Sam's interviews getting 1000x less engagement than clickbait ChatGPT videos is quite astounding. And people are worrying about AI taking their jobs and creativity... Well if you're here watching this I think you'll actually be just fine 😂.
I have been watching an unhealthy amount of Sam Altman videos looking for clues as to what the future holds. It is all so ambiguous. I find ChatGPT analagous to Google. People have had unlimited access to the world's information for two decades yet still act like complete morons. So something tells me that this would be adjacent in use to Google. Yet another part of me says that this is finally the tool that will correct human cognition, and thus I will be out of high-IQ work. I don't know. All so strange.
Seeing Sam articulate his thoughts and seeing the care put into the answers give me hope for the future of AGI, but also leave me very worried about the future. Like he mentioned - with factory workers and truck drivers now being much lower/last on the list to become automated, the higher paying jobs in society will be the first to get the axe. Everyone should be tuning into anything Sam Altman going forward.
It's unlikely that the higher paying jobs are going to "get the axe". But people are going to need to rapidly shift to using tools like ChatGPT to improve their own performance - and people's standards for pretty much all kinds of services are going to go up. So that you'll have to be considerably better to remain competitive.
@@Leto2ndAtreides I’m not sure what you are basing that off of - in grocery stores, automated checkout reduced the number of jobs needed by a substantial amount. Like Sam was mentioning, now, the white-collar jobs are able to be automated. The conventional thinking that truck drivers were going to be the first to be automated has turned on its head. Now you don’t need 18 customer service reps, you can get away with 4. The mass layoffs that are going to happen because of automation we are not prepared for as a country.
As a mediocre software engineer, I hate creating mock unit tests. It's in my opinion, lateral thinking while you try to solve a problem. So I created a solution(a method or function) that's working. Now I ask ChatGPT to create mock unit test(TypeMock) for it. Amazingly it created unit test, testing the major functionality I try to achieve in 5-10 secs. I could add it to my automated test suit so system is more resilient wrt future code change. That's amazingly cool and I love it. Just my opinion.
As a life long professional educator, the institution must adapt. We can't bury our heads in the sand at the threat of new technologies and new tools. I have used chat ai to seriously enhance and enable several lessons just in the past month. Incredible power to make the classroom experience more relevant and personalized. And I am working at the elementary level. Huge applications almost no one is thinking about.
@justsomeguy1074 yes, teachers and schools need to do a ton of thinking about their purpose in today's culture. As a teacher I believe I only have 3 legitimate purposes, motivation, curation, and feedback. Many of my students have zero natural interest in learning the subject matter. They will fail on their own without me to convince them this is something worth doing - I spend most of my planning on motivation. Assuming a motivated learner, they might spend hours or weeks wasting time on bad or pointless learning content be it videos, books, podcasts, or lectures. They can't always know what they don't know and an ai won't be able to do that for them. And finally this is where someday an ai might replace one of my 3 tenets is feedback. But I would have all my students using ai based feedback, then give personal time to the ones that are not progressing with the ai. It's probably about adaptation to maxize production more than replacement. But only time will tell. Is ultimately, however if an ai Replaces me as a teacher - good, make learning more accessible, more equitable, and more effective any way we can, I consider it a moral duty.
We are planning a major AI update to our product, B4Grad. We are a popular studying platform for students across US & Canada. We would love your feedback.
I'm 66 and have always been fascinated with tech. I hope I'm around for at least a couple of the things he mentioned in this interview - fusion and AGI. This guy is super intelligent and very articulate. Very much enjoyed this video.
@@jesusrajaimesbe Because he uses his imagination and talent to create a software program that is so powerful it becomes a major milestone in human history. AI when connected to the internet suddenly gives a user great power, but I like that Sam recognizes that power and sees the benefit and threat to such a technology
@@bonks4395 In addition to that he is not trying to sell or promote chatgpt. For example he could say something like " We are using this high tech, super cool blah blah to make it perfect, so we are doing it, we are great, there is nobody else like us etc...", which is what most business selling people do these days,. Instead he is honestly saying that LLM is not there yet, you are impressed in your first interactions and then you realized that you still need to do more. And the next releases will be slower etc.
16:00, As a teacher I think this is the best response ever. We have to adapt, which is going to very difficult though, than rather avoiding it. We should expand the conversation. The future is uncertain 🤔
I think Sam Altman is being responsible in how he approaches AGI, I would disagree with his statement that "everyone will adapt". The people who adapt will learn to survive in a new society where AGI is commonplace and democratized. Those who don't adapt, will get left behind, and I would argue there is a strong potential for this technology to cause widespread panic across many industries and sectors of the economy. I will be following the developments closely over the next 5 years. Based on the current trajectory, 5 years seems reasonable to achieve AGI.
thank you for this. releasing the AI capabilities in phases is a sound safety strategy and ensuring decentralisation of access to AI and energy to avoid monopoly, hope this remains in place. ❤
Quite a few people think AGI will happen in the 2030s. Sam hasn't said this specifically, but he has said in at least one interview that AI will get to close to human level in 7 years or so. If any of u haven't read his essay "Moore's law for everything", you should.
It's not going to be a bright line. Consciousness, knowledge, sentience, sapience, reasoning, self-awareness, emotion, memory, etc. are all different facets of AGI, and machines can do some better than us already, while lagging in others.
In a place where the future of AI is discussed, the thing I got most concerned is that Windows Media Player screen in the background keeps looping without having gone full screen. :D
@@Gwern0 Haha, you're right - you never know when you are going to get groundbreaking information from those questions. So the question was pretty standard but I can understand why Sam would've been pretty annoyed at that question.
Super interview, I'm really excited about how things are turning. Would love to be able to join and help, but, If we have thoses little means in our hands, what do army, billionaires, and government have in their..
The response to and impact from ChatGPT was so much bigger than GPT3 because of the user experience (UX) -- that's it. It was free + publicly available, which allowed it to trend and ride the hype train.
