Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google DeepMind Founder Demis Hassabis: Three Truths about AI

techrepublic.com

Post reply on HN
Print view View original

111–120 of 144 posts

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#111

Post by ableal »

mindcrime wrote: post #93

I've been spending a lot of time lately digging into "outdated" 80's era approaches to AI. Some people think I'm wasting my time and ask "why aren't you doing Deep Learning?" This is the reason why: "You can think about deep learning as it currently is today as the equivalent in the brain to our sensory cortices: our visual cortex or auditory cortex. But, of course, true intelligence is a lot more than just that, you…

> digging into "outdated" 80's era

I'd say that's a good idea. A 30-year cycle (say one human generation) seems to occur in many fields, for various reasons.

For instance, some of the first transistors were field-effect (FETs), then bipolar junction (BJTs) ruled the first integrated circuits, then MOS technology made FETs again the standard technology. Some of the design approaches were recycled ...

Sometimes things seem to go in a spiral, and the new level amplifies the previous. I'd just caution that knowing why the previous approach went out of fashion would be useful, but that's often hard to determine - acknowledging failure is not a popular activity. Lack of power may not be the only problem ...

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#112

Post by slm_HN »

tomd wrote: post #56
a008t wrote: post #40

Earlier quoted context omitted.

He is primarily a salesman, is he not?

> 'probably smart', 'primarily a salesman' "A child prodigy in chess, Hassabis reached master standard at the age of 13" "graduating in 1997 with a Double First[13] from the University of Cambridge" "obtain[ed] his PhD in cognitive neuroscience from University College London" "[his] theoretical account of the episodic memory system [...] was listed in the top 10 scientific breakthroughs of the year in any field by th…

Sure, PhD in cognitive neuroscience is OK, but Demis' most impressive feat is winning the Mind Sport Olympiad a record 5 times.

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#113

Post by the_greyd »

drabiger wrote: post #35

It's sad that a well paid person and probably smart person like Demis Hassabis is so wrong in his view about how the world is. To quote the article quoting Hassabis, "The reason I say that is that if you look at the challenges that confront society: climate change, sustainability, mass inequality — which is getting worse — diseases, and healthcare, we're not making progress anywhere near fast enough in any of these a…

That progress was set in motion by decisions we took years ago. I think all of us understand at some level, that we are failing to keep making good decisions. Books like "Better angels of our nature" and "factfulness" try to convince you that world is only going to get better, but I don't think it's obvious at all.

Re: enlightenment values, liberal democracy, trade etc. I believe governments and public institutions played a major role in the past that contributed to our success now.

We have moved from public institutions to private companies now. More capital is being invested now in selling us ads, selling you the next version of iphone etc. The companies with the most capital are tech companies. They are not solving the problem of healthcare, climate change, sustainability, mass inequality or any meaningful thing in any way. People are more excited for VR games. I feel right now, humanity is like a rich spoiled brat.

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#114

Post by EGreg »

mindcrime wrote: post #93

I've been spending a lot of time lately digging into "outdated" 80's era approaches to AI. Some people think I'm wasting my time and ask "why aren't you doing Deep Learning?" This is the reason why: "You can think about deep learning as it currently is today as the equivalent in the brain to our sensory cortices: our visual cortex or auditory cortex. But, of course, true intelligence is a lot more than just that, you…

I have one question for you and Demis

What if logic itself is just reduction to a simple binary system with few moving parts - the easy case?

Being able to logically explain how something works in special cases is what we have been able to do until now with pencil and paper.

Then you get approaches in the more general case, like New Kind of Science by Stephen Wolfram to “study” self-similarity. And you have fluid dynamics, which oncd again tries to find analytic closed-form solutions after averaging.

But what if - with computers - you can GENERALIZE REASONING ABOUT MODELS to something far beyond logical explanation? And it’s not reducible to a few main factors, but instead relies on precomputing HUGE VECTORS via gradient descent or MCTS (like AlphaZero) and store that information compactly. So you can’t really explain “why” this move is (one of) the best moves, it just is.

What if you could generalize AND handle a greater set of problems and “logic” was just a subset of it after all?

Ideas like “all swans are white” and syllogisms are just approximations of the real world. With deep learning you can find patterns that contain far more variables and complex interrlationships.

So maybe the 80s AI will suddenly emerge as a super-logical reasoning computer, not from efforts like Cyc which attempt to feed it data, but from general purpose deep learning with zero experience!

What if it is suddenly better than everything humanity has ever produced or reasoned about, similar to how AlphaZero beat every bespoke chess computer program ever written after a few hours of training?

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#115

Post by fsloth »

pas wrote: post #26
fsloth wrote: post #19

Earlier quoted context omitted.

Sorry, I think we are using the term 'technology' in a different ontological context. I like the brief description "Technology: the methods for using scientific discoveries for practical purposes". I.e. how to apply accrued knowledge practically. The opposites to technology-based problem solving would be trial-and-error, political or ideology based mandates, etc. This does not exclude adaption, nor does it mandate a…

Is trial-and-error outside science? Or do you mean a uniform prior distribution for trial-and-error?

Uh, in this case trial and error as an option to a situation where a solution is already generally known.

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#116

Post by oh_sigh »

mrhappyunhappy wrote: post #94
hyperbovine wrote: post #71

Earlier quoted context omitted.

One potential solution is to not have children.

What’s the point of living if you can’t achieve your biological function?

If procreation was the sole important biological function, gay people would be much rarer.

