Wes, this is one of the best videos I have seen this year. Please please please DO NOT make the shorter. You insights are incredible. I wish I could get more of you. You truly bridge that divide between the average person and the super geek
Fuck OpenAI. I'm tired of the rich get richer and poor get poorer, which we're all currently rushing towards with absolute joy. I'll release my own actual, proper AI assistant shortly as I'm tried of everything coming out being a Chat GPT wrapper.
Historically, societies that freed themselves from the burden of hard labor tended towards complete militarization and population decline. Early in Spartan history, they enslaved neighboring city states and were a completely militant society for over 1000 years. The Roman empire had 7 slaves for every 1 citizen at their military peak (and free citizens in the lower classes tended to be peasants waiting for their free bread allotments in the absence of work). WW1 and WW2 were only possible because engines enabled 1 person to do the farm and factory work of 300, and are examples where the entire social order was diverted to war efforts. Could list dozens of similar examples. I'm fascinated by AI, but I don't think it's going to be all peace and rainbows on the other side of things. The engine, electricity and computers have all undoubtedly made our lives higher quality just as AI will. However the flip side of technologies that disrupt labor balances seem to be periods of rapid militarization and war.
@@patrickkn7607 Yeah, but... like this isn't close to the same. This is a big fat question mark as far as history is concerned. There are too many variables, that not even those developing it know how fucked or not fucked we are. There are so many possible developments we couldn't foresee with this.
UBI will come with conditions. Compliance will be monitored 24/7. Noncompliance will get you cut off from everything in seconds. No human involved in this process. Good luck.
Great video, thanks! I would also prefer if the content could be trimmed into 12-15 minutes by avoiding some repetition and focusing on the most relevant points. Perhaps some videos could be split into two or three separate ones? I've slowly learned to ignore the clickbait titles, though I'm not sure they're worth the frustration they cause. Regardless, this is one of the top 3 AI channels on UA-cam!"
The same way that our brains have specialized centers of neurons that do specific tasks incredibly efficiently, will the the same way that we create efficiency in Ai. A ton of individually optimized clusters that talk to each other, instead of one giant model with an input and an output.
For sure, the compartmentalization of specialization is starting to look very brain like. I think some sort of sentience could arise from the interplay of dozens or thousands of systems.
But thats not possible with todays supercomputer...we want something big that this neuraton network handle with single data..hope nvidia its future creating artificial womb ..10 years away
@@piteshbhanushali1140of course it's possible, what do you mean?😂 Do you think OpenAI only does one fine tuned model? They have thousands of fine-tuned models AT LEAST, and hundreds of thousands of API requests every minute, etc... You think they can't handle 10 extra API calls from 10 expert finetuned "parts of a brain"?
AI Imagination Engines To enable AI systems to generate novel ideas and solutions, fine-tuned AI models could be developed as imagination engines. These models would learn from diverse knowledge sources and creative processes, and help other AI systems combine concepts, generate hypotheticals, and explore counterfactuals in ways that simulate human-like imagination.
And then you can take a bunch of small, fast, narrower domain models and hook them up in the right way to get more performant generality, potentially. And you could route to very large models for cases where you need integration of multiple domains. I think they can go beyond MoE arrangements to architectures with much higher degrees of modularity.
Increased Competition is the near-term result of implementing AI into regular business processes. The strong will become even stronger, and the weak will have even less chance (although there will be chances).
Wes, you nailed the first part completely -- but don't expect this finetuning to be available to open source models that aren't running in the cloud. they won't want to do that. That's why all the "startups" that "used" to have a plugin and got kicked in the face, realize they need to have their own compute, and their own models, because at any moment... "that finetuned" LLM environment could "change"... and "oops" we are no longer offering finetuned GPTs... we switched to "Neural AGI" -- which uses a new format, and requires X,Y,Z... These million dollar investors for the no longer present "plugins" are not going easily forget this lesson. What is "Laughable" is that people will be willing to let AI run their email -- you are simply wrong. Anyone doing that for anything other than "checking for spam" or writing form letters -- would be very foolish. People already despise that AI answers email questions with robotic replies...
