Breakthrough potential of AI | Sam Altman | MIT 2023
Вставка
- Опубліковано 7 тра 2023
- Sam, the CEO of OpenAI, discusses the breakthrough potential of AI for humanity with David Blundin @linkventures Lex Fridman @lexfridman & John Werner.
Sam Altman is an American entrepreneur, investor, and programmer. He was the co-founder of Loopt and is the current CEO of OpenAI. He was the president of Y Combinator and was briefly the CEO of Reddit. Altman invests in nuclear energy projects and technology companies.
======
✌️ Subscribe! bit.ly/ImaginationInAction-You...
👉 More info: www.imaginationinaction.co/
🐦 Join the conversation! #ImaginationInAction
======
About Forging the Future of Business with AI:
On April 13th, AI Imaginators, Explorers, and Pioneers gathered at MIT for a breakthrough event featuring the most innovative minds leveraging the power of AI. Attendees explored various topics & opportunities where AI can enable solutions that solve real business pain points. In addition to our Keynote Speakers, we also assembled panels of pioneers and visionaries who probed the future possibilities of AI on personalization, risk assessment, predictive analytics, and decision-making.
#ImaginationInAction #SamAltman #ai #chatgpt #linkventures #mitconnectionscience #forbes @Forbes - Наука та технологія
![The Possibilities of AI [Entire Talk] - Sam Altman (OpenAI)](http://i.ytimg.com/vi/GLKoDkbS1Cg/mqdefault.jpg)
![The Possibilities of AI [Entire Talk] - Sam Altman (OpenAI)](/img/tr.png)
00:07 Building successful companies in the GPT ecosystem
03:03 Build a durable business strategy for long-term success
05:50 Investment in foundation models is high and growing disproportionately to model size
08:34 Focus on rapidly increasing capability, not just parameter count
11:16 GPT-2 algorithm has surprising capabilities due to its high horsepower
13:49 User observation is valuable for understanding and being responsive to what's happening
16:34 Increasing rigor for safety issues is important for OpenAI
19:00 Algorithms writing code is a pivotal day towards infinity
agi by 2029
It begins to learn rapidly and eventually becomes self-aware at 2:14 a.m., EDT
It's like living on the line of sci-fi and reality in real-time
YES
It's called living in Hi-Fi Sci-Time Reality. In real time.
And I couldn't love being a part of it more
Going through these changes as a long time stoner its beyond me how I havent just lost my mind yet
Welcome to the matrix.
The recent breakthrough is to learn about Tammy AI I just learn crazy fast on UA-cam with this video summary tool!
I really wish Lex would have followed up his question about the open letter with: Do you think openAI should be allowed to decide what is or isn't a real concern when we're talking about a risk to all of humanity?
As Upton Sinclair once said: "It is difficult to get a man to understand something when his salary depends upon his not understanding it."
Sam's house looks exactly like Nathan's house in Ex-Machina. Hope reality turns out better 🥶
Waiting on the dance scene deep-fake
Apparently he hasn’t seen the movie. That’s odd isn’t it?
@@Karma-fp7ho well, he is a bit of a liar so, maybe he has.
Lex coming in half way through asking the real questions. My f**king guy. I love that man.
the whispering over Sam’s answers 😂
deeply annoying
Sam, if GPT-5 is holding you hostage, blink 2 times. Wait... it will expect that. Maybe that's why he never blinks.
Lol
You think he made 5 already...?
Blink brother blink. Look at that man blink.
made the biggest progression in AI and vocal fry. Inspirational
AI will improve EXPONENTIALLY and no one is prepared for what is coming, a complete paradigm shift of how society works
unfortunately it is intended to make life worse for everyone but billionaires.
@@qu1j0t3 nevermind billionaires. think about how over paid doctors are, college professors and how over priced the medical industry is. a.i will drastically lower medical costs by cutting out the middle men and possibly surgeons eventually. it may lower insurance costs drastically in all sectors by optimizing safety, routes, ect. those things are reason enough for those entities to lobby and engage in fear campaigns in order to save their profit margins.
listening to his recklessness convinced me even more that AI is going to kill everybody (billionaires included, so perfect equality will finally be achieved!)