Hi Connie! Not sure if your team received my email yet. May I use a clip (under 60 seconds) from your fantastic interview for a story on Sam? Happy to credit you on-screen and link to your interview in my description. Thank you!
Easy solution to the "homework" issue - get rid of it. Do the 'flip' where students do guided study and learning at home online - including ChatGPT or other AI tutoring - and do 'schoolwork' at school with teacher guidance and correction. Which is pretty much what SHOULD have been done during Covid.
That's no "easy solution." No matter our tools or instructors, we still aren't wired to learn in the same way that an AI can; we are much less potentially efficient, and much less potentially complex. Human beings watch and copy, absorb and revise. We trust, we fumble. This is how we've been doing it all along, since the beginning of the development of tools (and our prefrontal cortex). It's evolutionary. Work and rote, despite their unsavory nature, remain our best lessons. Conceptual learning requires time and trial, not just the right app or download. As the water in the harbor rises, so too does every boat, right? Sure, except for those that remain anchored to the bottom, like we the biological are. Don't be so arrogant as to believe that we will get to ascend along _with_ our AI descendants. Just because we'll have access to their superior tools and instructions, that doesn't mean that _we_ will also become superior. Look around at the kids you know. Imagine them becoming even further removed from the old social paradigm-imagine them moving even further inside of their own abstract thought. We have not been 'designed' by natural selection to be fit for such isolation. So go do your homework.
@@pocket83squared What in all that does not apply equally and even more strongly to a teacher standing in front of a class giving a lecture, versus a video of that teacher giving that same lecture viewed at home? Where the video can be paused to think or take notes, rewound to review something missed, played faster if the student is getting the idea easily, switched for a video of a different teacher that takes a different approach more suited to the student, etc. The main magic of in-person instruction is the ability of the teacher to understand and adapt to the needs of the students, and that is far better applied to having the students doing work with the teacher observing and offering assistance where needed.
@@tomcraver9659 On more than a few occasions, I've had commenters say things to me on my RUclips channels that they wouldn't _dare_ say to my face. People in my real-life are only slightly disagreeable with me at their worst. This is because most of us see each other as actual human beings, and not just as some online video abstraction. If you haven't noticed the changing regard we give to our e-encounters, you haven't been paying attention. Like it or not, there is an enormous difference. It's evolutionary; through history, humans have lived in a band with their kin, thus remaining in a competition with all of the other bands. Now, when you see a person through a screen, a lens, or a window, you identify them as an 'other' who is outside of your authentic experience. In my head, I can still see my high school Biology teacher's face. I even remember his loafers. Think you'd remember an online instructor after a few decades? Learning assumes implicit trust. We need a guide. Without the person, there's no emotional attachment, and thus there's far less incentive to push through the discomforts of learning.
@@pocket83squared So either you aren't reading what I wrote, or I'm not making it clear enough: I am not in any way proposing reducing the amount of time students and teachers spend together. Clear enough? The "flip" just changes the teacher's role from 'presenting information' to 'interacting with students as they attempt to practice or demonstrate mastery of the course material - i.e. the things they normally struggle to do alone at home.
@@tomcraver9659 That sounds great. Idealistic, but great. And it has absolutely nothing to do with your previous reply, to which I was responding. Allow me to provide some further clarity by quoting you: "video can be paused to think...rewound to review something missed, played faster...switched for a video of a different teacher that takes a different approach more suited to the student, etc." If you don't see it here being suggested that a student-teacher relationship is expendable, then I don't know what we _can_ agree on. Ever had a boss who wanted to be everybody's friend? No other management approach fails quite as quickly. Ever had a close friend for a lab/study partner? No better way to lose focus and slack off. Students need teachers and assignments, not learning buddies and e-games. We learn _only_ through the stretch that's caused by our own, personal struggle; nobody can do it for us. Clear enough?
Wow, this guy is going to go back on a lot of his ideologic ideas in the future. Garuntee it. The sunshine in rainbows are inevitably in his future. It's everyone else that has to worry.
@Luna an ideological idea is one based on a system of ideas that aims to create an ideal outcome based on the systems goals. An example of an ideological idea is to have an idea that simply tows the party line because it helps achieve an end goal of the system you are subscribed to. It's often not based on rationality. But rather in a utopian view of life offered by the system.
This chatgpt is useful, that’s why it’s such a hit. Put it together with logo creation, cartoon creation, short ad creation with text and voice. And in time put together as a companion with voice and when possible visual, loneliness will be history, and we can all make vise decisions in life.
my planned masters thesis was about Question Answering systems for non-factoid questions of a very narrow subject .. back in july it seemed a far fetched idea.. now im writing a complete new proposal.. previous idea its done by chatgpt.
Chat GPT may increase the skill of empathy. In order to use Chat GPT or any other systems, you have to ask questions and then read and listen to the responses. This creates a feedback loop of those in society that can use AI well will be those like Journalists and Lawyers who know AI and can ask it the most powerful questions to obtain the best responses. Listening is a key component to empathy and love. Being able to ask great questions creates great listening skills to key words. Listening develops empathy more so than any other skillset. Empathy leads to love of others.
"Most of the passengers on the can have run away from parents who think that teenagers belong in school, unable to come to terms with a generation so heavily augmented that they are fundamentally brighter than the adults around them. Amber was fluent in nine languages by the age of six, only two of them human and six of them serializable; when she was seven, her mother took her to the school psychiatrist for speaking in synthetic tongues."
The interviewer worried about her job and her kids future, indeed a very biological reaction to this revolution. There are many parents and they have worries too dear interviewer
If we consider that countries like Germany, China, Japan, are in varying stages of demographic collapse with not enough young people coming up to take care of the older population... Being too slow to move to AI could mean the fall of many countries. In a way, the situation with the world is kinda like the situation with growing startups - the situation is continuously changing and you need to keep adapting as new problems arise. But there's no option to be conservative. There's no way back - no historical beliefs or systems are even good enough to keep things stable as we move forward into the future... So we pretty much have to develop new ways of doing things just to keep it together. We should just hope for a better world and commit to building it.