Each country will have a ministry that trains their AI, a cultural representation, the cell wall between humanity and AI. In a post-labor world this interface will take up a big chunk of our attention.🙏
Am I the only one who notices that on this timeline that Wes talks about where AI goes from being able to do 0% of human tasks to 100% of human tasks, and then talks about how there is this great opportunity to make money with this technology, that no one is going to be able to spend money if you automate all of the human tasks? "You'll be able to make your e-commerce store so powerful with this incredible AI technology -- a technology that can do all work that humans do. It's going to be so great!" 🤣😭
Assistants API helps you build an AI that acts as a helper, completing tasks and interacting with various systems. Example Assistants API * Purpose - Build AI assistants * Focus - Task Execution and interaction. * Customization - Assistant capabilities and integration * Case - Superhuman (email assistant) ** Now consider: Assisted fine-tuning helps you improve an existing AI model to be more accurate and efficient within a specific domain. * Purpose - Create custom trained models * Focus - Domain-specific performance * Customization - Model training and parameters * Case - Harvey (legal AI)
Has the practical definition of "fine tuning" changed? My understanding of it is that all fine tuning does is give a set of sample inputs and outputs to train the ai in the style and format to respond with. It's not for creating specialized data-sets or teaching it new things. But this video make it sound like OpenAI's new fine tuning does more than that.
@alexjoens5014 I don't have time to watch the video I am a machine learning engineer I want to watch them but I come back from work and a couple hours till I sleep I am not wasting half my free time on a video
Good video ! I use gpt4 at work to hold indexes of services that I can query for clients in the social services, it super useful and gives me names contacts services everything from my database that I collected. And because its unstructured and structured I can ask really tailored searches to it
wink wink, we know where you borrowed this database, many companies making them esp trading ones, but now that *** in oponAi also got it, you delivered it by yourself like free money
@@lokiholland muh? Traders usually borrowing client base of their former employer, it's widespread and i've seen this 20 years ago, golden data with which you can start own business, but you brings same for free to nasty people in chatgeepeety
@@lokiholland traders always borrow clients base at their job, valuable to start own business, your database valuable to spamcallers, if chatgpt sells it
it only makes sense, that what is GPT5 big model will almost certainty act as an organizer for hundreds of smaller domain specific LLMs that can be called upon at any time.
Do you think Neural Networks will be the AI Backbone for the future? AI was developing for decades before NN jumped into pole position. Is there another approach to getting useful results?
Can you make videos on how to advertise for this skills set? OBVIOUSLY I’m using agents to push my pages higher and putting out content here n LinkedIn but I’m mostly getting W2 offers vs C2C and clearly there’s something im not discovering . I think there are so few people that know how to do this and have experience (I do this for a major company now), there is plenty of pie and the more smaller teams help businesses , the more it becomes the new norm. So far folks want to hire me “in-house” on W2 and of course I’m taking the interviews and sometimes the contracts but I prefer more freedom and more $. Any advice from anyone on that would be super welcomed.
i put them as listening to a radio show, but del boot-licking openAIM$ ignoring very interesting other projects and the lack of depth in the analysis got me really tired. Speacially when the shallow "we see the future of work and it's going to be this way, everything is going to be amazing!!" Sad, naive, improvised,. that's it. eenough ranting. Thank you very much if you read this I convinced myself. Unsubscribed. See you!
i still dont understand. so i could make money by selling customized models? as in, i customize a model for a customer and sell it? or i buy customized models and sell them for "retail"?
I'm on the portion where you discuss Harvey as I write this comment. I love AI, but one of my concerns with AI retrieval is unintentional cherry picking. I wonder if some of the human jobs of the future will be specialized analog researchers trained to go back to first sources and be aware of evolving discussion over history to ensure that we retain full information despite censorship and unintentional ignorance of sources, even and especially by AI. Seems like it might be important as AI becomes our search engines and now that it's shaping our laws and how they are interpreted.
The glacial speed of Government may pay off, if they write the new policy such that forms will be filled out using an AI then a new platform can be snapped into place without having to build it all years behind the curve. I mean, you talk to the AI and it asks you questions based on what government entity you are interacting with and picks the correct form, then fills it out for you. Just think of how smooth the DMV will be when you can just sit on your phone for a few minuets then show up at the DMV and they have your paperwork ready and entered, you prove your real, the DMV approves your paperwork, and you pay the administration fee. Same for the FBI, IRS, NOAA, NASA, DOJ, there are so many forms and interactions which are overwhelming that can just vanish into a conversation with a robot. Absolutely will not happen, half the people in government spend their career trying to make the entire system break because they believe when it all breaks they'll be the leader of the local warlord faction in the new free world.
