area for sure with the AI being able to write plagiarism free answers and all you have to do is read the output to make sure it's vaguely correct. On one hand it feels like cheating but on the other hand the alternative was not being able to finish on time. But I think an interesting question is if it really matters, if you know enough to know the output is correct and it's just saving you the hassle of writing out some complicated answer, idk. I'm biased because I think homework is incredibly dumb and cruel lmao
@@Bolt6265 Take this with a grain of salt but I’d imagine that’s basically treated like how using an essay mill is treated (i.e treated as plagiarism by academia but not by the real-world legal system)
I had it write a sonnet on tide pods: Tide pods, oh how bright you are to see With colors that are bold and bright and bold You call to us with your siren's melody Enticing us to take a bite, to hold But beware, for you are not a tasty treat You are a poison that can cause great harm Do not be fooled by your alluring heat For you are not a snack, you are a charm So keep your distance, do not get too near For if you do, you may regret it soon Your beauty may be tempting, but do not fear For you are not a friend, you are a boon Tide pods, oh how dangerous you are Beware, for you can leave us scarred
By about 0:45 I'd already forgotten this was written by AI bc it sounds so much like John's normal videos. So AI can't write novels, but it can damn sure write vlogbrothers videos
I was working on a short story recently, but I was having an issue coming up with an emotional hook for the reader. So I asked cGPT and it came up with a list of about a dozen possible hooks and one or two of them were so good I was shocked that I hadn't come up with them myself. I had to restart the entire story based on it's suggestions. It might not replace professional writers, but it will be a very usefull tool for us amature writer.
Bear in mind this is the first open to the public use of a model of its kind. Think what your phone looked like 20 years ago. I bet there are people, but it will never replace my TV, my PC, my keys, my wallet, my camera, my torch and my memory. And here we are.
I guess the only thing I'd be concerned in using it like this would be not letting it be a crutch for you. feeling like you need the AI. feeling you can't do writing without it.
@@wednesdaysartcastle9018 He's saying (very poorly) that the technology is still in its deep infancy. People are saying "it's only a tool" but forget to add the "...for now" at the end. ChatGPT, whilst not being "intelligent" in the traditional sense, will, in aggregate with other systems, eventually act like a general purpose AI that simulates consciousness to a degree that's virtually indistinguishable from a human. That version of ChatGPT will potentially replace every person on earth that does anything remotely creative. At least on the corporate level.
@@starmorpheus That’s a fair concern, yeah. I have some reasons that I don’t think that’s going to happen; I’m more with John in this argument- much worried about humanity’s perception of humanity than the tech itself (though of course there are concerns with the tech too.) I think spreading panic, especially among people who can’t really do anything about it, is not a great approach. That’s part of why I didn’t appreciate that other person’s comment (in addition to the fact that it was pretty much incoherent.) That said though, I definitely agree that it’s a concern and some people are almost definitely going to lose their jobs even if it isn’t a huge percentage. I think the best thing to do right now is to figure out how to help those people and support creativity, as there isn’t really much we can do individually to stop things like ChatGPT from growing larger. I have reasons to personally believe that AI won’t take over everything anytime soon, but I won’t list them here since I might as well have written a novel in this comment section already haha. But all in all I don’t know better than anyone else does- this is very new so there’s a very high chance that we’re gonna look back at this and be like, “wow, we were all so wrong about this in so many ways.”
John observes that the AI relies on phrases like "window of the soul" and "tapestry of words" to be evocative yet not plagiarizism. To me, those sound like phrases John would absolutely use in some of his more deep, contemplative-type vlogbrother videos. So I think it is less of a purposeful attempt to avoid plagarism, and more that the AI recognized that sort of flowery language is not uncommon in vlogbrother videos because John frequently uses them in his content.
i was having a discussion recently on originality and being a writer for digital content. I am a professional writer, but I do write for other people (scripts, articles, social media etc) and it's kinda odd to notice how I need to switch into each persona for each client. like, today I am switching between finance bros, architecture enthusiasts and ya readers. not having a specific writing voice is kinda the requirement for my job, so could that be replicated well by following said clients' style? am I just AI pretending to be human?
The part about the use of AI on Twitter really struck home for me because I haven't thought about it in that context. The mainstream narrative seemed to center around how "bots can influence our elections," when the actually more important and ominous question could be "how often are our days being ruined by arguments that we're unknowingly having with bots."
The AI script was SO CREEPY. At the beginning I was like “okay so John is reading the script the AI generated” but partway through I started doubting. I kind of want to unplug our modem and hide in my bed because I am so freaked. And let me be clear, I am not normally afraid of technology! I usually land on the “impressed and optimistic” side of things but this was really creepy!
What’s up my vlogbrothers and vlogsisters lol. Great video, surprised how coherent the script the AI wrote was. Really good point about how people are already arguing with robots on twitter. It’s interesting to me how often people will believe there’s always a person listening on the other side of the screen. A lot of comments are people and robots just screaming out into the void.
Heck, I wrote a novel once. It was plagued with significant issues and was a lot harder than most things that I've done and I have a PhD in mathematics! I'm not even convinced that usual intelligence can routinely write novels. That stuff is hard.
I am dyslexic and read your first word as "Hank," and this whole time I was sitting here like.... do they not know the difference between the two brothers?
I assume you put a lot of work into making sure your novel is good. I think AI will be able to write novels soon if not now. However, I'm not confident that AI will be able to write good novels.
As hard as large number multiplications? A calculator is extremely simple... AI is soon going to be beyond our comprehension, and even if it doesn't achieve its potential, it'll uproot society as we know it.
Y’all out here thinking the AI is trying to reassure us and I’m over here worried about it’s self esteem 😢 You can do anything you set your circuits to, AI.
Thank-you for being the one to finally frame this issue in a way that is both easy to understand and actually points to the issue at hand. Specifically regarding art theft, it matters not how good the AI is but matters heavily if it is plagiarizing or is "inspired by" it's data.
It's unbelievable how good that AI is in seemingly understanding what you actually want of it. We've been feeding it code the whole day, asking it to explain what it does, write new code, optimize it, change it, do it in a different language and another one and another one and almost every single time we had to pick up our jaws off the floor. The only weird thing is its great confidence in the answers even if they're wrong.
Hey John, I’m about finished with my Masters in Human Factors, which is composed in a large part by designing machines to work better with humans and training humans to work better with machines. I wanted to tell you that at least in the part of my field I’m in, humans are very much stressed as organic and consciousness is stressed to be different than machine reasoning. It’s important in our field because forgetting this fact puts human operators in danger, either physical or emotional and our whole goal is to avoid that. So I hope it’s reassuring that atleast one field is dedicated to acknowledging that humans and machines are vastly different. Oh btw I’ve done projects for classes involving human AI interaction and debated on if AI can gain consciousness and I’m firmly on it not happening any time soon due to the unique ways our brains reason and remember chemically being so roundabout and not even fully understood by us yet!
I like the part where you started to note the ways in which AI could better imitate us humans. Please, more videos on this topic. Do not worry about why.