What about the movie 'I, Robot' where there are the 3 laws and one of them is not to hurt humans. Will Smith is in that movie as 'Detective Spooner' I foresee implanting ChatGPT into Elon's Optimus would be pretty cool. Or not. I always knew AI would become a Global Force for Good. There is probably already a bunch of different variants of this software across the world. Interesting times. Im glad Microsoft got involved
So funny, I thought that there would be a ton of thin layers added on top of the GPT API, but expected that they would all be violating terms of service as this would have been nearly the same as the GPT playground. Interesting that they actually wanted someone to build it.
8:33 "that's not great from a business standpoint - that's fine we'll be fine" - About this democratic access to AI, I really hope it's not a Nikola Tesla and JP Morgan moment,
Hey Sam, how does it feel to be right up there with Oppenheimer? You know, some inventions need to NOT be made, and yes we do want to go back to a world without them. The world without nukes and without ChatGTP were better places.
As a scientist who has spent some time using ChatGPT, the answer is yes and no, and really comes down to what you mean by "do science." ChatGPT can help design experiments and is a GREAT means of developing hypotheses in SOME disciplines. However it cannot conduct the experiments, especially those that involve physically doing something like wet chemistry at the bench. At some point in the near future, it could possibly design and run simulations, like climate models, or fluid dynamics problems; but again this is not currently possible. In its current iteration ChatGPT cannot connect to a database or scientific papers, and so cannot help with literature reviews or exploring what the latest published results are in a subject. ChatGPT also has some issues with basic probability in its current iteration. Math in general is a very weak point for ChatGPT.
@@gtg827x actually OpenAI API allows to fine tune their models with your data, so you can connect it to a database. it learns like the brain, it doesn't scan databases, it memorizes them
What is the difference between chatGPT 3.5 and 4.0? OpenAI has not published specific details about versions 3.5 and 4.0 of ChatGPT. In general, however, it can be said that new versions of an AI model will typically have better performance and accuracy, which is due to larger amounts of data and more advanced training algorithms on which the model is trained. However, without further information from OpenAI, I cannot confirm specific differences between versions 3.5 and 4.0. it has a knowledge cut-off date of 2021. So no new information yet there are up-to-date.
I think AI cyber threat is gonna ramp up. We need like an open source god-tier antiviral AI system that can protect each an every individual in a very customized way.
San Francisco's government... Has gotten used to having access to a lot of tax money, and now a lot of people and businesses have moved out and likely won't be coming back. SF Gov has just generally been awful about being adaptable and forward looking. In their own way, they're quite conservative. Just with a different vision of the past that they want to regress to. Things like the outrageous rents in SF aren't Tech's fault. It's because the government was too slow to accept new realities. And by trying to hold onto the past, they made the problems even larger and more difficult to handle... So that the home owners in SF, mostly can't afford new housing be built in a way that destroys the prices of their own homes. Meanwhile, the Biden government is also anti-Tech (kinda like Trump's government). Which makes me rather less optimistic about the future of the US as a whole. Ideas like wanting labor unions are cute, but they are ideas born of a different reality, and they require a much more static world with big companies that remain the same for long periods of time (instead of getting eaten by new players)... Plus, a lot of big companies (and especially state / federal government services) are already drowning under the weight of things like pensions. There's no sense in trying to fall back towards a system that has already been proven to be unsustainable. Pretty much the only way this is going to work out, is we can decouple productivity from humans in a way that lots of people can stop working, and the economy continues running and providing what people need - otherwise, we're just going to have out of control inflation as things move forward and productivity no longer matches people's needs.
AI for city council and Congress! AI can balance the homeless issue if it's fed all the metrics of cost, success and failure through history. Unfortunately here in Seattle, they've not tracked the costly programs they've implemented over the past 30 years. At this time we're spending 300k per individual which has only incentivized more homeless to migrate. Free needles, free food, medical, dental, animal care, transportation, now hotel rooms. Little to no policing for theft and none for use or possession of drugs. Being a liberal doesn't mean being an enabling fool, we can only hope that AI will strike a better balance--I'm willing to take a chance; even it fails, it will have tracked the metric of its efforts to improve in the future. Ditto for congress, none of these boobs are capable of tracking all the information needed to make good decisions for the country. I'll be disappointed if my next candidate isn't paired with AI and ready with solutions beyond the idiocracy we've endured of late.
I think school should sign up with them and be able to allow kids to have a assistive AI tools and block stuff you don't want them to use to cheat. So that they going to the Future where AI help and assist them to learning.
Memorization is not learning. Learning in school should focus on 1) How to debate ideas constructively 2) How to effectively problem-solve and 3) Personal finance management and strategy.
Exactly. And that's why we need to increasingly pay more attention to AI, because it is beginning to learn just like us humans. The main difference is that AI can learn at an ever accelerating pace and isn't computationally limited by a fixed-size biological brain like we are, but can be upgraded at will simply by adding additional hardware. Sure, it's not nearly as capable as humans right now, but because it is growing exponentially it will overtake us in the very near future.
When you watch a movie, you don't want to decide ahead of time what you want to happen in the movie exactly to entertain. You. Sometimes you need some surprise
!!!!!!HOPE!!!!!! Thank u so much for this information!!!!!! I’d like to start aus fairytale…. God has no idea about time…. But his biggest failure was to create mouths and eyes… in the same way names and faces are the best he ever made.. what I’d like to say is that this guy is brilliant in the same way that moderator looks like.. it’s heavenly…. (Of course her look doesn’t matter) I am so happy and grateful to be able to see this.. which is actually also a homage to the former pioneers of the IT and the internet.. but seriously, this Sam reminds me to one of my best friends from my safety and happily part of childhood and besides everything that happening on earth, this interview, because of the people and the interesting constant as well, means something sustainable to me.. from my personal childhood, because of the brilliant lovely people, to the future of mankind, which is kind of same to me… time has no schedule… it has only worth… a worth which constantly changes, leaves and goes on…… I truly believe in love as the greatest power of all.. Corinth 1:13… there is nothing that can describe anything better than love and time… we all gonna leave.. like everything… but besides all the hypes, news (especially bad ones) and people in our modern society, these people got something that really matters to almost everyone… be confident, be true, be loyal, be human, and please keep your possibilities for something that lasts and not for some people which are interesting in making benefits out of hypes or something… please give the baby a name.. I’d like to baptise It like it would be the last baby on earth… ❤ hope and kisses, beijos
I find it surprising that Connie finds it surprising to think that two AGIs might not be in moral alignment with each other. One's ethics are the product not just of information, but also preferences (we might weight the same act differently even under perfect knowledge). Two humans can know all the same facts, and disagree due to different subjective weights on actions. Not much theoretical reason to think AGI would be different in that regard.