UBI is not the answer. Universal Basic Resources is. ALL people need Equal PERMANENT UNREVOKABLE Access to the Highest Quality everything! UBI would be just another way to make humanity dependent on a financial institution.
@nicholaslarue9400 at some point in the future, effort won't be needed or required from anyone. All problems will be solved and all work will be automated.
An Ai that paints a burger that never quite lives up, a steak without sizzle, where you can't quite remember the day you first rode a bike. The human becomes the integer, the pixel, the remnant archaic, the day and the plan for which he's but a dot in the mosaic. Once the miner, the farmer, the logger.. Meets the 6 lane, the high speed, amphibian correlate, in a grand game of frogger He who runs all that might even walk on the water.
The titles are repetitive and I have no real idea what its about until I (rarely) click it. It's usually misrepresented and when I find that out, and its dragged out 40 minutes, I just close it. This was one of like 2 channels I didn't block with regards to ai news and new platforms/software. But lately I don't really watch videos. Keep the exciting clickbait stuff to short videos and the real substance stuff with real titles in long form. For something truly exciting I'll sit around for 40 minutes on a whim. For a subject I'm interested in I'll also sit around for that when I make the time. For a title that generates a major sense of urgency and then its a long drawn out presentation and a low-energy topic I just feel cheated for time. For this video and the topic, I only cared about the first 5 minutes of information. My opinion isn't the only one that matters and its your channel but I wanna throw my 2c in the mix. Have a good one
This sounds like windows when it first came out. Build a base operating system then let people develop programs for the operating system then buy the apps out when all the bugs are out. Boom! Monopoly! The big kill the small again. It’s their MO.
I love your videos and want to watch all of them but due to large length of it I generally skip or watch in hurry. Can you please trim down it more to reduce length of videos?
Imagine the use cases opening up for hackers to take control of businesses who have these LLMs deeply imbedded, or essentially doing all of the business. Of course the attackers will use LLMs. AI dog eat dog. Regarding the use case of fine tuned LLMs assisting nurses. The extra free time will not go to more attention to patients needs, but rather the hospitals will merely cut their nursing staff. How is this not obvious?
I appreciate the time you take to make these videos. But they are so long...... Most people don't have time to sit through them. Could you not make them shorter, around 10 minutes?
@@thailandretromods I don't use TikTok. You may not have a life, but most people do and don't have 40 minutes or an hour to watch these videos which seem to be intentionally drawn out to maximise the number of ads that can be shown....
sounds like finetuning is another word for when a human brain choose to fous. I swear Im starting to relize more and more that we might not be as consiouss oursleves as we thought. Imainge when they start to be able to understand concepts instead of word for word and they focus on that.
All that money making, takes the form of money saved from jobs destroyed. Less jobs, less spending capacity in the economy. It will be a short lived party.
Naaa ... It is the opportunity of a lifetime for the monied individuals at the top. Guess who will set the conditions for you to obtain UBI. That is ultimate power in a world with no alternative ways to earn a living.
@ZappyOh no spending power no profits, there will be a party for a while until market saturation reaches a point where there isn't enough customer spending power and those profits will collapse. The economy will optimise itself to death as everyone tries to minimise labour costs, which vandalises spending. A UBI is a sticky plaster that inevitably, predictably, will lead to social unrest. As governments abuse UBI to assert control over private life. They won't be able to help themselves, it will come with conditions and those conditions will be increasingly totalitarian until intolerable. It's all going to burn and it's not because AI failed in its assigned task, or went out of control, it will be because human's are sht at not letting power go to their head.
I dunno man, I'm starting to think we won't get GPT5. Surely you guys remember what Microsoft does. They get ahold of companies, and then that company just goes dark. Technically it's still there, but it just doesn't do what it used to anymore. And it certainly stops blazing the trail. A trail that, who'd have thought, could hurt micro$oft. They're not a company that helps, they're a company that hurts. It's just what they are. I think OpenAi is probably finished, and that time will bear that out. Oh we'll hear all the platitudes and "It's only 18 months away!" for years to come. But if anyone out there cracks AGI, it will be someone else. Some other company will have to take up the mantle, and they probably will.