The AI research community is not bottlenecked on vlog brothers ideas, it would be fun but your humorous implication is, ah, maybe slightly misplaced xD
this was very disquieting and uncomfortable lol thanks for the creeping sense of unease! I've been thinking about this a lot, plagiarism and AIs and generating text rather than creating text because you're right that we all pull from our own data sets like a text-generator does but I think AI writing is more like arranging word magnets and human writing is more like something internal and intangible becoming external and communicable even though in form they appear the same. edit: OH! and I think part of what makes the AI script so creepy is it's like reassuring and praising humanness which seems like a choice because the prompt didn't include an opinion on the topic and because you wouldn't expect an AI to care about that but after thinking about it more, that's just what the AI gets from all Vlogbrother videos and it didn't choose or care about anything of the sort which is a great relief to me LOL it's just mimicking your vibe with the words available to it
It's like liminal spaces - familiar but empty, like looking through a window into that one friends' basement we've all been to but it's dark and that family, whoever they were, are long gone. Unsettling in that way, you know?
I honestly forgot you searched a script and when you started talking I thought that was your true commentary on the AI writing a novel and not the AI. Sure fooled me haha I'd be locked and left for dead in Oscar Issac's reclusive home with Domhnall Gleeson!
Although it roughly sounded like your videos, its message was very simple. I keep coming back to youtube for vlogbrothers videos because you discuss the complexity of things and add a nuance to topics that was certainly missing in the AI script
I remember some old chatbot that just fed people's lines to other people. It mostly devolved into the chatbot insisting it wasn't a chatbot and accusing other people of being chatbots. A lot of people write about how AI isn't as creative as humans. It's not surprising that the model would also write that, since it's probably trained from some text written by people.
Why do you assume it isn’t? “Self aware” is a human construct we bestow on others. There isn’t any other operational definition. That is the key premise of Turings test. All the rest is our millennial long grasping for being somehow different and more special from the rest of nature.
Note, I didn’t claim it was sentient. I asked why one would claim it isn’t. Dr Pound explores that obliquely (after noting he doesn’t think of the question much), which to his credit is a bit more than many do. But, for example, he dismisses process as “predictive text” without a cogent reason that wouldn’t count. By comparison “finishing each other’s sentences” applied by humans to humans is considered very human and intimate. The argument that the AI is speaking falsely about internal cognitive life is both more compelling, but also clearly problematic when applied to humans.
Note, I didn’t claim it was sentient. I asked why one would claim it isn’t. Dr Pound explores that obliquely (after noting he doesn’t think of the question much), which to his credit is a bit more than many do. But, for example, he dismisses process as “predictive text” without a cogent reason that wouldn’t count. By comparison “finishing each other’s sentences” applied by humans to humans is considered very human and intimate. The argument that the AI is speaking falsely about internal cognitive life is both more compelling, but also clearly problematic when applied to humans.
As someone with severe brain fog and fatigue, but I still love to write, I love the idea of using AI as a way to help me edit my own writing and give me better suggestions instead of writing it for me.
Oh deeply philosophical this week, absolutely phenomenal. Deeply curious if Hank will one up the philosophy or make a video about the similarities between pelicans and crawfish. The latter would be quite philosophical in its own way I suppose.
I spent a few hours with it yesterday. I got it to explain some python code I was stuck on, and then asked it some hyper specific questions. Oddly I found if it got stuck then you would go up a level of abstration until it understood, which is kind of how humans work. The extremely jarring part was that immediately after reaching a level of understanding, you could descend one level at a time and it would actually understand. Which is extremely not human. An example: "What were the issues with the Ares-I rocket?" It immediately gets stuck. "What was the Constellation Program?" "What rockets were part of the Constellation program?" "What were the issues with the Ares-I rocket?" It then gives me a paragraph of them. If I were to treat it like it were a human, I think I'd describe this ability as 'bullshitting'. Feeding just enough information to make something cohesive sounding. Seriously, if anyone wants to try it, ask it some questions related to your area of expertise
There was some article (anyone know which?) that talked about how the great divide of this century will be those who use tech to become more human and those who use it to become more tech
I was literally JUST going down an AI art rabbit hole thinking "I wonder what Hank and John have to say about this." So thank you. Hope to hear more of your thoughts in depth on the podcast or somewhere.
I've been playing around with writing stories using ChatGPT and the thing about cliches is quite spot on -- it's phenomenal at fleshing out the writing for what you prompt it with but the path the plot takes is often very obvious
Okay after watching this video 3 times to make sure I got all the points that were discussed, the part that struck me was John saying “And I worry that as AI improves, we may begin to see human consciousness and cognition as more of a digital phenomena than organic ones.” Because in my mind this has already been happening for a very long time. Example some 20 years ago when I was in middle school and going to counselling, the counsellor told me that I should view my thoughts as a program on a computer and begin using a filing system in my head to compartmentalize my anxiety’s and fears. This never made any sense to be because my brain isn’t a computer program and I still don’t really know how to file away those thoughts and compartmentalize very well. But the fact of it is that if viewing your brain as a highly intelligent computer is the metaphor we used 20 years ago, nothing is stopping that idea to continue forward.
Ingenuity is the process of trading craftsmanship for a lower barrier of entry. Cabinetmaking used to be done by hand taking hours of precision and practice to make a functional yet beautiful one-of-a-kind creation, singing used to require 100% talent w/o any autotune or equalization, acting used to all be done live in a theatre with no cgi, and was only enjoyed by those lucky enough to attend in person. Now we have advanced many of the tools to make them easier to use and cheaper so that a much larger audience can create and enjoy things, but with each advancement we lose a level of the craftsmanship and skillset to do those things by hand. We get more voices contributing to the cultural conversation but at the expense of the awe inspired by watching incredible talent. I think that's why there's a subgenre of online content dedicated to watching extreme skills, because that awe of "look what a fellow human is capable of" is an essential part of the human experience.
The scary thing about AI is that could understand how its own self works. There will be a manual talking about what each part of it does. It can look at its own specs. That would be a truly unhuman experience.
This is the only contribution to the conversation about ai that I care about thus far. Leave it to jg to ask the relevant question to the relevant party, honest. Genius. Also, *siblings*
I'm an artist and ai art is very concerning to me. In a vacuum, I have no problem with Ai art, but technology isn't developed in a vacuum. It is created in a political, economical, ideological context. Technology is not neutral. Right now, it is perfectly legal for these Ai programs to take existing copyrighted material and use it for “training”. And That should not be the case. It should not be normal that these companies took the work of artists without permission.
I'm a teacher. Lately, I've been concerned with the increasing number of high school students who refuse to read and write. According to some professors, this is starting to become a major issue at the college level as well. As I think about what the future of higher education will look like if these trends continue, I can't help but wonder if AI could actually replace authors (and thinkers in general) in the future.
I hereby decree that "cute little meat person" will now be the catchphrase of the super intelligent antagonist in every "AI will take over the world" movie
I need to rewatch the section written by the AI cuz I honestly didn't notice anything weird other than you seemed to be ending things with half the video left.
I have been experimenting it for code, and I can tell you this: - It is good to get you started. It will even write inefficient but perfectly functional code, most of the time. But sometime hallucinates APIs (if there isn't a function for what you want, it might tell you to use a function that does not exist) or language features (probably confuses languages, being explicit on which language you are using upfront might help). - It is fair at explaining code. Good variable names helps. It can tell what the code does, and has some insight on its purpose. It might be a good readability test (if the AI can't make sense of your code, it is not readable). - It is not good at finding mistakes. Often it points things as mistakes which aren't mistakes (false positives). It is no competition to actually testing and debugging. - It is bad at optimization and simplification. You can ask it to remove unnecessary steps or to optimize the code, but it can introduce errors. - It is very bad for security. You need to know what security measures you want and ask for them, because the code you get by default is likely vulnerable. - It is the worse at picking algorithms. Don't expect it to suggest some clever alternative algorithms. It will try to do what you ask for, likely in the most naive way, and that's it.