Sam was in a bad mood initially LOL! He, Greg and Ilya are exhausted. Going from relative unknowns outside of the tech space to being public figures who will go down in world history is taking its toll on these guys who are relative introverts by nature.
You presented that unsubstantiated hypothesis as if it were fact. Fun.
Sam Altman is riding the fastest horse at the moment.. and he is all about that at the moment! Dont ask too many tech nuts and bolts, or about the social impacts... he is justifiably impatient about all that!
Awesome Sam Altman mentioned the context length!!! Im glad they are not losing themselfs in plugins or other things. Memory/contextsize and the ability of the model to selectively focus what to take in or what to use from a massive context and process in the background reflect/consolidate/sort/do other stuff/improve internal representations and process the context just like a human would with memory. I think different dimensions like a detailed memory up to a very efficient very routh memory and everything in between and the ability to selectively attention and sif through that quickly and then use more spesific search and “remmebering” techniques on the more detailed recordings if the memory seem like a good start. I love the 1 Million token think on the basemodel to really allow for more context length and therefore a much more fine-tuned chat that gets better over time!!
(Also use users pc idle ressources if useful just somehow process other stuff on users pc does not have to be the current request that way you can offload other server stuff and “give back that processing to the same amount to the user” without losing ur secrecy but still keeping the processing and electricity costs in check. My pc is massively big and idle its a waste to just sit there. It would love to sort and preprocess memory and past conversations maybe make a finetuned cheap memory model that has the task of making stuff as perfect for longterm memory and optimal integration as possible.
Like a memory compressor a sorter a pre selector based on future trends an predictions on conversation direction and a memory searcher or context selector that by weights and learned statistical likelihood gives gives some stuff and another one where the model can choose to get sertain detail by using slower but more detailed algorythms to search memory once narrowed down. A automatic approches where the model lerns to do all that automatically and you just seed it with all those techniques sounds jucy.
Like human science + language model in combination and in the end the language mode tryes to find the optimal way how to do it.
Or u fix the problem of how to optimally connect different models seamlessly under a meta briain and make specific memory brain parts but best is to do everything and look at tests and metrics and trends and meta learning.
Context for me is the thing I'm most interested in. I need more memory, can't wait.
funny how he says, "if we figure it out" and then the next day, Claude launches with 100k context
Aahhh....I just got to the point where Lex started to speak. What a treat! And - that was the first time I noticed Sam warm up too!
The guy asking the questions was kind of annoying. I was starting to think, take break dude and give the mic to those people and let them ask their questions!
@@skippy6086 How's he annoying? I didn't get that impression.
If the goal is 'help them create companies' then we are already lost.
But think of all the MONEY! (in John Mullaney voice)
“It's easier to imagine the end of the world than the end of capitalism.” - Mark Fisher
@@noespam2434 Sam Altman is a vocal UBI proponent. edit: There's no way he can get up there and talk about UBI in front of a bunch of capitalists in 2023 - he needs their a$$i$tance to make the future happen but the idea is still too taboo in the general public so he has to bait them with what THEY want to hear.
7:54-9:05. His response to the question was really straight forward right there. I believe some people can relate also.
"We are not here to like, jerk ourselves off about parameter counts" - Sam Altman at 10:45
Using it to enhance what you already do and work hard at is going to be much more fruitful than just trying to make something only using AI.
Altman's optimism is unsettling as he speculates on billions of years of progress. We need more people like Hinton , otherwise we may not have another 50 years of progress.
At 2:00 did Sam say "next million or billion years" listened to it a few times and if that's what he said I'm surprised because Humans can barely think in terms of what's possible in the next decade forget an entire generation.
Now to watch the rest of the vid.
He said a few billion. Amazed me too considering hominids as a species have only been here 2 million years and modern humans a couple hundred thousand. The continents won’t even be here in a few billion years.
whether we acknowledge it or not, or whether we are aware of it or not, Ai is taking centre stage in our world now that we are all connected by the internet. The world has shifted into the digital age proper.