I get what you're saying but I don't think It's going to be the case this time. Microsoft is looking to get ahead even further. They are pushing limits beyond regulations and what they are building in terms of compute clusters just for AI is insane and the biggest in the world. They have by far the best chance to make AGI and OpenAI is hiring people like crazy. This is the opportunity of a century and microsoft knows it. They are putting everything on the line, perhaps the biggest gamble in history and I think it they are on the money.
Fool me once … . They are just the next generation of Microsoft. Those who fail to know history are destined to repeat it. Anyone remember Officewriter? Probably not.
The process is like this: 1. Do a agent to do what you want, write some examples and put all of that in a prompt. Costly because have there a lot of explanations on what to do, but it's easy to change. 2. After your software work and your flow is stable, get examples of successful run of you software. 3. Fine tune and remove the need to explain what you want. Its harder to change, but there low cost. 4. Get more data and enhance your model to be better in the edge cases.
@@justinwescott8125 The latest Chat GPT fails hilariously even at writing a simple gomoku game... and you suckers believe that will just suddenly turn into AGI? 😆 grow up hypeoids.
![Zuckerberg and Jensen Discuss The Future of AI [SUPERCUT]](/img/n.gif)
Wes, this is one of the best videos I have seen this year. Please please please DO NOT make the shorter. You insights are incredible. I wish I could get more of you. You truly bridge that divide between the average person and the super geek
Your business centric viewpoint on these new developments are valuable ! Thank you.
Can@wes Roth make more content on how to position oneself as one of these agencies?
Only stunned today. Good. I don't think I could take another shock.
Your videos are informative and the duration of your videos are just fine.
To Serve All Mankind.
😂
To wrench every penny out of mankind you mean.
So basically you put the sweat into building their llm, then they steal all your customer and say thanks for fine tuning our product.
"It's a cookbook"
@@babbagebrassworks4278who knows knows
Fuck OpenAI. I'm tired of the rich get richer and poor get poorer, which we're all currently rushing towards with absolute joy. I'll release my own actual, proper AI assistant shortly as I'm tried of everything coming out being a Chat GPT wrapper.
Forget these goldfish attention span having spanners, keep up the great summaries bro!
I cannot wait for a future free of human suffering through necessity of work. Can't wait for UBi and freedom for everyone
Historically, societies that freed themselves from the burden of hard labor tended towards complete militarization and population decline. Early in Spartan history, they enslaved neighboring city states and were a completely militant society for over 1000 years. The Roman empire had 7 slaves for every 1 citizen at their military peak (and free citizens in the lower classes tended to be peasants waiting for their free bread allotments in the absence of work). WW1 and WW2 were only possible because engines enabled 1 person to do the farm and factory work of 300, and are examples where the entire social order was diverted to war efforts. Could list dozens of similar examples.
I'm fascinated by AI, but I don't think it's going to be all peace and rainbows on the other side of things. The engine, electricity and computers have all undoubtedly made our lives higher quality just as AI will. However the flip side of technologies that disrupt labor balances seem to be periods of rapid militarization and war.
Well, ai could be promising to eliminate the peasant status. Unlikely but possible in a eutopia
@@patrickkn7607 Yeah, but... like this isn't close to the same. This is a big fat question mark as far as history is concerned. There are too many variables, that not even those developing it know how fucked or not fucked we are. There are so many possible developments we couldn't foresee with this.
The chances that the Libertarians and neoliberal economic types that dominate the AI revolution will distribute such a UBI is exactly ZERO.
UBI will come with conditions.
Compliance will be monitored 24/7.
Noncompliance will get you cut off from everything in seconds.
No human involved in this process.
Good luck.
Great video, thanks! I would also prefer if the content could be trimmed into 12-15 minutes by avoiding some repetition and focusing on the most relevant points. Perhaps some videos could be split into two or three separate ones?
I've slowly learned to ignore the clickbait titles, though I'm not sure they're worth the frustration they cause. Regardless, this is one of the top 3 AI channels on UA-cam!"
20min for me
Thank you. Your videos are insightful and you make your points brilliantly.
Seems like a good way to out-source agi. If all those refined agents get good enough/varied enough, then OpenAI can just combine them all. Boom AGI.
Yes. That is how it will work.
A high level, pondering master-AI, querying a plethora of dumb domain-expert AIs.
The pondering AI is Q*.
Yeppp. But not possible right now because that kind of AGI will have the electric bill of an entire country.
AI first task: zero cost energy discovery
@@ZappyOh your mom is Q*
@@ZappyOh I like that. Pondering AI.