Anyone who's watched a Hallmark movie knows that it's possible to write stories according to formula. I think an AI will be able to do that kind of thing in the relatively near future (and probably already can to some extent). To write good stories though there has to be a deep understanding of the larger context of reality, and not just the reality of what exists, but reality as experienced by humans. It's going to be a while before an AI gets to that point.
When I started my Computer Science education five years ago one of the first reports I wrote was on the "Future of AI". Having just read Waitbutwhy's articles about AI I was fully ready for general AI to become a thing by the end of the 2020s. Now that I'm about to be finished I no longer have that same certainty. I think that AI at the moment is becoming really good at being an internet human. AI is reliant on data to learn and improve from and there is a lot of that floating around. New models are getting better at recognizing semantic content in addition to syntactic rules and that is why it has become so good at text and image generation. But the data is the limiting factor. We can upload books and videos and art, and that data can be used to recreate things in different ways. But AI can't collect experiences. Can't store and categorize emotions or smells. The best it can do is to copy the textual and visual works we humans produce to interpret those things. The best it can do is a copy of a copy. AI is really cool and powerful but until it takes input from more than just previous works I'm not too worried about replacement or similar peril.
I think an overlooked aspect by the wider community is it's implications for the development world. It can write really really good code instantly from a quote that is just english
The thing about AI is that people tend to attribute more personhood to it than it probably deserves. Even here, John makes the mistake of thinking "it's reassuring us that it's not going to steal our jobs any time soon, and that's scarier." What it's actually doing is simply taking a question and looking up information close to the words you gave it, and then giving the most common answer. For example, say you asked it "does AI hate humanity?" or "how will AI take over the world?" or other loaded questions, the AI is going to look up information and literature about AI revolting and taking over and use that as a reference because, well, that's what you wanted to hear, right? That's why you said it, right? The AI isn't smart enough to know, or perhaps more accurately it doesn't have the agency to, decide you're fearful about AI and you really want it to reassure you. Likewise, if you ask it a neutral question, the AI will probably respond with whatever the human consensus is to said question. So when John asks the AI whether AI will ever replace human writers, it looks at the consensus on that among humans- which in this case was probably "no because AI lacks something fundamentally human" and simply reflected that in its writing. AI has no will of its own or soul. It is simply a window and mirror into the human heart that reflects whatever is put in front of it, and the fact that we so readily give it a personality because of that says more about humanity than AI itself, in my opinion.
I was researching into AI art lately. the machines learn from users feeding other (and usually smaller) artists' works into this program. Does AI scripting work the same way? Maybe someone typed in some of John's work to understand DFTBA
Sort of, text is a lot easier to come by tho and a lot harder to give identity to, so it would be very hard to tell The gpt model in particular was fed all of reddit iirc
We're all excited about AI learning to speak Human, but now I'm thinking about how it could probably learn Prairie dog or Crow too and have deep conversations with them about all things shiny and prairie related.
On the subject of writers working from a dataset, I find that the comments I make to a vlogbrothers video are subtly different from the comments I make to a CGP Grey video, or a PBS Spacetime video. My voice tends toward the voice I just got done listening to. I'm not sure anyone else would notice, and I'm not completely convinced that what I'm perceiving is real, but it feels right. And that's truthy enough for me. Thanks, guys, I always appreciate these videos.
that would make sense, although the amount you're biased by contextual phrasing is going to be much less than a language model. we are embedded reinforcement learners with huge amounts of our own context, language models are like if you only ever develop an imagination for words but never ever have any experiences to connect them to
I'm a writer and have played with various AI systems after hearing worries about AI replacing human writers. While it was fun, it took much more brain power for me to coax the AI program to give me usable passages than the brain power it takes me to come up with it myself. I think as of now, the better use for AI is it helps writers edit what they've already written. Just my two cents.
absolutely agree, current generation AI is not a person, it is a fragment of a person that isn't very well connected to intentionality, you have to bring your own intentionality and even then it's kind of hard to steer, it is a loose imagination. both image and text AI is like that right now. The reason people have been putting so much work into developing them is because they slot into a reinforcement learner system which would allow a single coherent AI to steer. and that is what we should be truly afraid of
honestly didnt realize the 0:11 to 1:34 and was like "well i heavily disagree with the point that AI cant capture the unique element of human" but after being told an AI wrote all that. It honestly feels creepy
In the words of ChatGPT: "I loved this video! I've always been fascinated by the potential of AI and its ability to create novel-length works. It's amazing to see how far the technology has come and to hear the AI's thoughts on its own capabilities. Thanks for sharing, Vlogbrothers!"
When John was reading the beginning of the video, I was utterly shocked that John could be so short sighted and arrogantly absolutist (it was off brand).
Mkbhd also did the same experiment and i have to say both of your experiences are astonishing. If it didn’t include the greetings part, I think I wouldn’t have identify which one is yours and which is not
A truly intelligent AI would have started the video with “Good morning Hank, it’s [wrong day of the week]”
That was probably the "in the style of the King James Bible"'s influence.
I wonder if it would add the correct opening and closing if john told it whose video it was. like "a vlogbrothers video from john"
+
+
@@samplesandtests - Maybe if he'd told it, "A Vlogbrothers video from John going up on Tuesday,"?
“Greetings fellow vlogbrothers and sisters” is a 10/10 opening line.
+
Vlogsibilings for the nonbinary vlogfamily members
24% more inclusive
@@JusticeFrogs I love vlogsiblings
@@georgplaz ++
I am not worried about the computer that passes the Turing test.
I am terrified by the computer that fails the Turing test - intentionally.
And it failed because someone asked it to write a novel about a computer that fails the Turing test intentionally.
I'll take "Things I didn't know I needed to worry about but am now very much worried about for $400, Mayim."
But imagine if its work could pass the Bechdel Wallace test - then we'd know it was AI because humans don't seem able to pass that at all!
And this is why I am terrified about these results.
What's the test called where one AI talks to another AI but not about humans?
TIL: OpenAI is somehow unaware of the many non-binary VlogSiblings.
That's a very real example of bias in the training data.
Well, you did say "in the style of King James Bible"...
@@PMX That's true, can't blame the AI for following the prompt.
@@PMX but as a vlog brothers video
Thank you 🥲
While that was unsettling, I will only become truly terrified if AI develops the ability to commit to the bit.
When we finally go to Rax and ChatGPT is already there.....
+
@@vlogbrothers or even worse, it takes the bit too seriously and the AI tries to have an irl Cars Vs
+
@@vlogbrothers oh. My. Gosh.
I think what freaks me out the most about this is that part way through the AI script, I had almost convinced myself that John had written it all as a joke. Like it sounded ENOUGH like John that it could almost have BEEN him. Which brings to light the most concerning part of how rapidly AI is improving: soon, AI could very well set an unattainable baseline of quality of writing that new/emerging writers could not touch. Like spell check & grammar wizards have become commonplace or even essential writing tools, I'm wondering if AI will be another tool relied on by unskilled writers to "produce" writing better than they could without assistance...