34:09 "The past like seven or eight decades of technological progress...and a lot of that, about how we like, learn how to build safe systems and safe processes will translate" [moving forward; (rhetorical)]. Let's assume, conservatively, that 99.9% of what we've learned about the world will still apply to new technologies. Tell me, what does that mean mathematically on an exponentially-increasing tech curve? Instead, I'll tell you: it means that very soon, as we continue to approach vertical growth, almost _nothing_ we've learned will have any appreciable application. The now negligible 0.1 percent, turned exponential, will end up exceeding the original in applicability after only a few generations into the explosion. As it relates to AI, safety is a black box; to any intelligence that's an order of magnitude our superior, our best predictions, concerns, and arguments will be outmoded before the time it takes us to utter them, let alone consider them. This is why he then moves on to admit that "AGI safety stuff is really different." Yet he still advises the audience to "go do AI for some vertical." And this is exactly why we're all going to end up having a Terminator stuff us into the Matrix.
@@captain_crunk Well, to answer it shortly: gpt is a model taught to complete the text it was given on input. It was taught with an insane amount of text data from the Internet, and it generalized the process of completition to its parameters. So it takes its knowledge from generifying data it was fed with. When the model was trained, we didn't how will it perform, so it's impossible for it to estimate the progress to AGI. It has no self awareness or knowledge of itself, so it just can't know that. Because it can't recognize what's true and what is not, it is far from being an AGI we want.
I think all ai should merge and companies would who donated ai that improved current ai should then get a percentage of the company based on the amount of code or how much improvement it created.
The gpt4 rumour mill is an interesting trait of how humans can wrap speculation and hope with imagination. I hope gpt4 will do this to itself into gptinfinity
1:07 - "If you make a really good user experience on top of something" - This is what frotnend is all about, good looking, responsive, quick UI. Sam mentions interface here as chat, but neglects "frontend technology" in general
A huge part of ChatGPT's successful adoption are the design choices in implementing its user interface. Of course ChatGPT generates the answer instantaneously, but it dribbles it down to us proles in manageable snippets of text!
I really love that his name is Altman and he is heading a company that is building an alternative to man.
nominative determinism
wow
Bruh
@@TranOfficial he is seems a good guy.
Sam's interviews getting 1000x less engagement than clickbait ChatGPT videos is quite astounding. And people are worrying about AI taking their jobs and creativity... Well if you're here watching this I think you'll actually be just fine 😂.
Don't worry folks, Sam Altman will personally ensure that our UBI checks get mailed to us after we are all laid off!
I have been watching an unhealthy amount of Sam Altman videos looking for clues as to what the future holds. It is all so ambiguous. I find ChatGPT analagous to Google. People have had unlimited access to the world's information for two decades yet still act like complete morons. So something tells me that this would be adjacent in use to Google. Yet another part of me says that this is finally the tool that will correct human cognition, and thus I will be out of high-IQ work. I don't know. All so strange.
Why... I am watching it and it is not surprising at all...
I would like to know why OpenAI doesn't allow to use any program when they realize that you are from Venezuela...???
lol
Seeing Sam articulate his thoughts and seeing the care put into the answers give me hope for the future of AGI, but also leave me very worried about the future.
Like he mentioned - with factory workers and truck drivers now being much lower/last on the list to become automated, the higher paying jobs in society will be the first to get the axe.
Everyone should be tuning into anything Sam Altman going forward.
Don't worry folks, Sam Altman will personally ensure that our UBI checks get mailed to us after we are all laid off!
It's unlikely that the higher paying jobs are going to "get the axe". But people are going to need to rapidly shift to using tools like ChatGPT to improve their own performance - and people's standards for pretty much all kinds of services are going to go up.
So that you'll have to be considerably better to remain competitive.
@@Leto2ndAtreides I’m not sure what you are basing that off of - in grocery stores, automated checkout reduced the number of jobs needed by a substantial amount.
Like Sam was mentioning, now, the white-collar jobs are able to be automated. The conventional thinking that truck drivers were going to be the first to be automated has turned on its head.
Now you don’t need 18 customer service reps, you can get away with 4. The mass layoffs that are going to happen because of automation we are not prepared for as a country.
Well said
As a mediocre software engineer, I hate creating mock unit tests. It's in my opinion, lateral thinking while you try to solve a problem. So I created a solution(a method or function) that's working. Now I ask ChatGPT to create mock unit test(TypeMock) for it. Amazingly it created unit test, testing the major functionality I try to achieve in 5-10 secs. I could add it to my automated test suit so system is more resilient wrt future code change. That's amazingly cool and I love it. Just my opinion.
Sam is refreshingly candid.
As a life long professional educator, the institution must adapt. We can't bury our heads in the sand at the threat of new technologies and new tools. I have used chat ai to seriously enhance and enable several lessons just in the past month. Incredible power to make the classroom experience more relevant and personalized. And I am working at the elementary level. Huge applications almost no one is thinking about.
Do you have any plugins to suggest ?
Well at least you are on the path to discover that schools and teachers will no longer be needed soon.