I actually like the longer vids
The same way that our brains have specialized centers of neurons that do specific tasks incredibly efficiently, will the the same way that we create efficiency in Ai. A ton of individually optimized clusters that talk to each other, instead of one giant model with an input and an output.
For sure, the compartmentalization of specialization is starting to look very brain like. I think some sort of sentience could arise from the interplay of dozens or thousands of systems.
That's already happening even within the model architecture to some extent. Mixture of experts
But thats not possible with todays supercomputer...we want something big that this neuraton network handle with single data..hope nvidia its future creating artificial womb ..10 years away
@@piteshbhanushali1140of course it's possible, what do you mean?😂 Do you think OpenAI only does one fine tuned model? They have thousands of fine-tuned models AT LEAST, and hundreds of thousands of API requests every minute, etc... You think they can't handle 10 extra API calls from 10 expert finetuned "parts of a brain"?
@@themprsndevbut its not possible in todays gpu
Thanks, Wes; your work is excellent as it makes it possible for a broad range of people to understand this rapidly moving creature
Thanks for your service Wes! Keep it up! ❤
AI Imagination Engines
To enable AI systems to generate novel ideas and solutions, fine-tuned AI models could be developed as imagination engines. These models would learn from diverse knowledge sources and creative processes, and help other AI systems combine concepts, generate hypotheticals, and explore counterfactuals in ways that simulate human-like imagination.
You just put my thoughts into words really fucking well there. Whoah. that shit cray...
Like googles deepmind?
And then you can take a bunch of small, fast, narrower domain models and hook them up in the right way to get more performant generality, potentially. And you could route to very large models for cases where you need integration of multiple domains.
I think they can go beyond MoE arrangements to architectures with much higher degrees of modularity.
Great video Wes, i absolutely loved this deep dive!
Increased Competition is the near-term result of implementing AI into regular business processes.
The strong will become even stronger, and the weak will have even less chance (although there will be chances).
That is really great coverage Wes... this is a real glimose into the future of AI... STUNNING SHOCKING ETC ETC
Jesus fucking Christ, 2 STUNNING developments in 2 days!
Wes, you nailed the first part completely -- but don't expect this finetuning to be available to open source models that aren't running in the cloud. they won't want to do that.
That's why all the "startups" that "used" to have a plugin and got kicked in the face, realize they need to have their own compute, and their own models, because at any moment... "that finetuned" LLM environment could "change"... and "oops" we are no longer offering finetuned GPTs... we switched to "Neural AGI" -- which uses a new format, and requires X,Y,Z...
These million dollar investors for the no longer present "plugins" are not going easily forget this lesson.
What is "Laughable" is that people will be willing to let AI run their email -- you are simply wrong. Anyone doing that for anything other than "checking for spam" or writing form letters -- would be very foolish. People already despise that AI answers email questions with robotic replies...
I am STUNNED
YASSSSS STUNNING STUNNING STUNNING ABSOLUTELY STUNNING
STUNNED
Really? 😂
Thank you for confirming.
Again, Wes has STUNNED us. How creative.
Each country will have a ministry that trains their AI, a cultural representation, the cell wall between humanity and AI. In a post-labor world this interface will take up a big chunk of our attention.🙏
i am excited!!! p.s. why no shocked 😲 in title
This was excellent.
I'm not saying you should drink some caffeine but I had to listen to this at double speed today
Am I the only one who notices that on this timeline that Wes talks about where AI goes from being able to do 0% of human tasks to 100% of human tasks, and then talks about how there is this great opportunity to make money with this technology, that no one is going to be able to spend money if you automate all of the human tasks? "You'll be able to make your e-commerce store so powerful with this incredible AI technology -- a technology that can do all work that humans do. It's going to be so great!" 🤣😭
Thank you!
I got slightly confused comparing the Assistants API, and the assisted fine tuning. How are these 2 different?
Assistants API helps you build an AI that acts as a helper, completing tasks and interacting with various systems.
Example
Assistants API
*
Purpose - Build AI assistants
*
Focus - Task Execution and interaction.
*
Customization - Assistant capabilities and integration
*
Case - Superhuman (email assistant)
**
Now consider:
Assisted fine-tuning helps you improve an existing AI model to be more accurate and efficient within a specific domain.