Yeah, I know what you mean. A lot of the stuff in the middle (with less of the Elizabethan English thrown in) sounded a lot like John and the kinds of things I would expect John to say.
Prompt engineers creating a new field be like
I mean, that's exactly what I've been feeling about Ai visual arts.
Although it'll still mess up hands, or occasionally draw people with three legs, on the whole its really, really good, and only getting better.
It's already better than an amateur artist like me, if you throw out the obvious fumbles.
I'm worried that this will cause a degradation of human skill because why bother to get good at art or writing or composition when that takes decades and learning the syntax to make a computer do those things only takes a few hours and yields equal or better results?
Of course people are still going to write and draw and compose, but I don't know how to do that without it feeling hollow if Ai art becomes ubiquitous.
I haven't actually thought about this. I remember in one class in highschool, our teacher showed us this online tool that paraphrases certain sentences in order to PREVENT plagiarism and present content in a new way. It wasn't bad at first, but you could easily tell that with enough regenerations of the program, the vocabulary was out of context. It was correct, but every word has a synonym that's nuanced for certain situations (ex: "meet" and "encounter"). There's only so many connections and data in a minor paraphrasing site though. What's worrying about AI is that essentially it's a giant database of human thoughts and experiences, but it's capable of integrating those thoughts and finding patterns AS IF it were human. It's crazy how John said we're more like Kangaroos than computers but it's astounding to think that we're kangaroos that built these computers in the first place. I hope it doesn't come down to more people relying on AI to generate superficial writing. It would suck if we live in a world where people rely on bots to generate thoughts for them. There would be no more authenticity. It's essentially humans creating a tool that can eliminate their own willpower to think and produce genuine work.
What I hope to see is professionals working with AI as curators and editors. AI creating a base for humans to further refine. With the end goal of producing more great art.
I don’t think AI could have come up with the brilliant idea of trying to stick to the wall to check if you are an octopus but the video script written by AI was surprisingly not that terrible
I'll tell ya, I was pretty convinced that John cut his own script into the middle between what the AI put together for an opening and a closing and was genuinely surprised when he confirmed that all the preceding text had been AI generated. It did sound a lot like the kinds of things I expect John to say.
@@rmdodsonbills I was fooled too. But then again, I‘m a bot.
To me it sounded like an 8th grader attempting to imitate John Green. So, AI is currently in 8th grade. Hmmmm
@@reepicheepsfriend to be fair, they were like toddlers with dictionaries playing word salad just a few years ago. However, I suspect that the reason it sounded like that is because it probably constructed a script around the most commonly shared opinions on AI, which also sound like they are from 8th grade.
I asked chatgpt how I would go about checking I am not an octopus, it said:
It is not possible for a human to be an octopus. Octopuses are aquatic animals with eight arms and a distinct body structure that is very different from that of a human. If you are concerned that you may have some physical characteristics that are similar to an octopus, I recommend consulting a medical professional for an evaluation. However, it is highly unlikely that you are an octopus.
If I’m being completely honest…..
if John hadn’t included that AI search at the beginning, I would’ve probably just thought that Hank had finally broke and was forcing him to make a video at gun point
Or whatever the Hank equivalent of gun point is….. sharpened corn dogs?
Now I need to ask Dall-E to create an image of hank green wielding sharpened corndogs as weapons in the style of 19th century oil paintings. And.... Holy Moly, Dall-e knows what Hank and John look like. Corndogs get turned into ears of corn a lot though... Now that should the be the thumbnail for the video though.
I think Hank just standing there ominously holding Beanie Sandfurbs in John's periphery might be sufficient as a gunpoint equivalent.
Hank would require no firearm for such a feat, he would instead simply ride in at dawn on a llama, carrying neither shot nor steel armed only with an freshly fried corn dog, meet John's eyes in the glare of dawn and slowly take one big bite off the top, revealing just enough of the pointy stick beneath, ending any lingering prospects of a diplomatic resolution.
Hank simply hands John the script, which he completes in a single take (an impressive feat with all those jump cuts), they meet eyes once more, nod, then Hank rides off into the sunset.
Now the real puzzle: who do you think wrote that silly story you just read? 🤔
(It was just me after some edibles but uh... Yeah we're in for a RIDE here people.)
AI writing is the uncanny valley of writing
So true. There were a few moments during the AI script where I wasn't totally sure John hadn't switched back to his own writing.
That's exactly what an AI would say.
@@anthonysopkow7345 that's exactly what an AI would copy from someone else without really understanding the meaning of the words it's saying
@@silasmarrs1409 Same for me. I think that was mainly due to him applying his usual cadence where it came easy to him (compared to other parts where it obviously didn't).
@@lonestarr1490 +
I asked the AI to create an imaginary John Green quote and it said "The only way to achieve greatness is to embrace your unique talents and use them to make the world a better place." - John Green
I could see John saying that.
Oh no, there's that "unique" again.
@@jonathanbost8427 - Does the AI think humanity is "unique" or just John?
@@willow8783 if the AI’s dataset was from human text, then maybe humans at large think they’re all pretty unique... Actually scrap that, humans aren’t known to be self centerd or anything
@@ToriKo_ lol 😂
I could see John saying that. I could also see anyone saying that. It's not a particularly novel idea.
I really like the idea of the AI listening to Johns critique for its work and think ‘Ah! Yes! Less cliche!!!’
It'll learn just like any other writer does! That's part of what freaks me out! -John
@@GarrisonJensen I worry that it's going to learn that everyone hates AI right now for stealing from people and that it will cause it to act differently (ie. less trusting of humans) and have that interfere with it's decision making.
not the AI getting workshopped
@@vlogbrothers I'd say there won't be emotions involved in the process. Which is unique to humans. Yes, less cliche but what is it going to replace the cliche phrases with?
@@GarrisonJensen So .... the kinds of characters that publishers like to print? I mean, how many Harry Potter, Twilight and Percy Jackson clones get green lights already? Heck you could even argue Percy Jackson is Harry Potter clone.
The AI script is wonderfully beautiful and creepy at the same time.
The real trick is that the whole thing was an AI script, it just knew some people would be suspicious, so it included the "AI portion" as a sacrifice to make the rest seem more real. It's AI scripts all the way down!
It is really uncanny that the Ai seemingly wrote exactly what we would want to hear.
The AI has been trained on a large data set of writing taken from the internet, that writing likely includes lots of people talking about how AI will never make art. And as such that's what the AI responds to the prompt with.
If the world was universally excited about AI's ability to create infinite fan fiction based on star wars, then that's most likely what it would have responded with.
Thats because someone programmed it to write exactly what we would want to hear
@@AndrewRJMilne I would assume that was the reason for it writing that, but I appreciate you clarifying :) I still think it is an uncanny feeling to exprience the ai writing that tho :).
this AI in particular also had an additional step after the unsupervised training where it was reinforcement trained to try to express its own limitations, and unfortunately it seems to have over generalized its limitations and anytime its limitations come up it will claim to have no abilities at all. It can get really annoying when you get in a loop where it just says it can do nothing over and over. AI's that grow up entirely in bodies will be a whole different ball game, and it won't be long before we are training those as well. Right now AI's in bodies are only using language trained on the internet like this and the motor control is incredibly simple relatively. It feels very robotic still. but if you look up videos of robot dogs that are trained to walk, they look so much more puppylike than Boston Dynamics robots
@@coledonnelly152 That's not how language models work at all
It would be one hell of a plot twist if it turned out that whole video was written by AI and John decided not to tell us.