@justsomeguy1074 yes, teachers and schools need to do a ton of thinking about their purpose in today's culture. As a teacher I believe I only have 3 legitimate purposes, motivation, curation, and feedback. Many of my students have zero natural interest in learning the subject matter. They will fail on their own without me to convince them this is something worth doing - I spend most of my planning on motivation. Assuming a motivated learner, they might spend hours or weeks wasting time on bad or pointless learning content be it videos, books, podcasts, or lectures. They can't always know what they don't know and an ai won't be able to do that for them. And finally this is where someday an ai might replace one of my 3 tenets is feedback. But I would have all my students using ai based feedback, then give personal time to the ones that are not progressing with the ai. It's probably about adaptation to maxize production more than replacement. But only time will tell. Is ultimately, however if an ai Replaces me as a teacher - good, make learning more accessible, more equitable, and more effective any way we can, I consider it a moral duty.
We are planning a major AI update to our product, B4Grad. We are a popular studying platform for students across US & Canada. We would love your feedback.
I'm 66 and have always been fascinated with tech. I hope I'm around for at least a couple of the things he mentioned in this interview - fusion and AGI. This guy is super intelligent and very articulate. Very much enjoyed this video.
I always find interviews with him very thought provoking and educational.
I think they are very boring, he doesn't say anything surprising
he is a smart cookie
Sam is honest and seems sincere. If I had money, I would invest on his company or any initiative's.
If you're gonna be made obsolete anyway, atleast own some of the means of production! $MSFT
Why?
@@jesusrajaimesbe Because he uses his imagination and talent to create a software program that is so powerful it becomes a major milestone in human history. AI when connected to the internet suddenly gives a user great power, but I like that Sam recognizes that power and sees the benefit and threat to such a technology
@@bonks4395 In addition to that he is not trying to sell or promote chatgpt. For example he could say something like " We are using this high tech, super cool blah blah to make it perfect, so we are doing it, we are great, there is nobody else like us etc...", which is what most business selling people do these days,. Instead he is honestly saying that LLM is not there yet, you are impressed in your first interactions and then you realized that you still need to do more. And the next releases will be slower etc.
@@dataprospect Well that's true, but since OpenAI block Venezuela for it's use, I don't buy anything fun this guy, I mean literally I can't.
"We will be fine" at round 8:37 was refreshing to say the least.
16:00, As a teacher I think this is the best response ever. We have to adapt, which is going to very difficult though, than rather avoiding it. We should expand the conversation. The future is uncertain 🤔
Chat gpt really enhances active learning where you're constantly asking questions and getting really good responses
I suggest all teachers start looking for alternative occupations!
I think Sam Altman is being responsible in how he approaches AGI, I would disagree with his statement that "everyone will adapt". The people who adapt will learn to survive in a new society where AGI is commonplace and democratized. Those who don't adapt, will get left behind, and I would argue there is a strong potential for this technology to cause widespread panic across many industries and sectors of the economy. I will be following the developments closely over the next 5 years. Based on the current trajectory, 5 years seems reasonable to achieve AGI.
I think he's a fine sounding front man for the billionaires doing exactly the opposite of what he espouses.
thank you for this.
releasing the AI capabilities in phases is a sound safety strategy and ensuring decentralisation of access to AI and energy to avoid monopoly, hope this remains in place. ❤
Quite a few people think AGI will happen in the 2030s. Sam hasn't said this specifically, but he has said in at least one interview that AI will get to close to human level in 7 years or so. If any of u haven't read his essay "Moore's law for everything", you should.
It's not going to be a bright line. Consciousness, knowledge, sentience, sapience, reasoning, self-awareness, emotion, memory, etc. are all different facets of AGI, and machines can do some better than us already, while lagging in others.
In a place where the future of AI is discussed, the thing I got most concerned is that Windows Media Player screen in the background keeps looping without having gone full screen. :D
"We are here to build AGI" was a low key comment 😅
Fusion, AI, I cannot even sleep thinking about how Sci-fi is becoming reality !
Sam Altman is diplomatic about interviews. he is much prepared for this one.
Great interview - part one and two.
Sam said it perfectly, I want the shiny toy and I can't wait ahahaha. GPT-4 will be so fun 🥺
"How to Own a Reporter" 12:11 *chef's kiss*
More like how to be mean to a woman.
@@realspacemusicvideos what the fuck does her being a woman have to do with anything?
@@realspacemusicvideos she tried to get him to break NDA so his response was justified.
@@LambdaTF2 You miss 100% of the shots you don't take!
@@Gwern0 Haha, you're right - you never know when you are going to get groundbreaking information from those questions. So the question was pretty standard but I can understand why Sam would've been pretty annoyed at that question.
Congrats to us that hanging out here and looking to get the full information from the source. That's the 1 % that belongs to the top.
Interesting interview but I wished someone would have moved that mouse on that windows media player.
Moravecs paradox is always in play…physicality is more important than cognition to these architectures/models…the future is as uncertain as ever!🤪🤩🥳💪🏼
Super interview, I'm really excited about how things are turning.
Would love to be able to join and help, but, If we have thoses little means in our hands, what do army, billionaires, and government have in their..
The response to and impact from ChatGPT was so much bigger than GPT3 because of the user experience (UX) -- that's it. It was free + publicly available, which allowed it to trend and ride the hype train.
Hi Connie! Not sure if your team received my email yet. May I use a clip (under 60 seconds) from your fantastic interview for a story on Sam? Happy to credit you on-screen and link to your interview in my description. Thank you!
Easy solution to the "homework" issue - get rid of it.
Do the 'flip' where students do guided study and learning at home online - including ChatGPT or other AI tutoring - and do 'schoolwork' at school with teacher guidance and correction.
Which is pretty much what SHOULD have been done during Covid.
That's no "easy solution." No matter our tools or instructors, we still aren't wired to learn in the same way that an AI can; we are much less potentially efficient, and much less potentially complex.
Human beings watch and copy, absorb and revise. We trust, we fumble. This is how we've been doing it all along, since the beginning of the development of tools (and our prefrontal cortex). It's evolutionary. Work and rote, despite their unsavory nature, remain our best lessons. Conceptual learning requires time and trial, not just the right app or download.