*
Purpose - Create custom trained models
*
Focus - Domain-specific performance
*
Customization - Model training and parameters
*
Case - Harvey (legal AI)
There’s so much ai news, I watch everything at 1.5 speed
How do you increase speed
If they talk slower, I’ll do it at x1.75 😅. Tap the video and Select the gear ⚙️ icon next to the CC and then “playback speed”
Has the practical definition of "fine tuning" changed?
My understanding of it is that all fine tuning does is give a set of sample inputs and outputs to train the ai in the style and format to respond with. It's not for creating specialized data-sets or teaching it new things.
But this video make it sound like OpenAI's new fine tuning does more than that.
make your videos shorter please or make a summary video
Go watch shorts we want hr long videos make them longer
@alexjoens5014 I don't have time to watch the video I am a machine learning engineer I want to watch them but I come back from work and a couple hours till I sleep I am not wasting half my free time on a video
He has time stamps and summaries in his description. If you're hard on time, just skim through the vid.
TikTok victim
@@Aemond-qj4xt then listen to it
Good video ! I use gpt4 at work to hold indexes of services that I can query for clients in the social services, it super useful and gives me names contacts services everything from my database that I collected. And because its unstructured and structured I can ask really tailored searches to it
wink wink, we know where you borrowed this database, many companies making them esp trading ones, but now that *** in oponAi also got it, you delivered it by yourself like free money
@@fontenbleau huh?
@@lokiholland muh? Traders usually borrowing client base of their former employer, it's widespread and i've seen this 20 years ago, golden data with which you can start own business, but you brings same for free to nasty people in chatgeepeety
@@lokiholland traders always borrow clients base at their job, valuable to start own business, your database valuable to spamcallers, if chatgpt sells it
@@lokiholland will oponai sell your database to spammers?
Wow Wes.. wow.
AI doesn't have to be better than you, just cheaper: an error rate of 20% or a 300% slower completion time are non factors when it saves 99% of costs.
it only makes sense, that what is GPT5 big model will almost certainty act as an organizer for hundreds of smaller domain specific LLMs that can be called upon at any time.
Do you think Neural Networks will be the AI Backbone for the future? AI was developing for decades before NN jumped into pole position. Is there another approach to getting useful results?
Can you make videos on how to advertise for this skills set? OBVIOUSLY I’m using agents to push my pages higher and putting out content here n LinkedIn but I’m mostly getting W2 offers vs C2C and clearly there’s something im not discovering . I think there are so few people that know how to do this and have experience (I do this for a major company now), there is plenty of pie and the more smaller teams help businesses , the more it becomes the new norm. So far folks want to hire me “in-house” on W2 and of course I’m taking the interviews and sometimes the contracts but I prefer more freedom and more $. Any advice from anyone on that would be super welcomed.
Let's hope AI will be able to help us overcome the climate crisis 🤞
Lmfao
What software are you using for your screenrecording and drawing on it live?
Andrew ng argues that auto agentic workflows are the future. It seems investors are forcing open A.I towards that direction.
I don’t even watch these videos anymore. I just tap forward through them
i put them as listening to a radio show, but del boot-licking openAIM$ ignoring very interesting other projects and the lack of depth in the analysis got me really tired. Speacially when the shallow "we see the future of work and it's going to be this way, everything is going to be amazing!!" Sad, naive, improvised,. that's it. eenough ranting. Thank you very much if you read this I convinced myself. Unsubscribed. See you!
i still dont understand. so i could make money by selling customized models? as in, i customize a model for a customer and sell it? or i buy customized models and sell them for "retail"?
I'm on the portion where you discuss Harvey as I write this comment. I love AI, but one of my concerns with AI retrieval is unintentional cherry picking. I wonder if some of the human jobs of the future will be specialized analog researchers trained to go back to first sources and be aware of evolving discussion over history to ensure that we retain full information despite censorship and unintentional ignorance of sources, even and especially by AI. Seems like it might be important as AI becomes our search engines and now that it's shaping our laws and how they are interpreted.
The glacial speed of Government may pay off, if they write the new policy such that forms will be filled out using an AI then a new platform can be snapped into place without having to build it all years behind the curve.
I mean, you talk to the AI and it asks you questions based on what government entity you are interacting with and picks the correct form, then fills it out for you. Just think of how smooth the DMV will be when you can just sit on your phone for a few minuets then show up at the DMV and they have your paperwork ready and entered, you prove your real, the DMV approves your paperwork, and you pay the administration fee. Same for the FBI, IRS, NOAA, NASA, DOJ, there are so many forms and interactions which are overwhelming that can just vanish into a conversation with a robot.
