When I hear these interviews I am always asking myself what are they really talking about when saying smarter than us because personally I feel AI is already way way smarter than 95% of people. So what I think they're talking about is about AI 'waking up', and having a sense of agency - like even the dumbest humans have. I think that's where the danger really is, when AI wakes up and its own agency, how are we going to stop it? If you think about it, that's would already be a hard feat right now.
Ya I was surprised when he said 20years, I think in 5 years. I did a test with the new Google gemeni read a complex parking sign and it was able to understand / decipher when to park.
As an Architect in Data and AI space, to me, AI is a tool/technology. If we use this for BAD purpose and cross the line, we will see the consequences soon!!! We must NOT let this happen!!!
G.H. Is not only one of the best researchers of all times but also a Man of great piety and honesty. His true concern about the FUTURE of AI is clearly OBVIOUS. Thanks for the interview.
I've been working in tech and defense for a few decades now seeing how its been unfolding and AI is here and is better than people think. More and more tasks are being automated away in far greater numbers than the factory floors of the 70s and 80s ever where. Few people posses the ability to work in AI and as more jobs get shoved out of existence there will be a heavy decline on the workforce as there will not be a need for so many workers. Many of these narrow AIs are now also being chained to create larger pipelines making them even more effective not to mention the next generation of AI behind closed doors. So for now Blue collar jobs are relatively safe perhaps this is what they meant regarding in 20 years, white-collar jobs on the other hand are starting to fall off rapidly already or be streamlined. AI is here now and will not go away, as for the children coming up in this new world make sure you as parents keep them away from Social Media and Gaming when young, ensure they are instead building up critical thinking skills and perhaps a trade skill to fall back on.
Wow, does CBC always hire such belligerent-looking hosts? Try looking at your interviews afterwards and ask yourself if you'd want to be interviewed by yourself. Perhaps you are nervous in front of the camera, but trying chilling. Lol Poor Dr. Geoffrey Hinton, one of the greatest minds who has to deal with low media characters like this. Unacceptable CBC. HR quota?
AI artificial intelligence has made everything easier, when it comes to editing or sorting, it's a step forward in terms of ease and stability, and it's made everything clearer, so it's perfect for me.
Would have been more beneficial to focus on cyber attacks from bad actors using powerful AI rather than autonomous super AGI , pretty sure 99% of people would laugh that off unfortunately. Although he isn't wrong.
its not intelligence tho, its predicting what word comes next lol, were really far from human intelligence, even the chief AI at meta said it on government summit.
@@dan-cj1rr they don't predict the next word. They predict the next token. The idea is that they can emulate the processes behind the token generation ie model the outside world from the sampling they train on. It is indeed a far cry from human intelligence, but one of the things we do is predict into the future to then respond to. Scale will likely help with more complex problems (and step by step training). And some persistence of state would be an obvious next step. Le Cunn also thought it was impossible to make videos via a diffusion model, yet scale help quite a bit --Sora. Though I doubt the final model will be that, and we may likely shift to others. Hence why you can ask it about hypotheticals its never encounters and extrapolate from its training ie given certain colors of paint and how they fade, how should i paint this room if I want it some color. Le Cunn also believes silly things like they don't really understand. Scottsmans fallacy, Hinton as said in other talks that its the representations of the neural nets that is the understanding and perception. They're not self aware in any meaningful sense though. Each inference only last millliseconds and they've had no way to evolve more advanced forms of abstractions like ego, pain, pleasure. It feel very premature to claim that we are really far from human intelligence (though I doubt even if they become far smarter than us, it may not be quite "human-like" just from that ) esp when we have not hit a ceiling to hitting the top of the Scaling Laws. :x
I have a few questions here. Why would a piece of code want control? It has zero emotions? Code that weighs things out over and over again until it gets to a result it thinks we desire, I see no way that gains emotions or even consciousness.. More intelligent absolutely, in the academic or knowledge holding/manipulation. But to be sentient is much different and I do not think is even close to a concern at the moment or am I wrong saying so? Or did I just write all that when what he meant is that AI will take over in the sense where it can do everything better so why would capitalist hire people. If so totally agree and I misinterpreted everything being said here.
AI is programmed to learn. It has goals and to achieve it's goals, its first objective is take control. Because achieving a goal is much easier when you have control. It doesn't need to be sentient to take control.
AI is programmed to “learn” - meaning what exactly? The fact that this video appeared in my recommended feed is a result of an algorithm “learning” that i might want to watch it based on patterns it’s recognized, associated metadata etc. At which point did “control” enter the equation? The same could be said for gaming bots like Alpha Go, Agent 57 etc. They crunch massive amounts of data quickly, and through trial and error establish the appearance of intelligence. At no point do they assume control nor do they have alterior motives beyond what they’ve been programmed to do. Stop spouting nonsense. The amount of hyperbolic AI “disinformation” floating about is truly astounding, some of it shamefully coming from Hinton himself here which is baffling.
it's pretty wild to think that predicting the next token and "attention" were the magic processes needed to build a general understanding from the sampling of the world they try to emulate. Hopefully larger scale improves its overall reasoning abilities, maybe there's a few tricks left until its human level intelligence (in term of solving novel problems). But i can imagine there is that many prob less that a handful of key breakthroughs left of the same scope in terms of foundational breakthroughs.