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#117

Post by mindcrime »

EGreg wrote: post #114
mindcrime wrote: post #93

I've been spending a lot of time lately digging into "outdated" 80's era approaches to AI. Some people think I'm wasting my time and ask "why aren't you doing Deep Learning?" This is the reason why: "You can think about deep learning as it currently is today as the equivalent in the brain to our sensory cortices: our visual cortex or auditory cortex. But, of course, true intelligence is a lot more than just that, you…

I have one question for you and Demis What if logic itself is just reduction to a simple binary system with few moving parts - the easy case? Being able to logically explain how something works in special cases is what we have been able to do until now with pencil and paper. Then you get approaches in the more general case, like New Kind of Science by Stephen Wolfram to “study” self-similarity. And you have fluid dyn…

So maybe the 80s AI will suddenly emerge as a super-logical reasoning computer, not from efforts like Cyc which attempt to feed it data, but from general purpose deep learning with zero experience!

That's a fair point, and I've spent some time thinking about roughly the same thing. My current thought on it is this: "general intelligence" may well turn out to be an emergent property of very simple lower-level primitives, and it may be that what we call "deep learning", done with a sufficiently deep and wide network, might actually produce "general intelligence".

But... I feel like we're still pretty far from having the compute power to run an ANN of such scale (that's just a guess though). So where we can "help" the process by hand-engineering modules that do specific elements of what we call intelligence, combine those modules, that's one way to short-cut the process. And in my view, things like Deep Learning, DRL, MCTS, etc. would also be included, along with older approaches: PCT for abductive inference, logical inference, maybe something like symbolic regression with GA's, etc., etc.

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#118

Post by munificent »

mindcrime wrote: post #75
adwhit wrote: post #49

I always find it strange to hear people talk about climate change as if it is some sort of tricky challenge we need to solve, like improving education or curing cancer. It is a 'problem' of completely different magnitude. It is on track to end civilization as we know it within two generations. A rational society would think about nothing else. Our children will be astonished and mystified - and furious - to see all t…

Our children will be astonished and mystified - and furious - to see all the things we got up to instead of dealing with this thing that is the only thing that matters. That's been happening since Generation X kids were old enough to understand what a mess our parents made of the world. Unfortunately we haven't done a very good job of making a difference. Frankly I'm a little disappointed in my cohort in that regard.…

I am really optimistic about the younger generation, actually.

I'm a gen-Xer. Our mantra was "whatever". Instead of trying to solve problems, we just tried to detach from them. As long as we were self-aware and cynical about what we were being sold, we thought things would be fine.

But that's obviously stupid in retrospect. People in power love cynicism — why bother defeating other people when they'll defeat themselves?

Millenials, on the other hand, seem way more organized and engaged. For their generation "woke" doesn't just mean being aware of what's going, it means doing something about it. So I hope that they (and us too) can start changing some of the fundamental structural flaws in the current systems of power.

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#119

Post by jobigoud »

lixtra wrote: post #62
adwhit wrote: post #55

Earlier quoted context omitted.

Well that's OK then (???)

Yes, learn to accept your mortality. Long term the human race will die. Some think by robots taking over. Some think by earth getting as warm as it has not been a long time much faster than ever (although live was thriving at that time). Some think a bit meteor might hit us or a super volcano could erupt and cause 100 years of darkness. Some think a global nuclear war will end us. Some think humans will just procreat…

> Yes, learn to accept your mortality. Long term the human race will die.

Don't give up?

We can accept our mortality as individuals but why accept that of the species? There is no precedent for what we are. We are the first species that we know of that could be actively trying to survive for as long as possible, at the species level, by concerted effort of intelligent individuals and long term planning.

The average lifespan of mammalian species seems to be 1 to 2 million years. Very short. We should try to survive for much longer than that, and this needs planning and not fucking everything up.

It's also kind of important that we or a descendant species (including a robotic one) survive because we might be alone in the Universe.

Re: Google DeepMind Founder Demis Hassabis: Three Truths about AI

#120

Post by gowld »

DennisP wrote: post #64
exceptione wrote: post #57

Earlier quoted context omitted.

> It is on track to end civilization as we know it within two generations. A rational society would think about nothing else. If you mean by "to end" something like "to extinguish" than I think this hyperbole is enormous. If you mean by "to end" that the civilization will change, than there is nothing new under the sun. If I look at pictures from 19th century, which is very recent, I am always baffled at how much hav…

A while back I read Six Degrees by Mark Lynas, who read 3000 peer-reviewed papers on the effects of climate change and summarized them, one chapter per degree, with extensive references. My takeaway from that book was that three degrees was pretty terrible but life would go on, but at four degrees modern civilization would have a hard time surviving. Go all the way to six and there won't be many humans left on the pl…

How is that possible? From the Amazon blurb, it seems that the book talks about what would be destroyed by weather change, but nothing about what would be created. If temperature increases, some current cold areas would be more habitable. If sea levels rise, some inland area would become coastal. These changes would happen over generations, so populations would move. There would be tumult in human and animal life, which is bad but not a novel part of human experience (we've always had hurricanes, tsumanis, earthquakes, and plenty of manmade disasters), but a new equilibrium would be reached. There was once an Ice Age that supported human life! I don't think the overall human environment 6 degrees in the future would be worse than say 200 years ago.

Also, it's been 16 years since the book was published. Have temperatures risen? Have the predictions come to pass?

https://www.amazon.com/Six-Degrees-Future-Hotter-Planet/dp/1...

Post reply on HN