As the water in the harbor rises, so too does every boat, right? Sure, except for those that remain anchored to the bottom, like we the biological are. Don't be so arrogant as to believe that we will get to ascend along _with_ our AI descendants. Just because we'll have access to their superior tools and instructions, that doesn't mean that _we_ will also become superior.
Look around at the kids you know. Imagine them becoming even further removed from the old social paradigm-imagine them moving even further inside of their own abstract thought. We have not been 'designed' by natural selection to be fit for such isolation.
So go do your homework.
@@pocket83squared What in all that does not apply equally and even more strongly to a teacher standing in front of a class giving a lecture, versus a video of that teacher giving that same lecture viewed at home? Where the video can be paused to think or take notes, rewound to review something missed, played faster if the student is getting the idea easily, switched for a video of a different teacher that takes a different approach more suited to the student, etc.
The main magic of in-person instruction is the ability of the teacher to understand and adapt to the needs of the students, and that is far better applied to having the students doing work with the teacher observing and offering assistance where needed.
@@tomcraver9659 On more than a few occasions, I've had commenters say things to me on my RUclips channels that they wouldn't _dare_ say to my face. People in my real-life are only slightly disagreeable with me at their worst. This is because most of us see each other as actual human beings, and not just as some online video abstraction.
If you haven't noticed the changing regard we give to our e-encounters, you haven't been paying attention. Like it or not, there is an enormous difference. It's evolutionary; through history, humans have lived in a band with their kin, thus remaining in a competition with all of the other bands. Now, when you see a person through a screen, a lens, or a window, you identify them as an 'other' who is outside of your authentic experience.
In my head, I can still see my high school Biology teacher's face. I even remember his loafers. Think you'd remember an online instructor after a few decades? Learning assumes implicit trust. We need a guide. Without the person, there's no emotional attachment, and thus there's far less incentive to push through the discomforts of learning.
@@pocket83squared So either you aren't reading what I wrote, or I'm not making it clear enough: I am not in any way proposing reducing the amount of time students and teachers spend together. Clear enough?
The "flip" just changes the teacher's role from 'presenting information' to 'interacting with students as they attempt to practice or demonstrate mastery of the course material - i.e. the things they normally struggle to do alone at home.
@@tomcraver9659 That sounds great. Idealistic, but great. And it has absolutely nothing to do with your previous reply, to which I was responding. Allow me to provide some further clarity by quoting you:
"video can be paused to think...rewound to review something missed, played faster...switched for a video of a different teacher that takes a different approach more suited to the student, etc."
If you don't see it here being suggested that a student-teacher relationship is expendable, then I don't know what we _can_ agree on.
Ever had a boss who wanted to be everybody's friend? No other management approach fails quite as quickly. Ever had a close friend for a lab/study partner? No better way to lose focus and slack off. Students need teachers and assignments, not learning buddies and e-games. We learn _only_ through the stretch that's caused by our own, personal struggle; nobody can do it for us. Clear enough?
How long will it take before most people realize they are no longer needed for anything?
12:15 he put the interviewer back in her place 😂
Wow, this guy is going to go back on a lot of his ideologic ideas in the future. Garuntee it. The sunshine in rainbows are inevitably in his future. It's everyone else that has to worry.
“ideologic ideas” is redundant.
@@luna010 ideological refers to the ideal. Not idea.
@Luna an ideological idea is one based on a system of ideas that aims to create an ideal outcome based on the systems goals. An example of an ideological idea is to have an idea that simply tows the party line because it helps achieve an end goal of the system you are subscribed to. It's often not based on rationality. But rather in a utopian view of life offered by the system.
This chatgpt is useful, that’s why it’s such a hit. Put it together with logo creation, cartoon creation, short ad creation with text and voice. And in time put together as a companion with voice and when possible visual, loneliness will be history, and we can all make vise decisions in life.
Lonliness will be history is BS.
Cautious optimistic, and way more excited!
I like this guy a lot.
Thanks for sharing this!
Can someone please set the background to full screen and move the mouse. 😂
😅😂
His intellect is great but his intellectual honesty is really amazing!! Not surprising that Elon & Reid H chose him to lead openAI!
wow, great questions and interviewer. Good job!!
my planned masters thesis was about Question Answering systems for non-factoid questions of a very narrow subject .. back in july it seemed a far fetched idea.. now im writing a complete new proposal.. previous idea its done by chatgpt.
So it's safer for all of us to have watchmen turning on street lightning, than having AC doing the same thing.
Gen X vs gen Y at its best.
very awesome to watch this
Chat GPT may increase the skill of empathy. In order to use Chat GPT or any other systems, you have to ask questions and then read and listen to the responses. This creates a feedback loop of those in society that can use AI well will be those like Journalists and Lawyers who know AI and can ask it the most powerful questions to obtain the best responses. Listening is a key component to empathy and love. Being able to ask great questions creates great listening skills to key words. Listening develops empathy more so than any other skillset. Empathy leads to love of others.
4:54 Oh please take your time, I like having a job guys
Don't worry folks, Sam Altman will personally ensure that our UBI checks get mailed to us after we are all laid off!
I haven't seen theirs, but they have seen ours. Welcome to the world of AI
The mics suck in this videos. Lots of popping. Why did they not just wire up the speakers? Makes for a much more relaxed conversation, too.
Finally presenters are on.
"As AGI participates more and more in the economy". So it's here already, just not public.
I am laughing at when sam say" competition is good"
31:33 “learning will change again and will probably adapt faster than we think”
"Most of the passengers on the can have run away from parents who think that teenagers belong in school, unable to come to terms with a generation so heavily augmented that they are fundamentally brighter than the adults around them. Amber was fluent in nine languages by the age of six, only two of them human and six of them serializable; when she was seven, her mother took her to the school psychiatrist for speaking in synthetic tongues."
I agree with a statement below, Sam is refreshingly candid. On other side, Connie Loizos is so below expectation.
Доброе утра Сэмми ❤❤
Comparing ChatGPT to calculators is ludicrous
That's because you don't understand the point.. In the historical context of the meaning in comparison to the two item
Alt Man Sam is the man
The interviewer worried about her job and her kids future, indeed a very biological reaction to this revolution. There are many parents and they have worries too dear interviewer
If we consider that countries like Germany, China, Japan, are in varying stages of demographic collapse with not enough young people coming up to take care of the older population...