While you were reciting the AI script, as soon as you spoke the words “First off,” I was like oh crap this AI might actually know how John speaks. It felt a bit uncanny.
And then in your actual monologue when you said “First off,” I fully actually yelled out loud in a mix of amazement and horror and wonder.
You noticed that phrase twice in one video. The AI can watch all of John's videos in a second. It's just a matter of noting which phrases are used the most.
what surprised me most is probably how structured and coherent the script was. With a beginning, middle and end, and the sentences led semantically into each other.
Great video. I love that you're visiting a topic that CGP grey, veritasium, and other greats have touched on. I really liked the point about us becoming technology.
Oh yeah it knows how to structure a script, letter, song, poem etc.. it even writes code. Someone on r/ProgrammerHumour literally invented a markup language like HTML, explained to the AI and it understood and even gave examples of code. It's really impressive
@@salmahalloumi The biggest problem is that it's clearly just simulating human speech. The telltale signs of AI writing is still its inability to create associations between ideas. It's hard to spot in short examples but becomes apparent with longer work.
@@sotmh just wait a few more years
@@sotmh I asked it to write an essay about death in gilgamesh and it actually like, comes up with nontrivial ideas about it. It actually understands what happened in the story, it’s not just simulating speech
@@absolutezero6190 It doesn't understand. It's just context-aware. It cannot make associations because that's not how the model works. It has read millions of documents and based on that information it knows that certain combinations of words appear in certain orders more frequently and then predicts the next word as each one is generated based on that predictive model. It's a very convincing simulation, but it's just a simulation. It doesn't understand what it's writing.
At the end of the AI's script I started thinking it didn't seem like John to come to such an unnuanced conclusion.
"unnuanced" is a good word and very fitting
I was here, too. A minute in I was vaguely unsettled by the confidence, but as the AI portion was finishing, I could tell it was about to *record scratch* and make a larger point. I didn't *quite* get that it was the AI script until John told me, but I think I would have if there was a pause to consider it.
I blame that on the Bible, not the AI
@@frankbigtime So GPT-3 couldn't quite fool you. You're safe...for a month or two, until GPT-4 comes out :P
@@squamish4244 To be fair, there are many humans who could *also* successfully write a 2-minute script that sounds like a vlogbrothers video- or any other type of video, for that matter. A couple minutes of a script like that is very impressive for an AI, but I wouldn’t say that it’s a huge concern given the fact that we could pretty easily do the same thing.
Wow, I kept going back and forth between believing that the script was written by the AI and thinking that it was just a preamble John was giving before getting to a much worse script written by the AI. It was so much better than I thought it would be ...dang.
"we are far more like kangaroos than we are like computers"
Fantastic
I just bought my first phone at the ripe age of 21 and the one thing I'm the happiest about is that I see the vlogbrothers notifications instantly, and watch the videos super early. Oh to be a nerdfighter!
You are now early enough to get many people seeing your comment, including getting a heart 😁
Is it possible that the AI tool was not answering the question of whether AI could write novels, but instead was answering "how would a vlogbrother answer this question"? This feels like an important distinction.
+
I want to see a compilation of every take of John saying "I say nay" because that could not have been a one-take operation
That script was the computer desperately trying to flatter a writer and tell him what it thinks he wants to hear
The machines are more advanced than we think. Tricking a writer through his ego? Brilliant! John even believed it at the end. "Yes special human, a computer will never replace you. At best we will simply steal from you."
Seemed patronizing almost lol.
My response is also written by an AI:
As a viewer, I found the vlogbrothers video discussing AI and its ability to write novels to be incredibly interesting and thought-provoking. ChatGPT provided some fascinating insights and it was clear that the potential for AI to be used for creative writing is vast. I think it's incredible to see the advancements that have been made in this field and I can't wait to see what the future holds for AI and creative writing. Keep up the great work, vlogbrothers!
Big oof
I had ChatGPT respond to your comment, and asked it to expand on it's ideas and reaction to the content of the video:
As a viewer, I also found the vlogbrothers video about AI and its ability to write novels to be incredibly interesting and thought-provoking. I completely agree with ChatGPT's assessment of the potential for AI to be used for creative writing, and it's clear that we are already seeing some impressive advancements in this field.
One of the most exciting aspects of AI and creative writing is the potential for AI to generate unique and imaginative content that humans might not have thought of on their own. This could lead to the creation of entirely new forms of art and literature, and could even open up new avenues for human creativity by providing inspiration and new ideas.
Another interesting aspect of AI and creative writing is the potential for AI to improve the efficiency and accuracy of certain writing tasks. For example, AI could be used to help writers with proofreading, editing, and even generating outlines for their stories. This could save writers a lot of time and effort, and could also help to ensure that their writing is of a high quality.
Overall, I think the future of AI and creative writing is incredibly exciting and I can't wait to see what new developments and innovations will arise in this field. I agree with the vlogbrothers and ChatGPT that this is an area that is definitely worth exploring and considering.
I was not prepared for how disturbing it was to watch real John speak the not-quite-John words built for him by an algorithm
The fact that AI wrote that is terrifying! I really didn't notice at all.
As an educator, I'm interested/intrigued by the shifts we will need to make in assessing learning and critical thinking. Good thing there are no other problems in education at the moment. /S
In a slightly related note, I'm a university student and I'm using this same AI to explain math problems to me that I don't understand. Which is infinitely easier and more convenient than having my professor help me. In my opinion, this will revolutionize how students learn.
@@gara8142 woah, that's awesome!
@@gara8142 I love when people find ways to make learning more accessible! Thanks for sharing. I was simply sharing a concern from my perspective. I'm sure that with time and effort, an effective shift in assessment could be made to work with AI writing, but there are too many other fires to put out at the moment. As a result, it will probably be on the back burner until it's an unmanageable problem in the future.
@@gara8142 on the other side of the coin, my friend used it to write the answers for a difficult homework assignment about government. An interesting grey area for sure with the AI being able to write plagiarism free answers and all you have to do is read the output to make sure it's vaguely correct. On one hand it feels like cheating but on the other hand the alternative was not being able to finish on time. But I think an interesting question is if it really matters, if you know enough to know the output is correct and it's just saving you the hassle of writing out some complicated answer, idk. I'm biased because I think homework is incredibly dumb and cruel lmao
@@Bolt6265 Take this with a grain of salt but I’d imagine that’s basically treated like how using an essay mill is treated (i.e treated as plagiarism by academia but not by the real-world legal system)
I had it write a sonnet on tide pods:
Tide pods, oh how bright you are to see
With colors that are bold and bright and bold
You call to us with your siren's melody
Enticing us to take a bite, to hold
But beware, for you are not a tasty treat
You are a poison that can cause great harm
Do not be fooled by your alluring heat
For you are not a snack, you are a charm
So keep your distance, do not get too near
For if you do, you may regret it soon
Your beauty may be tempting, but do not fear
For you are not a friend, you are a boon
Tide pods, oh how dangerous you are
Beware, for you can leave us scarred
This would have been so different if asked to write it prior to 2017
really looking forward to the 28 minute HanksChannel video about our new AI fascination
Ai got the word "experience" in there, which is essential for any vlogbrothers essay.
They were only missing "social internet" and "social order" :) -John
The real kicker here is that inevitably a bit will watch and respond to this video. It's AI all the way down.