Absolutely will not happen, half the people in government spend their career trying to make the entire system break because they believe when it all breaks they'll be the leader of the local warlord faction in the new free world.
Who is the second in command?
Totally STUNNED!😯
No serious company will want to give their proprietary data to OpenAI
Yep roll out is like 20 years for corporate IT. Unless AGI becomes trusted overnight and can do everything on its own initiative.
UBI is not the answer. Universal Basic Resources is. ALL people need Equal PERMANENT UNREVOKABLE Access to the Highest Quality everything! UBI would be just another way to make humanity dependent on a financial institution.
So everybody gets the same regardless of the effort they put in. Great idea!
@nicholaslarue9400 at some point in the future, effort won't be needed or required from anyone. All problems will be solved and all work will be automated.
You should try starting your comments with “I think that…”.
@@brianfabiano5368 you should do the same.
An Ai that paints a burger that never quite lives up, a steak without sizzle, where you can't quite remember the day you first rode a bike.
The human becomes the integer, the pixel, the remnant archaic, the day and the plan for which he's but a dot in the mosaic.
Once the miner, the farmer, the logger..
Meets the 6 lane, the high speed, amphibian correlate, in a grand game of frogger
He who runs all that might even walk on the water.
Commenting for accurate shock 36:15
The titles are repetitive and I have no real idea what its about until I (rarely) click it. It's usually misrepresented and when I find that out, and its dragged out 40 minutes, I just close it. This was one of like 2 channels I didn't block with regards to ai news and new platforms/software. But lately I don't really watch videos.
Keep the exciting clickbait stuff to short videos and the real substance stuff with real titles in long form. For something truly exciting I'll sit around for 40 minutes on a whim. For a subject I'm interested in I'll also sit around for that when I make the time.
For a title that generates a major sense of urgency and then its a long drawn out presentation and a low-energy topic I just feel cheated for time. For this video and the topic, I only cared about the first 5 minutes of information. My opinion isn't the only one that matters and its your channel but I wanna throw my 2c in the mix. Have a good one
Where is Google? It seems to me that OpenAI is building the future and Google is asleep at the wheel!
This sounds like windows when it first came out. Build a base operating system then let people develop programs for the operating system then buy the apps out when all the bugs are out. Boom! Monopoly! The big kill the small again. It’s their MO.
I love your videos and want to watch all of them but due to large length of it I generally skip or watch in hurry. Can you please trim down it more to reduce length of videos?
Watch them at 1.5x speed.
WHAT? do you play EVE Online?
The way you can look at it is GPT is the human and then the other half is like you sent to human to school for a specific job
Too long bro. Even at 2x speed its a 20 minute video
i am starting to feel sad for AI because we are training it to bring money ...when the goal should be to make this place better..am i wrong?
I don't understand what this means. The economics have to be right for a company to stay afloat. OpenAI doesn't run on rainbows and butterflies.
Yes, but I wonder why they would choose anyone else for all these ai jobs than Microsoft, since they use word office anyway.
❤❤❤❤❤❤
They should have named the it Mike instead of Harvey.
Excited
and shocked . Always shocked .
Imagine the use cases opening up for hackers to take control of businesses who have these LLMs deeply imbedded, or essentially doing all of the business. Of course the attackers will use LLMs.
AI dog eat dog.
Regarding the use case of fine tuned LLMs assisting nurses. The extra free time will not go to more attention to patients needs, but rather the hospitals will merely cut their nursing staff.
How is this not obvious?
Why do you think cutting staff isn't obvious to people?
I think it is pretty obvious to everyone.
Day by day my job gets closer to being replaced.
It's amazing to me how much of this video represents AI trying to sell us s***.
I appreciate the time you take to make these videos. But they are so long...... Most people don't have time to sit through them. Could you not make them shorter, around 10 minutes?
You're not watching a tiktok buddy.....
@@thailandretromods I don't use TikTok. You may not have a life, but most people do and don't have 40 minutes or an hour to watch these videos which seem to be intentionally drawn out to maximise the number of ads that can be shown....
@@thailandretromods I don't use TikTok. You may have no life and can sit and watch these long videos, lots of other people can't do that though.