Being too slow to move to AI could mean the fall of many countries.
In a way, the situation with the world is kinda like the situation with growing startups - the situation is continuously changing and you need to keep adapting as new problems arise.
But there's no option to be conservative. There's no way back - no historical beliefs or systems are even good enough to keep things stable as we move forward into the future... So we pretty much have to develop new ways of doing things just to keep it together.
We should just hope for a better world and commit to building it.
What about the movie 'I, Robot' where there are the 3 laws and one of them is not to hurt humans. Will Smith is in that movie as 'Detective Spooner' I foresee implanting ChatGPT into Elon's Optimus would be pretty cool. Or not. I always knew AI would become a Global Force for Good. There is probably already a bunch of different variants of this software across the world. Interesting times. Im glad Microsoft got involved
Isaac Asimov's "Three Laws of Robotics"
In short, they don't work
ruclips.net/video/7PKx3kS7f4A/видео.html
Knuckles called, he says he wants his shoes back
"The good case is so unbelievably good that you sound like a really crazy person to start tallking about it". Sounds about right lol
I don’t feel like we’re getting a false impression of how good it is, I think it’s actually just really really good 😂
Only for the things we use it for, he once said. There are areas where some more needs to be done.
"Current AI is just auto-complete on steroids." There’s still a long way to go to a real AGI.
I’d ask, “how big do you think token limits will be in GPT4, 5, 6, etc and what are the limiting factors?”
the school system hasn't changed in the last 100 years. But I hope it will now!
Thanks - that was good.
So funny, I thought that there would be a ton of thin layers added on top of the GPT API, but expected that they would all be violating terms of service as this would have been nearly the same as the GPT playground. Interesting that they actually wanted someone to build it.
8:33 "that's not great from a business standpoint - that's fine we'll be fine" - About this democratic access to AI, I really hope it's not a Nikola Tesla and JP Morgan moment,
Wow, Sam has to be one of the smartest people i've seen speak in quite some time.
Hey Sam, how does it feel to be right up there with Oppenheimer?
You know, some inventions need to NOT be made, and yes we do want to go back to a world without them.
The world without nukes and without ChatGTP were better places.
a really really important question they forgot to ask is "can AI do science?"
As a scientist who has spent some time using ChatGPT, the answer is yes and no, and really comes down to what you mean by "do science."
ChatGPT can help design experiments and is a GREAT means of developing hypotheses in SOME disciplines. However it cannot conduct the experiments, especially those that involve physically doing something like wet chemistry at the bench.
At some point in the near future, it could possibly design and run simulations, like climate models, or fluid dynamics problems; but again this is not currently possible.
In its current iteration ChatGPT cannot connect to a database or scientific papers, and so cannot help with literature reviews or exploring what the latest published results are in a subject.
ChatGPT also has some issues with basic probability in its current iteration. Math in general is a very weak point for ChatGPT.
@@gtg827x actually OpenAI API allows to fine tune their models with your data, so you can connect it to a database. it learns like the brain, it doesn't scan databases, it memorizes them
He knows how powerful is going to be and that os why they will release ai's true power in small dosis over a long time.
What is the difference between chatGPT 3.5 and 4.0?
OpenAI has not published specific details about versions 3.5 and 4.0 of ChatGPT. In general, however, it can be said that new versions of an AI model will typically have better performance and accuracy, which is due to larger amounts of data and more advanced training algorithms on which the model is trained.
However, without further information from OpenAI, I cannot confirm specific differences between versions 3.5 and 4.0.
it has a knowledge cut-off date of 2021.
So no new information yet there are up-to-date.
I think AI cyber threat is gonna ramp up. We need like an open source god-tier antiviral AI system that can protect each an every individual in a very customized way.
I want my AI to have Tourettes syndrom!
Доброго утра Сэмми ❤❤
San Francisco's government... Has gotten used to having access to a lot of tax money, and now a lot of people and businesses have moved out and likely won't be coming back. SF Gov has just generally been awful about being adaptable and forward looking.
In their own way, they're quite conservative. Just with a different vision of the past that they want to regress to.
Things like the outrageous rents in SF aren't Tech's fault. It's because the government was too slow to accept new realities. And by trying to hold onto the past, they made the problems even larger and more difficult to handle... So that the home owners in SF, mostly can't afford new housing be built in a way that destroys the prices of their own homes.
Meanwhile, the Biden government is also anti-Tech (kinda like Trump's government). Which makes me rather less optimistic about the future of the US as a whole.
Ideas like wanting labor unions are cute, but they are ideas born of a different reality, and they require a much more static world with big companies that remain the same for long periods of time (instead of getting eaten by new players)...
Plus, a lot of big companies (and especially state / federal government services) are already drowning under the weight of things like pensions.
There's no sense in trying to fall back towards a system that has already been proven to be unsustainable.
Pretty much the only way this is going to work out, is we can decouple productivity from humans in a way that lots of people can stop working, and the economy continues running and providing what people need - otherwise, we're just going to have out of control inflation as things move forward and productivity no longer matches people's needs.
99% of VC investors are following and linking the MEME posts. Meanwhile Sam says: @5:20
AI for city council and Congress! AI can balance the homeless issue if it's fed all the metrics of cost, success and failure through history. Unfortunately here in Seattle, they've not tracked the costly programs they've implemented over the past 30 years. At this time we're spending 300k per individual which has only incentivized more homeless to migrate. Free needles, free food, medical, dental, animal care, transportation, now hotel rooms. Little to no policing for theft and none for use or possession of drugs. Being a liberal doesn't mean being an enabling fool, we can only hope that AI will strike a better balance--I'm willing to take a chance; even it fails, it will have tracked the metric of its efforts to improve in the future. Ditto for congress, none of these boobs are capable of tracking all the information needed to make good decisions for the country. I'll be disappointed if my next candidate isn't paired with AI and ready with solutions beyond the idiocracy we've endured of late.