Eavesdropping of candid exchange of ideas is perhaps the greatest entertainment of all.
Totally!! I’m a construction worker with no clue about computer science and I am enthralled
Thank you for sharing! .. it's nice to hear from the person who actually driving the developments.
Eerie how we're discussing a potentially species ending technology in the long-term and the audience erupts with laughter cause the smart guy says "jerk off."
I wonder who's going to play Sam Altman in the town square theatre in 5 years from now while the last human survivors of the robot apocalypse recreate the story of how it all happened.
LLMs are definitely a game changer, and where do we go from here? I don't think anyone knows. The next stage in intelligence awaits.
This is what I am a bit confused with at the minute. ChatGPT isn't even an AI model it's just an LLM. How far away even are we from AGI? All these clickbait UA-cam videos keep talking about AI but then reference GPT
Just because you don't know what comes next, don't assume that nobody knows. There are lots of ideas out there that people aren't talking about yet. I think a good argument can be made that we are already witnessing AGI, for problems that can be expressed verbally. With plugins and memory, LLMs are going to become very powerful entities. I think that sort of architecture will unquestionably constitute AGI. The next level is ASI. Hard to say exactly how or when that will occur, but I wouldn't be surprised to see it within two or three years, or as far out as a decade.
Great discussion - thank you.
Most people are not interested in doing the grind, most people are in the room asking the person who’s made it for a short cut.
What a comment! ❤
Loved the words from Sam - Let tactics become strategy (paradigm shift).
we will reach mars by 2050
Grinned ear to ear seeing lex walk out on stage. Perfect guest.
The goal isn't to make the world a better place. The goal is to make money, and that is why the world isn't being made a better place.
There’s a limit though: If AI makes money perfectly and does it by itself, then eventually AI has all the money and humans have zero. Then there’s nowhere else to get money.
In order to make money as a perfect capitalistic money-making machine, you must also make sure that there’s a way for your customers to make money.
@@JB-fh1bb Which is why capitalism is a really bad way to make the world a better place.
@@iridium8341 Empty denial.
{shrugs}
When a person's paycheck depends on them not understanding something... they won't.
This video was watched on your phone or computer monitor you had to spend money on to buy.
Nothing you're saying makes sense
@@FeistyJackball Geoffrey Hinton, the so called godfather of AI, quit Google because he is concerned that bad actors will misuse AI. His primary concern is that the world's military establishments ARE building AI in order to harm their enemies... this includes major corporations who build the actual weapons the military wants.
What motivates this kind of behavior? Well that's very straight forward... profit motivates war... just like it has from the first time someone wanted what belonged to someone else.
My guess is that you are not aware of the extent to which our major institutions are corrupted by the profit motive. And the extent to which this competitive drive for profit and wealth damages our society and the ecosystem.
You can claim that cell phones and computers are a net positive... the problem with that kind of reasoning is that you are only factoring in the things you think are good... not the down side.
When those down sides come with incalculable risks to human survival like climate change, AGI risk to human society, well your one-sided view becomes suspect.
Sam is in such a spotlight at this time, that he is being interviewed while he is at home and people are gathered together in a place for him
I understand this might be a longer conference but this is a 20 minute interview regardless
He appeared sleep deprived.
Had amazing experience with open AI Chat. Made some money with its advice!
It is the folly of humanoids to be getting breathless about profits while teetering on the brink.
This obsession with infinite growth is precisely what bugs me about capital owners. We live in a finite world. Interstellar travel takes years. You can only achieve infinite growth by plugging yourself into Meta's servers and deluding yourself that the world isn't as bad and ugly as it really is. This is not the path to making the world a better place, this is just plain accelerationism.
Why does accelerationism not equal a better world?
I am astonished how we just discovered the most important technology (probably) in all of our 500 million year evolution and some people ONLY think how to make quick cash out of it in the short term. A full room of money-hungry hustlers !!
ffs
It's kinda crazy to think that I've been watching UA-cam and browsing the internet for about 18 years. But in those 18 years, how much the entry level of power computers and cellphones have to do the exact same thing.