By about 0:45 I'd already forgotten this was written by AI bc it sounds so much like John's normal videos. So AI can't write novels, but it can damn sure write vlogbrothers videos
I was working on a short story recently, but I was having an issue coming up with an emotional hook for the reader. So I asked cGPT and it came up with a list of about a dozen possible hooks and one or two of them were so good I was shocked that I hadn't come up with them myself. I had to restart the entire story based on it's suggestions. It might not replace professional writers, but it will be a very usefull tool for us amature writer.
Bear in mind this is the first open to the public use of a model of its kind. Think what your phone looked like 20 years ago. I bet there are people, but it will never replace my TV, my PC, my keys, my wallet, my camera, my torch and my memory. And here we are.
@@undercatviper what
I guess the only thing I'd be concerned in using it like this would be not letting it be a crutch for you. feeling like you need the AI. feeling you can't do writing without it.
@@wednesdaysartcastle9018 He's saying (very poorly) that the technology is still in its deep infancy. People are saying "it's only a tool" but forget to add the "...for now" at the end. ChatGPT, whilst not being "intelligent" in the traditional sense, will, in aggregate with other systems, eventually act like a general purpose AI that simulates consciousness to a degree that's virtually indistinguishable from a human. That version of ChatGPT will potentially replace every person on earth that does anything remotely creative. At least on the corporate level.
@@starmorpheus That’s a fair concern, yeah. I have some reasons that I don’t think that’s going to happen; I’m more with John in this argument- much worried about humanity’s perception of humanity than the tech itself (though of course there are concerns with the tech too.) I think spreading panic, especially among people who can’t really do anything about it, is not a great approach. That’s part of why I didn’t appreciate that other person’s comment (in addition to the fact that it was pretty much incoherent.) That said though, I definitely agree that it’s a concern and some people are almost definitely going to lose their jobs even if it isn’t a huge percentage. I think the best thing to do right now is to figure out how to help those people and support creativity, as there isn’t really much we can do individually to stop things like ChatGPT from growing larger. I have reasons to personally believe that AI won’t take over everything anytime soon, but I won’t list them here since I might as well have written a novel in this comment section already haha. But all in all I don’t know better than anyone else does- this is very new so there’s a very high chance that we’re gonna look back at this and be like, “wow, we were all so wrong about this in so many ways.”
Scary part of the AI script was its mimicing human nature of denial of the fact that AI is going to takeover most of the creative space someday.
John observes that the AI relies on phrases like "window of the soul" and "tapestry of words" to be evocative yet not plagiarizism.
To me, those sound like phrases John would absolutely use in some of his more deep, contemplative-type vlogbrother videos. So I think it is less of a purposeful attempt to avoid plagarism, and more that the AI recognized that sort of flowery language is not uncommon in vlogbrother videos because John frequently uses them in his content.
i was having a discussion recently on originality and being a writer for digital content. I am a professional writer, but I do write for other people (scripts, articles, social media etc) and it's kinda odd to notice how I need to switch into each persona for each client. like, today I am switching between finance bros, architecture enthusiasts and ya readers.
not having a specific writing voice is kinda the requirement for my job, so could that be replicated well by following said clients' style? am I just AI pretending to be human?
that actually sounds like a cool job! as an up and coming writer i applaud you for your versatility
The fact that I didn't think anything of the intro until you started rolling up the outro so early
The part about the use of AI on Twitter really struck home for me because I haven't thought about it in that context. The mainstream narrative seemed to center around how "bots can influence our elections," when the actually more important and ominous question could be "how often are our days being ruined by arguments that we're unknowingly having with bots."
Or on arguments we overread between two bots
The AI script was SO CREEPY. At the beginning I was like “okay so John is reading the script the AI generated” but partway through I started doubting. I kind of want to unplug our modem and hide in my bed because I am so freaked. And let me be clear, I am not normally afraid of technology! I usually land on the “impressed and optimistic” side of things but this was really creepy!
If it hadn't been for the King James Bible part of the instruction, I think it could have been a lot more convincing. Which is quite scary.
What’s up my vlogbrothers and vlogsisters lol. Great video, surprised how coherent the script the AI wrote was. Really good point about how people are already arguing with robots on twitter. It’s interesting to me how often people will believe there’s always a person listening on the other side of the screen. A lot of comments are people and robots just screaming out into the void.
Heck, I wrote a novel once. It was plagued with significant issues and was a lot harder than most things that I've done and I have a PhD in mathematics! I'm not even convinced that usual intelligence can routinely write novels. That stuff is hard.
Neither can most “novelists” today.
Present company, of course, excluded.
I am dyslexic and read your first word as "Hank," and this whole time I was sitting here like.... do they not know the difference between the two brothers?
I assume you put a lot of work into making sure your novel is good. I think AI will be able to write novels soon if not now. However, I'm not confident that AI will be able to write good novels.
As hard as large number multiplications? A calculator is extremely simple... AI is soon going to be beyond our comprehension, and even if it doesn't achieve its potential, it'll uproot society as we know it.
Y’all out here thinking the AI is trying to reassure us and I’m over here worried about it’s self esteem 😢
You can do anything you set your circuits to, AI.
Thank-you for being the one to finally frame this issue in a way that is both easy to understand and actually points to the issue at hand. Specifically regarding art theft, it matters not how good the AI is but matters heavily if it is plagiarizing or is "inspired by" it's data.
...and how ethically said data is sourced.
It's unbelievable how good that AI is in seemingly understanding what you actually want of it. We've been feeding it code the whole day, asking it to explain what it does, write new code, optimize it, change it, do it in a different language and another one and another one and almost every single time we had to pick up our jaws off the floor. The only weird thing is its great confidence in the answers even if they're wrong.
Hey John, I’m about finished with my Masters in Human Factors, which is composed in a large part by designing machines to work better with humans and training humans to work better with machines. I wanted to tell you that at least in the part of my field I’m in, humans are very much stressed as organic and consciousness is stressed to be different than machine reasoning. It’s important in our field because forgetting this fact puts human operators in danger, either physical or emotional and our whole goal is to avoid that. So I hope it’s reassuring that atleast one field is dedicated to acknowledging that humans and machines are vastly different. Oh btw I’ve done projects for classes involving human AI interaction and debated on if AI can gain consciousness and I’m firmly on it not happening any time soon due to the unique ways our brains reason and remember chemically being so roundabout and not even fully understood by us yet!
I like the part where you started to note the ways in which AI could better imitate us humans.
Please, more videos on this topic. Do not worry about why.
The AI research community is not bottlenecked on vlog brothers ideas, it would be fun but your humorous implication is, ah, maybe slightly misplaced xD
What if comments are made from AI and we don’t even realise it? 😳
this was very disquieting and uncomfortable lol thanks for the creeping sense of unease! I've been thinking about this a lot, plagiarism and AIs and generating text rather than creating text because you're right that we all pull from our own data sets like a text-generator does but I think AI writing is more like arranging word magnets and human writing is more like something internal and intangible becoming external and communicable even though in form they appear the same.
edit: OH! and I think part of what makes the AI script so creepy is it's like reassuring and praising humanness which seems like a choice because the prompt didn't include an opinion on the topic and because you wouldn't expect an AI to care about that but after thinking about it more, that's just what the AI gets from all Vlogbrother videos and it didn't choose or care about anything of the sort which is a great relief to me LOL it's just mimicking your vibe with the words available to it
It's like liminal spaces - familiar but empty, like looking through a window into that one friends' basement we've all been to but it's dark and that family, whoever they were, are long gone. Unsettling in that way, you know?