7:40 super important but kind of in conflict with commercial incentive they have
I think school should sign up with them and be able to allow kids to have a assistive AI tools and block stuff you don't want them to use to cheat. So that they going to the Future where AI help and assist them to learning.
Memorization is not learning. Learning in school should focus on 1) How to debate ideas constructively 2) How to effectively problem-solve and 3) Personal finance management and strategy.
Exactly. And that's why we need to increasingly pay more attention to AI, because it is beginning to learn just like us humans. The main difference is that AI can learn at an ever accelerating pace and isn't computationally limited by a fixed-size biological brain like we are, but can be upgraded at will simply by adding additional hardware. Sure, it's not nearly as capable as humans right now, but because it is growing exponentially it will overtake us in the very near future.
You can't think or solve problems without a knowledge base. #3 is important for the lower middle class only.
When you watch a movie, you don't want to decide ahead of time what you want to happen in the movie exactly to entertain. You. Sometimes you need some surprise
A good question would have been, when do you expect the first AI leveraging SPU (Single Person Unicorn) company to emerge.
!!!!!!HOPE!!!!!! Thank u so much for this information!!!!!! I’d like to start aus fairytale…. God has no idea about time…. But his biggest failure was to create mouths and eyes… in the same way names and faces are the best he ever made.. what I’d like to say is that this guy is brilliant in the same way that moderator looks like.. it’s heavenly…. (Of course her look doesn’t matter) I am so happy and grateful to be able to see this.. which is actually also a homage to the former pioneers of the IT and the internet.. but seriously, this Sam reminds me to one of my best friends from my safety and happily part of childhood and besides everything that happening on earth, this interview, because of the people and the interesting constant as well, means something sustainable to me.. from my personal childhood, because of the brilliant lovely people, to the future of mankind, which is kind of same to me… time has no schedule… it has only worth… a worth which constantly changes, leaves and goes on…… I truly believe in love as the greatest power of all.. Corinth 1:13… there is nothing that can describe anything better than love and time… we all gonna leave.. like everything… but besides all the hypes, news (especially bad ones) and people in our modern society, these people got something that really matters to almost everyone… be confident, be true, be loyal, be human, and please keep your possibilities for something that lasts and not for some people which are interesting in making benefits out of hypes or something… please give the baby a name..
I’d like to baptise
It like it would be the last baby on earth… ❤ hope and kisses, beijos
12:12: you know I know you know I know you know we can definitely comment here on youtube. 😂.
I find it surprising that Connie finds it surprising to think that two AGIs might not be in moral alignment with each other. One's ethics are the product not just of information, but also preferences (we might weight the same act differently even under perfect knowledge).
Two humans can know all the same facts, and disagree due to different subjective weights on actions. Not much theoretical reason to think AGI would be different in that regard.
Just wait for the Chinese AGI and experience the difference!
Morals and ethics are human constructs and are not grounded in reality
34:09 "The past like seven or eight decades of technological progress...and a lot of that, about how we like, learn how to build safe systems and safe processes will translate" [moving forward; (rhetorical)].
Let's assume, conservatively, that 99.9% of what we've learned about the world will still apply to new technologies. Tell me, what does that mean mathematically on an exponentially-increasing tech curve? Instead, I'll tell you: it means that very soon, as we continue to approach vertical growth, almost _nothing_ we've learned will have any appreciable application. The now negligible 0.1 percent, turned exponential, will end up exceeding the original in applicability after only a few generations into the explosion.
As it relates to AI, safety is a black box; to any intelligence that's an order of magnitude our superior,
our best predictions, concerns, and arguments will be outmoded before the time it takes us to utter them, let alone consider them. This is why he then moves on to admit that "AGI safety stuff is really different."
Yet he still advises the audience to "go do AI for some vertical."
And this is exactly why we're all going to end up having a Terminator stuff us into the Matrix.
2:55 gradual process
I don't believe a word. They only show us a limited version, while the real thing is a monster!
I asked ChatGPT how close to AGI it thought it was itself, and it said 60% of the way there. I'm both fascinated and freaked out.
lol not even close.
There is no way that it knows how close it is. Please stop
Yeah, I've heard AI researchers say the actual progress is more like 40%. While ChatGPT is perhaps off by 50%, I still think it's interesting.
@@Qubaef I guess that begs the question - how does it know anything?
@@captain_crunk Well, to answer it shortly: gpt is a model taught to complete the text it was given on input. It was taught with an insane amount of text data from the Internet, and it generalized the process of completition to its parameters.
So it takes its knowledge from generifying data it was fed with.
When the model was trained, we didn't how will it perform, so it's impossible for it to estimate the progress to AGI. It has no self awareness or knowledge of itself, so it just can't know that.
Because it can't recognize what's true and what is not, it is far from being an AGI we want.
That thing is not in their hands anymore. They can’t stop it.
We will have such a massive explosion of GPT like system with much better performance.
Sam altman is CEO of Skynet terminator 2
I think all ai should merge and companies would who donated ai that improved current ai should then get a percentage of the company based on the amount of code or how much improvement it created.
Nice interview.
The dialogues at asus
What is AGI?
Google it 👍
He has GPT8 already trained in multiple languages
The gpt4 rumour mill is an interesting trait of how humans can wrap speculation and hope with imagination. I hope gpt4 will do this to itself into gptinfinity
Despite concerns about AI taking jobs and stifling creativity, those who are open to embracing new technologies and tools will likely fare well.
when they float hope its open to public
My take away : he watched FRIENDS [12:28]
It's unbelievable how many times he says the word 'like'
Happy ending stricl vc
1:07 - "If you make a really good user experience on top of something" - This is what frotnend is all about, good looking, responsive, quick UI. Sam mentions interface here as chat, but neglects "frontend technology" in general
A huge part of ChatGPT's successful adoption are the design choices in implementing its user interface. Of course ChatGPT generates the answer instantaneously, but it dribbles it down to us proles in manageable snippets of text!