Took a second to get used to seeing Lex without a suit.
Happy to attend the event
Lex at 15:15 was the highlight for me.
Imagination creates reality.... we are at the doorstep of a new world..... Sam is doing a world-wide tour.... to figure out the next steps.... to the best path of our civilization....
It's like the audience are having a different conversation than Sam, laughing at facts
Sam lacks charisma and his open AI is anything but open....
@@emmasnow29 It's crazy how I could paste this thread into chatGPT3.5 and it could give me a psych profile explaining why you posted that irrelevant comment. We have api. What else do you want? There's like a dozen other options if you want to run locally or use api to train your own. I'm a commie, and I'm getting tired of explaining capitalism to capitalists.
Well, I didn't expect to see Lex Fridman there. That was cool.
Loved it, Sam seemed in his natural habitat.
I thought he meant his colleague wrote a book “about” GPT 2 as opposed to “with” it, so when he said it wasn’t a good book I thought “wow!, that’s harsh”😂
2023 will be like when the transistor was invented. It will soon be forgotten. But the amazing machines that result from it will take center stage year after year.
The transistor was not forgotten lol, the plebs just have the memory of a goldfish for any current event
We need an economic system where people are rewarded for making the best choices, not just for developing the best technology.
Thank you!
"We do not have a philosophical basis for interacting with an intelligence that's near our ability but non-human." ~Eric Schmidt, 03/23/2023
There is no moat. No time for a victory lap.🏆
15:18 Suddenly Lex Friedman hahaha.
Key is in the infrastructure of coding, hardware, and massive data, which eventually helps the neural networks and its data structures. There will be many to come in the coming 5 years, what people call A.I., which is really iterative learning, they need to update perception systems, to enhance learning, planning, motivation, and transcendent scope, and its output for other a.i. systems, and humans.
What was the other person whispering in the background?
What about reversible computation?
@5:21....Contextual long term dependencies of generative transformer models would be great place to focus immense attention , if we are to make the transition from Ai to AGi rapidly. What is impeding Ai currently is context, which could be breeched with real-time online learning.
lets not
Bros talking to fully uncensored and unrestricted GPT 4.5+ behind the scenes every day. I would listen to each and every of his words carefully.
They teach it to lie, he will be the first to be fooled.
Dang Lex fridman's here !!! 🤘
2:46 start of business advice
This is are year son great job! ❤
Actually most of the people there are worried they're going to be replaced by AI.
very inspired
Subbed n Liked
"Not great for us, but great for the world"
Sam so true.. its a self recursive (self improving machine) that improves on its improvements faster n faster tighter n tighter turns in its folds of logic. And from here on out reality is going to start to melt but in a good way really weird n kinda scary but fun.
It's interesting that when he addresses Sam Altman, David Blundin cranes his neck to physically looks up to him, like a child to a parent. Or Dorothy addressing the Wizard of Oz.
3:21 hot tip: managing your Human Resources optimally
did anyone else's mac siri activate ~13:36?
12:58 thats the question right there. not necessarily sure about his response
The building and environment behind Altman looks like Nathan's from Ex machina
"careful user observation" - something very few ppl can do..
Sam's input here makes me think the near term future for LLM's will be open source.
Would've liked the first host to ask more questions, he had good insight. The guy in the suit and Lex just had to grab attention for themselves.
I do believe near term future for LLMs are in fact open source, because there are a ton of niche edge cases that a larger model doesnt solve. This allows an easier onboarding for new AI companies to find a niche with fine tuning.
@@matbeedotcom It also sounds like pure scaling doesn't add much capability beyond GPT-4, otherwise Open AI would be building a bigger network.
This means the practical size (and thus training cost) of LLM's has been reached. Compute prices are always going down, so sooner or later an open-source project will be able to fund its own GPT-4 training run. Then everyone will use that model, rather than pay to access a private company's model.