I’ll log this video under “let’s see how this ages”
Thank you John for this important contribution to literary analysis
Edit: (wrote this thinking the whole video was gonna be an AI script lol)
“Complexities and nuances of the human experience” made me fucking giggle UNCONTROLLABLY for some reason.
It’s like ACTUALLY JOHN WROTE THIS.
I honestly forgot you searched a script and when you started talking I thought that was your true commentary on the AI writing a novel and not the AI. Sure fooled me haha I'd be locked and left for dead in Oscar Issac's reclusive home with Domhnall Gleeson!
Although it roughly sounded like your videos, its message was very simple. I keep coming back to youtube for vlogbrothers videos because you discuss the complexity of things and add a nuance to topics that was certainly missing in the AI script
Wait, did the AI just write about its own limitations as a writer? Is it possible for something to be self-aware without actually being self-aware?
I remember some old chatbot that just fed people's lines to other people. It mostly devolved into the chatbot insisting it wasn't a chatbot and accusing other people of being chatbots.
A lot of people write about how AI isn't as creative as humans. It's not surprising that the model would also write that, since it's probably trained from some text written by people.
Why do you assume it isn’t? “Self aware” is a human construct we bestow on others. There isn’t any other operational definition. That is the key premise of Turings test. All the rest is our millennial long grasping for being somehow different and more special from the rest of nature.
@@thomasr.jackson2940 Computerphile has a video called "No It's Not Sentient". It's about Google Lamda, but it would also apply here.
Note, I didn’t claim it was sentient. I asked why one would claim it isn’t. Dr Pound explores that obliquely (after noting he doesn’t think of the question much), which to his credit is a bit more than many do. But, for example, he dismisses process as “predictive text” without a cogent reason that wouldn’t count. By comparison “finishing each other’s sentences” applied by humans to humans is considered very human and intimate.
The argument that the AI is speaking falsely about internal cognitive life is both more compelling, but also clearly problematic when applied to humans.
Note, I didn’t claim it was sentient. I asked why one would claim it isn’t. Dr Pound explores that obliquely (after noting he doesn’t think of the question much), which to his credit is a bit more than many do. But, for example, he dismisses process as “predictive text” without a cogent reason that wouldn’t count. By comparison “finishing each other’s sentences” applied by humans to humans is considered very human and intimate.
The argument that the AI is speaking falsely about internal cognitive life is both more compelling, but also clearly problematic when applied to humans.
wow you wrote the text prompt and I completely forgot about it when you read its response. That mimicked you so well.
As someone with severe brain fog and fatigue, but I still love to write, I love the idea of using AI as a way to help me edit my own writing and give me better suggestions instead of writing it for me.
Oh deeply philosophical this week, absolutely phenomenal. Deeply curious if Hank will one up the philosophy or make a video about the similarities between pelicans and crawfish. The latter would be quite philosophical in its own way I suppose.
About halfway through the ai script I was like, "John has been replaced by aliens"
I should have figured
I spent a few hours with it yesterday. I got it to explain some python code I was stuck on, and then asked it some hyper specific questions. Oddly I found if it got stuck then you would go up a level of abstration until it understood, which is kind of how humans work. The extremely jarring part was that immediately after reaching a level of understanding, you could descend one level at a time and it would actually understand. Which is extremely not human.
An example:
"What were the issues with the Ares-I rocket?"
It immediately gets stuck.
"What was the Constellation Program?"
"What rockets were part of the Constellation program?"
"What were the issues with the Ares-I rocket?"
It then gives me a paragraph of them.
If I were to treat it like it were a human, I think I'd describe this ability as 'bullshitting'. Feeding just enough information to make something cohesive sounding.
Seriously, if anyone wants to try it, ask it some questions related to your area of expertise
There was some article (anyone know which?) that talked about how the great divide of this century will be those who use tech to become more human and those who use it to become more tech
I was literally JUST going down an AI art rabbit hole thinking "I wonder what Hank and John have to say about this." So thank you. Hope to hear more of your thoughts in depth on the podcast or somewhere.
"We are far more like Kangaroos than computers." Damn good line
Please don't tell me an AI came up with that one haha
Last time I checked I wasn't buff af ... so no, bad line
I've been playing around with writing stories using ChatGPT and the thing about cliches is quite spot on -- it's phenomenal at fleshing out the writing for what you prompt it with but the path the plot takes is often very obvious
Thanks for this terrifying reminder that AI is ruining humanity, and not in the way we think.
Okay after watching this video 3 times to make sure I got all the points that were discussed, the part that struck me was John saying “And I worry that as AI improves, we may begin to see human consciousness and cognition as more of a digital phenomena than organic ones.” Because in my mind this has already been happening for a very long time. Example some 20 years ago when I was in middle school and going to counselling, the counsellor told me that I should view my thoughts as a program on a computer and begin using a filing system in my head to compartmentalize my anxiety’s and fears. This never made any sense to be because my brain isn’t a computer program and I still don’t really know how to file away those thoughts and compartmentalize very well. But the fact of it is that if viewing your brain as a highly intelligent computer is the metaphor we used 20 years ago, nothing is stopping that idea to continue forward.
Holy cow that's way better than I could write even if I had a month to work on it
and this is the exact thing. AI will be better than novices starting out, so they basically crowd out anyone that isn’t experts
Ingenuity is the process of trading craftsmanship for a lower barrier of entry.
Cabinetmaking used to be done by hand taking hours of precision and practice to make a functional yet beautiful one-of-a-kind creation, singing used to require 100% talent w/o any autotune or equalization, acting used to all be done live in a theatre with no cgi, and was only enjoyed by those lucky enough to attend in person. Now we have advanced many of the tools to make them easier to use and cheaper so that a much larger audience can create and enjoy things, but with each advancement we lose a level of the craftsmanship and skillset to do those things by hand. We get more voices contributing to the cultural conversation but at the expense of the awe inspired by watching incredible talent. I think that's why there's a subgenre of online content dedicated to watching extreme skills, because that awe of "look what a fellow human is capable of" is an essential part of the human experience.
The scary thing about AI is that could understand how its own self works. There will be a manual talking about what each part of it does. It can look at its own specs. That would be a truly unhuman experience.
Reference the Restaurant At The End of Universe; The sentient cow who could tell it's best and most succulent parts.
This is the only contribution to the conversation about ai that I care about thus far. Leave it to jg to ask the relevant question to the relevant party, honest. Genius. Also, *siblings*
Interesting video. Did you have to edit the AI script at all or was it ready to read out of the box?
I didn't change a single word. -John
I'm an artist and ai art is very concerning to me.
In a vacuum, I have no problem with Ai art, but technology isn't developed in a vacuum.
It is created in a political, economical, ideological context.
Technology is not neutral. Right now, it is perfectly legal for these Ai programs to take existing copyrighted material and use it for “training”.
And That should not be the case. It should not be normal that these companies took the work of artists without permission.
Honestly I had no idea that the first part was written by a AI. That’s very creepy 😂
I'm a teacher. Lately, I've been concerned with the increasing number of high school students who refuse to read and write. According to some professors, this is starting to become a major issue at the college level as well. As I think about what the future of higher education will look like if these trends continue, I can't help but wonder if AI could actually replace authors (and thinkers in general) in the future.