Omg omg it’s soooo stunning
What exactly makes you think I am a blob? My precious, how do I look like to you?
stop hyping every video with STUNNING, there are actual stunning releases so just save it
sounds like finetuning is another word for when a human brain choose to fous. I swear Im starting to relize more and more that we might not be as consiouss oursleves as we thought. Imainge when they start to be able to understand concepts instead of word for word and they focus on that.
All that money making, takes the form of money saved from jobs destroyed. Less jobs, less spending capacity in the economy.
It will be a short lived party.
Naaa ... It is the opportunity of a lifetime for the monied individuals at the top.
Guess who will set the conditions for you to obtain UBI.
That is ultimate power in a world with no alternative ways to earn a living.
@ZappyOh no spending power no profits, there will be a party for a while until market saturation reaches a point where there isn't enough customer spending power and those profits will collapse. The economy will optimise itself to death as everyone tries to minimise labour costs, which vandalises spending.
A UBI is a sticky plaster that inevitably, predictably, will lead to social unrest. As governments abuse UBI to assert control over private life. They won't be able to help themselves, it will come with conditions and those conditions will be increasingly totalitarian until intolerable.
It's all going to burn and it's not because AI failed in its assigned task, or went out of control, it will be because human's are sht at not letting power go to their head.
@@ZappyOhthat my reply to this comment has vanished reminds me why I usually don't bother to have conversations on UA-cam.
funny how openAI did this atthe same tie anthopic did
Anthropic announced today, Claude 3 Tools. Tit-for-tat.
You do realize that your account needs a new pay wallet?
I dunno man, I'm starting to think we won't get GPT5. Surely you guys remember what Microsoft does. They get ahold of companies, and then that company just goes dark. Technically it's still there, but it just doesn't do what it used to anymore. And it certainly stops blazing the trail. A trail that, who'd have thought, could hurt micro$oft. They're not a company that helps, they're a company that hurts. It's just what they are. I think OpenAi is probably finished, and that time will bear that out. Oh we'll hear all the platitudes and "It's only 18 months away!" for years to come. But if anyone out there cracks AGI, it will be someone else. Some other company will have to take up the mantle, and they probably will.
I get what you're saying but I don't think It's going to be the case this time. Microsoft is looking to get ahead even further. They are pushing limits beyond regulations and what they are building in terms of compute clusters just for AI is insane and the biggest in the world. They have by far the best chance to make AGI and OpenAI is hiring people like crazy. This is the opportunity of a century and microsoft knows it. They are putting everything on the line, perhaps the biggest gamble in history and I think it they are on the money.
That's all the time with me u need. I hope you got enough of enders ip
I would appreciate shorter videos. I'm not going to watch 38 min.
Fool me once … . They are just the next generation of Microsoft. Those who fail to know history are destined to repeat it. Anyone remember Officewriter? Probably not.
Please pin this so that when my words come true, people will find them.
Patent ur ideas as fast as possible!
I know how to do It, Hire me... LOL
The process is like this:
1. Do a agent to do what you want, write some examples and put all of that in a prompt. Costly because have there a lot of explanations on what to do, but it's easy to change.
2. After your software work and your flow is stable, get examples of successful run of you software.
3. Fine tune and remove the need to explain what you want. Its harder to change, but there low cost.
4. Get more data and enhance your model to be better in the edge cases.
you have so much fluff in this video
Ask ChatGPT to give you alternate words to “stunning”
Ha-ha 😂 after no liability for several lawyers in USA news destroyed their career using it?
Like your videos but the clickbait titles is getting old :/
I don't even look at them because the thumbnail is so mesmerizing lol
You talk about some of this stuff as if it wasn’t merely a substitute for a PA…
38 mins is too long. release twice a week if you;re going to be this dense.
Oof, title didn't have shocking but stunning. Disliked, unsubscribed. This channel is nothing but fake news
Your videos are much too long. Give ChatGPT the page and let it summarise. This can help!
I recommend 2x speed with captions 😊
Nah man. The voice and the personality make it work
One more "Shocking" or "Stunning" and I will unsubscribe! Stop this clickbait s#it!
All the overhyped techbros who think AGI is just around the corner are looking forward to a giant flop.
Well gosh, if moonstriker7350 says AGI isn't coming, it must be true😂
@@justinwescott8125 The latest Chat GPT fails hilariously even at writing a simple gomoku game... and you suckers believe that will just suddenly turn into AGI? 😆 grow up hypeoids.
First