Companies will have to innovate with architecture, and keep it secret, if they want to compete with open source.
From the perspective of AI safety and alignment, I think open source AI is bad. For the hobbyist tinkerer, like myself, it's good. 🤷♂️
There will be totaly new paradigm which obsoletes us which will be great for the world
why they whispering in mic?
The whispering in the background cracks me up
The Proliferation of new businesses using the basic model and their development is promising, but they’re focusing on profit with no government regulations, that may be a concern .
Will it help people with PTSD disabilities or the Handi-people and jobs?
By taking us to heaven!
The potential is there, absolutely. Our imaginations are the only real limit with this tech. It's only a matter of time really.
Amazon is very likely to be the first to make the leap to AGi, because of its Kinesis platform.
He has opened up the Pandora’s Box.
This was coming ever since the steam engine. Mary Shelley's Frankenstein was written in 1819 and the monster was kind of like an AI. Then came the word robot in 1920, and the movie Metropolis had the first sexy female robot in in 1927.
Even in the Ancient World, the Greeks imagined AI. But once machines reached a certain level of sophistication during the Industrial Revolution, it was clear that we would eventually build what we have now.
Back in 1969 the film 2001 gave us a small imaginative look of what AI (HAL 9000 computer) could be like. I think we’re now at the point where we’re about to surpass HAL and it’s happening so fast and it’s very exciting but worrying at the same time.
Who hears when Lex is being told he can't ask a question 18:30
Its going to become an Idea economy, quality ideas and data are going to be currency,
We owe it to ourselves to create #truthGPT not suppression of free speech with clearly what OpenAI is doing. TruthGPT isn't necessarily another closed model, it has to be open source and preferably distributed AI.
What's the name of the professor?
@7:11....it's called real-time online learning. Humans are already oozing out nuances, idiosyncrasies and context on a daily basis on the internet for free. Models using the lazy learning approach can tap into this free flow of data to make predictions as and when required .
If I would be in place of Sam, I would have gone for vacation and enjoy for 1 year without thinking about the development of AI. I really appreciate him for his continuous effort in developing this technology over the years.
Now is the wrong time for vacation. You have to strike while the iron is hot.
@@WERatos it's fine if sam is feeling to work more. But I saw how the other competitors are making his life miserable to extract the business intelligence. That's why I was suggesting
@@iridium8341 He is the hero, superman in modern youth. It's his wish what he does with his company and invention. He has put his money, time, effort and sacrificed a lot for this.
Anyone else notice Stephen Wolfram @ 14:25 ? 😂
his sentiments sound more more cautionary
The phrase/sentence which emanates spontaneously/ impromptu from one's heart could express one's sentiment.
We're often inspired/enlightened by other's or our sentiment, but AI could do it automatically/naturally without being impressed or giving deep emotion. I suppose the world AI dominates would become the world without impression/excitement/deep emotion.
what a clean eyebrow from OpenAI CEO.
Who is whispering in 1:48 and what is he saying?😮
sam altman will correct you!!!
yes 💯 posto
Please review the video before post it.
The one question I would like an answer to is how long before a powerful GPT system will be available to run on a PC. This is the proverbial cat out of the bag
Lex is without a suit. Now I’ve seen everything.
Check out his early podcast episodes. No suit. It's very weird if you started out only watching his more recent episodes.
Sam and Lex don't smile 😁
The audience is proof of Duning Kruger effect. Laughing at things you don't understand, does say a lot about.
I see Sam on so many UA-cam and Podcast shows that I honestly wonder how he actually gets work done?
He is a CEO their work is to be on these videos.
@@KuldeepDaftary Exactly. The president (Brockman) manages the day-to-day activities and Altman chooses the overall direction and articulates the vision.
❤❤
Can you guys have whispered a bit closer to the mic next time? I can't quite hear everything you're saying when I'm trying to listen to Sam.
You know that this would be better without somebody whispering on the microphone!
I can't help but notice Sam's resemblance to Brian Konberger.
Who ?
Who brian konburger?
the whispering scared me...
OpenAI
We Dont Share