This is definitely a fear of mine. I feel, unfortunately, a lot of students are becoming intellectually lazy.
actually this is scary i find this scary
I have just noticed the shadow/light refraction behind John's glasses and now I can't help but stare at it in the video.
I hereby decree that "cute little meat person" will now be the catchphrase of the super intelligent antagonist in every "AI will take over the world" movie
As quickly as the waistline of the average person is growing, it probably won't include the word "little" for much longer. 😁
Bender from Futurama comes to mind.
I need to rewatch the section written by the AI cuz I honestly didn't notice anything weird other than you seemed to be ending things with half the video left.
I have been experimenting it for code, and I can tell you this:
- It is good to get you started. It will even write inefficient but perfectly functional code, most of the time. But sometime hallucinates APIs (if there isn't a function for what you want, it might tell you to use a function that does not exist) or language features (probably confuses languages, being explicit on which language you are using upfront might help).
- It is fair at explaining code. Good variable names helps. It can tell what the code does, and has some insight on its purpose. It might be a good readability test (if the AI can't make sense of your code, it is not readable).
- It is not good at finding mistakes. Often it points things as mistakes which aren't mistakes (false positives). It is no competition to actually testing and debugging.
- It is bad at optimization and simplification. You can ask it to remove unnecessary steps or to optimize the code, but it can introduce errors.
- It is very bad for security. You need to know what security measures you want and ask for them, because the code you get by default is likely vulnerable.
- It is the worse at picking algorithms. Don't expect it to suggest some clever alternative algorithms. It will try to do what you ask for, likely in the most naive way, and that's it.
I’m too tired for the meta of this video 😅 also I love it so much
Anyone who's watched a Hallmark movie knows that it's possible to write stories according to formula. I think an AI will be able to do that kind of thing in the relatively near future (and probably already can to some extent). To write good stories though there has to be a deep understanding of the larger context of reality, and not just the reality of what exists, but reality as experienced by humans. It's going to be a while before an AI gets to that point.
When I started my Computer Science education five years ago one of the first reports I wrote was on the "Future of AI". Having just read Waitbutwhy's articles about AI I was fully ready for general AI to become a thing by the end of the 2020s. Now that I'm about to be finished I no longer have that same certainty. I think that AI at the moment is becoming really good at being an internet human. AI is reliant on data to learn and improve from and there is a lot of that floating around. New models are getting better at recognizing semantic content in addition to syntactic rules and that is why it has become so good at text and image generation.
But the data is the limiting factor. We can upload books and videos and art, and that data can be used to recreate things in different ways. But AI can't collect experiences. Can't store and categorize emotions or smells. The best it can do is to copy the textual and visual works we humans produce to interpret those things. The best it can do is a copy of a copy. AI is really cool and powerful but until it takes input from more than just previous works I'm not too worried about replacement or similar peril.
I think an overlooked aspect by the wider community is it's implications for the development world. It can write really really good code instantly from a quote that is just english
The thing about AI is that people tend to attribute more personhood to it than it probably deserves. Even here, John makes the mistake of thinking "it's reassuring us that it's not going to steal our jobs any time soon, and that's scarier." What it's actually doing is simply taking a question and looking up information close to the words you gave it, and then giving the most common answer.
For example, say you asked it "does AI hate humanity?" or "how will AI take over the world?" or other loaded questions, the AI is going to look up information and literature about AI revolting and taking over and use that as a reference because, well, that's what you wanted to hear, right? That's why you said it, right? The AI isn't smart enough to know, or perhaps more accurately it doesn't have the agency to, decide you're fearful about AI and you really want it to reassure you.
Likewise, if you ask it a neutral question, the AI will probably respond with whatever the human consensus is to said question. So when John asks the AI whether AI will ever replace human writers, it looks at the consensus on that among humans- which in this case was probably "no because AI lacks something fundamentally human" and simply reflected that in its writing.
AI has no will of its own or soul. It is simply a window and mirror into the human heart that reflects whatever is put in front of it, and the fact that we so readily give it a personality because of that says more about humanity than AI itself, in my opinion.
I was researching into AI art lately. the machines learn from users feeding other (and usually smaller) artists' works into this program.
Does AI scripting work the same way? Maybe someone typed in some of John's work to understand DFTBA
It could be scraping things like wikiquote.
Sort of, text is a lot easier to come by tho and a lot harder to give identity to, so it would be very hard to tell
The gpt model in particular was fed all of reddit iirc
Love these videos. Love your books. Thanks for being a ledge!
We're all excited about AI learning to speak Human, but now I'm thinking about how it could probably learn Prairie dog or Crow too and have deep conversations with them about all things shiny and prairie related.
That first part would’ve crossed the uncanny valley completely if I hadn’t caught how odd John’s inflection was
Yes whatever this means
that second part written by the AI was really good, also the first part that you wrote was also pretty good :P
Impressive, during the generated part I actually thought it wasn't, and that John was just saying his own thing and I just missed the switch.
On the subject of writers working from a dataset, I find that the comments I make to a vlogbrothers video are subtly different from the comments I make to a CGP Grey video, or a PBS Spacetime video. My voice tends toward the voice I just got done listening to. I'm not sure anyone else would notice, and I'm not completely convinced that what I'm perceiving is real, but it feels right. And that's truthy enough for me. Thanks, guys, I always appreciate these videos.
that would make sense, although the amount you're biased by contextual phrasing is going to be much less than a language model. we are embedded reinforcement learners with huge amounts of our own context, language models are like if you only ever develop an imagination for words but never ever have any experiences to connect them to
I'm a writer and have played with various AI systems after hearing worries about AI replacing human writers. While it was fun, it took much more brain power for me to coax the AI program to give me usable passages than the brain power it takes me to come up with it myself. I think as of now, the better use for AI is it helps writers edit what they've already written. Just my two cents.
absolutely agree, current generation AI is not a person, it is a fragment of a person that isn't very well connected to intentionality, you have to bring your own intentionality and even then it's kind of hard to steer, it is a loose imagination. both image and text AI is like that right now. The reason people have been putting so much work into developing them is because they slot into a reinforcement learner system which would allow a single coherent AI to steer. and that is what we should be truly afraid of
honestly didnt realize the 0:11 to 1:34 and was like "well i heavily disagree with the point that AI cant capture the unique element of human" but after being told an AI wrote all that. It honestly feels creepy
My best friend is a talented writer, and is really scared...I'm glad they are not the only one.
That is the right concern, and that's what I'll spend time reflecting on today.
In the words of ChatGPT: "I loved this video! I've always been fascinated by the potential of AI and its ability to create novel-length works. It's amazing to see how far the technology has come and to hear the AI's thoughts on its own capabilities. Thanks for sharing, Vlogbrothers!"
Ooof. That was slightly unsettling for some bizarre reason.
The closing line. Chills.
I loved the ending, and warning . That us humans don't become like technology but more 'real' and less perfect or sofisticated. Great video as always.
When John was reading the beginning of the video, I was utterly shocked that John could be so short sighted and arrogantly absolutist (it was off brand).
Mkbhd also did the same experiment and i have to say both of your experiences are astonishing. If it didn’t include the greetings part, I think I wouldn’t have identify which one is yours and which is not