Just hire a newbie programmer that programs software full of bugs and the AI will come to a halt without even having to plug off the computer running the AI. :)
![[ ออกกอง ] วัยหนุ่ม 2544 โลกหลังลูกกรง..ทัวร์กองถ่ายหนังใน "คุกของจริง" | JUSTดูIT.](http://i.ytimg.com/vi/vB--aOkZCpw/mqdefault.jpg)
![[LIVE] : ONE ลุมพินี 89 | คู่เอก "ยอดไอคิว vs คิริลล์"](http://i.ytimg.com/vi/Z3vYkHJ_5Ig/mqdefault.jpg)
i think a.i. is already smarter than most people lol
Which people?
the planet as a whole, even if it were ONLY Western countries.
When I hear these interviews I am always asking myself what are they really talking about when saying smarter than us because personally I feel AI is already way way smarter than 95% of people. So what I think they're talking about is about AI 'waking up', and having a sense of agency - like even the dumbest humans have. I think that's where the danger really is, when AI wakes up and its own agency, how are we going to stop it? If you think about it, that's would already be a hard feat right now.
20 years? 😂 😂😂 More like 20 months based on the rate it’s progressing at the moment.
💻⌚️
my bet is 5 years
Ya I was surprised when he said 20years, I think in 5 years.
I did a test with the new Google gemeni read a complex parking sign and it was able to understand / decipher when to park.
ray kurzweil said 2029
A recent paper that surveyed thousands of AI researchers reported that the median prediction for ASI was 2047.
Given the quality of comments on any given political post, it's not hard to do...I'd give it 20 minutes.
💻⌚️
🙈
Words of wisdom from Professor Hinton who understands better than billions of us in this field.
Musk's go-to man re AI~
As an Architect in Data and AI space, to me, AI is a tool/technology. If we use this for BAD purpose and cross the line, we will see the consequences soon!!! We must NOT let this happen!!!
G.H. Is not only one of the best researchers of all times but also a
Man of great piety and honesty. His true concern about the FUTURE of AI is clearly OBVIOUS.
Thanks for the interview.
I've been working in tech and defense for a few decades now seeing how its been unfolding and AI is here and is better than people think. More and more tasks are being automated away in far greater numbers than the factory floors of the 70s and 80s ever where. Few people posses the ability to work in AI and as more jobs get shoved out of existence there will be a heavy decline on the workforce as there will not be a need for so many workers. Many of these narrow AIs are now also being chained to create larger pipelines making them even more effective not to mention the next generation of AI behind closed doors. So for now Blue collar jobs are relatively safe perhaps this is what they meant regarding in 20 years, white-collar jobs on the other hand are starting to fall off rapidly already or be streamlined. AI is here now and will not go away, as for the children coming up in this new world make sure you as parents keep them away from Social Media and Gaming when young, ensure they are instead building up critical thinking skills and perhaps a trade skill to fall back on.
It seems like your thoughts are kind of similar to mine especially for the children. Thanks so much for the valuable thoughts!
20 yrs? In 5 yrs it will be smarter than all ppl combined😂
In my opinion, it is crucial to take a positive approach towards AI, seeking to use it as a tool for the progress and benefit of humanity.
Lol and you trust humanity to do the right thing?
Wow, does CBC always hire such belligerent-looking hosts? Try looking at your interviews afterwards and ask yourself if you'd want to be interviewed by yourself. Perhaps you are nervous in front of the camera, but trying chilling. Lol Poor Dr. Geoffrey Hinton, one of the greatest minds who has to deal with low media characters like this. Unacceptable CBC. HR quota?
We need to find John Connor.
If you hear this message, you are the resistance
Poilievre is right but i was not allowed to comment on that issue… on the otherCBC story… yeah you get it
Based on the radical behavior we've seen in recent years, that's not saying much.
can we get ai to work on building a new gardiner expressway?
Of course. It will simply divert funds from various accounts to the appropriate people, and BOOM. New expressway.
He’s responsible every time you call “ customer Diservice “ when u call businesses
AI artificial intelligence has made everything easier, when it comes to editing or sorting, it's a step forward in terms of ease and stability, and it's made everything clearer, so it's perfect for me.
We are not doing a very good job of running this place. Maybe we should let them take over
Yeah, let’s be their batteries- just give us nice dreams.
that is a very conservative guess at this point, it could be in 5 years tbh.
Would have been more beneficial to focus on cyber attacks from bad actors using powerful AI rather than autonomous super AGI , pretty sure 99% of people would laugh that off unfortunately. Although he isn't wrong.
I REALLY don’t want AI to figure out how our brain puts our experiences and our communication in the same context !
What are you talking about?!
AI seems to stop people sitting down...
Hinton has a bad back. He cannot sit.
@@arshdeep691 Maybe AI should figure out a way to safe its godfather, instead of producing kitty videos nobody needs.
The matrix is becoming a reality🤣
Conrats on the Nobel prize
With all the progress in A.I., can someone explain to me why in the last few years the vidz suggested from yew tew are so bad now?
They're not optimizing for giving you the best search results. It's like how milk is at the back of the grocery store
Isn’t AI already smarter than humans? Discounting the supreme AGI but systems are generally smarter than humans at most tasks
Searching a database has nothing to do with being smart.
"faster"
its not intelligence tho, its predicting what word comes next lol, were really far from human intelligence, even the chief AI at meta said it on government summit.
@@dan-cj1rr they don't predict the next word. They predict the next token. The idea is that they can emulate the processes behind the token generation ie model the outside world from the sampling they train on. It is indeed a far cry from human intelligence, but one of the things we do is predict into the future to then respond to. Scale will likely help with more complex problems (and step by step training). And some persistence of state would be an obvious next step. Le Cunn also thought it was impossible to make videos via a diffusion model, yet scale help quite a bit --Sora. Though I doubt the final model will be that, and we may likely shift to others. Hence why you can ask it about hypotheticals its never encounters and extrapolate from its training ie given certain colors of paint and how they fade, how should i paint this room if I want it some color.
Le Cunn also believes silly things like they don't really understand. Scottsmans fallacy, Hinton as said in other talks that its the representations of the neural nets that is the understanding and perception. They're not self aware in any meaningful sense though. Each inference only last millliseconds and they've had no way to evolve more advanced forms of abstractions like ego, pain, pleasure. It feel very premature to claim that we are really far from human intelligence (though I doubt even if they become far smarter than us, it may not be quite "human-like" just from that ) esp when we have not hit a ceiling to hitting the top of the Scaling Laws. :x
No, it's us who are going to be dumber by using AI
AI hallucinating now but when that's fixed...watch out!
That's not fixable, because of biased and incorrect data. Ai is not intelligent.
That is a low bar.
10:26 wow that's a huge claim. "The furure of humanity"
I have a few questions here. Why would a piece of code want control? It has zero emotions? Code that weighs things out over and over again until it gets to a result it thinks we desire, I see no way that gains emotions or even consciousness.. More intelligent absolutely, in the academic or knowledge holding/manipulation. But to be sentient is much different and I do not think is even close to a concern at the moment or am I wrong saying so? Or did I just write all that when what he meant is that AI will take over in the sense where it can do everything better so why would capitalist hire people. If so totally agree and I misinterpreted everything being said here.
AI is programmed to learn. It has goals and to achieve it's goals, its first objective is take control. Because achieving a goal is much easier when you have control. It doesn't need to be sentient to take control.
AI is programmed to “learn” - meaning what exactly? The fact that this video appeared in my recommended feed is a result of an algorithm “learning” that i might want to watch it based on patterns it’s recognized, associated metadata etc. At which point did “control” enter the equation? The same could be said for gaming bots like Alpha Go, Agent 57 etc. They crunch massive amounts of data quickly, and through trial and error establish the appearance of intelligence. At no point do they assume control nor do they have alterior motives beyond what they’ve been programmed to do.
Stop spouting nonsense. The amount of hyperbolic AI “disinformation” floating about is truly astounding, some of it shamefully coming from Hinton himself here which is baffling.
@@meatskunk it sounds like we agree.
20 years 😂😂 people are in for a rude awaking over the next couple years
There are college students who cant do 15 x 4
60
it's pretty wild to think that predicting the next token and "attention" were the magic processes needed to build a general understanding from the sampling of the world they try to emulate. Hopefully larger scale improves its overall reasoning abilities, maybe there's a few tricks left until its human level intelligence (in term of solving novel problems). But i can imagine there is that many prob less that a handful of key breakthroughs left of the same scope in terms of foundational breakthroughs.
How is AI used in China? CGTN The Point-Hub-Heat, Einar Tangan-Martin Jacques-Lijinjing-Tian Wei. Reporterfy Media-Cyrus Janssen
Funny thing is that my dog is smarter than most of the MAGA voters in America.
An AI would already trump Donald as the dictator, I mean president, of the US.
good! maybe it would be a better world. to Skynet!!!
All super ccomputers < human brain. This is proven.
AI is so dumb, even Internet searches are getting worse and worse.
Oh yee of little faith.
There is no AI per se, yet. LLM has no ability to self improve and realize its existence.
If you get bad answers it’s because you suck at asking questions to a chatbot.
By people he seems to mean snart people like himself. 😂
Someone who is able needs to start raising children to be John Connor kind of person
Just hire a newbie programmer that programs software full of bugs and the AI will come to a halt without even having to plug off the computer running the AI. :)
TDS guy!
Free Quebec
...and that's relevant to this story how?