What does /sci/ think about Mr. Demis, the guy who claimed he'll

>>7900175
If he was really smart, he would of studied math and wouldn't be wasting his time with games. You didn't even post his IQ. Maybe because it's too low?

>>7900183
kek

>>7900175
if he's so smart, why isn't he the richest man on earth?

>>7900187
i know you're trolling but

>In January 2014 DeepMind was acquired by Google for a reported £400 million (approximately $625 million), where Hassabis is now Vice President of Engineering leading their general AI projects.

>>7900193
I'm not trolling.

>>7900303
Me too. But due to my high intelligence I'm too apathetic.

I know this is a meme thread but Demis will probably go down as our Einstein. In fact he's probably even more important than that. He's leading the team that will create human-level AGI.

>>7900342
Wouldn't Hawking go down as our Einstein considering he's a physicist like Einstein was?

>>7900362

I just mean "le important science man".

>>7900342
this fucking retard will doom us all if it will not be friendly
hope he is familiear with jewdkowsky

>CS

stopped reading right there ;)

>>7900362
Except that Hawking did nothing of importance. He is literally only famous for sitting in a wheel chair.

>>7900175
>Cognitive Neuroscience
Isn't that the field where publishing untestable pseudo-philosophy garbage on consciousness and free will is allowed and even encouraged?

>>7900385

Friendly AI will be solved, I'm sure of it. Too much is on the line and too many smart people are working on it for it to fail.

What I'm afraid of is meme countries getting their hands on the tech and altering it to cause harm. This isn't a 50+ year away danger, this is right now. Demis and Shane Legg admit (in private company) that they see human-level AI being viable by 2025.

Geoff Hinton agrees. Rich Sutton agrees. Yoshua Bengio. Blaise Aguera y Arcas. Dileep George. etc etc. This is happening and the normies don't see it coming at all.

Don't know much about him, but he seems like a very insightful and interesting person.
Sadly, the modern academia lacks ambition, because its members are afraid of deviating from the norm and being criticized for that (usually by a small minority, which feels it can boast its self-esteem by bringing down more ambitious people than them).

>>7900175

Are you his publicist?

File: 1428416749361.jpg (1MB, 1830x1598px) Image search: [Google]

1MB, 1830x1598px

>>7900175
>rationalists being proud of their eccentricities

He looks like a nice guy. I like him.

>>7901119
Do you like me? My IQ is only 85.

>>7900362
If Stephan Hawkins is so smart, how come he doesn't just stand up?

>>7901139

his brain is too heavy, he merged with his computer chair and now his mind is the internet

i want to be him

>>7900341
OMG me too. I actually could do anything if I tried.

File: tumblr_nqbcgfL62K1tdpeqjo1_400.png (27KB, 372x344px) Image search: [Google]

27KB, 372x344px

>>7900175
Deepmind does really interesting research. In my (somewhat educated) opinion, Neural Turing Machines are future of AI. NTM learned a sorting algorithm just from seeing examples of sorted and unsorted vectors!

>>7900342
Geoff Hinton will probably be more influential desu.

>>7900385
>Yudkowsky
lol

>>7900413
>I'm sure of it.
Unless you are a competent AI researcher you shouldn't have a strong opinion on that, you just don't know enough. Even AI experts have proven to be terrible at predicting the future of AI.

>Geoff Hinton agrees. Rich Sutton agrees. Yoshua Bengio. Blaise Aguera y Arcas. Dileep George

Jurgen Schmidhuber disagrees. Peter Norvig Disagrees. Stuart Russell disagrees. Richard Sutton. Murray Shanahan. Steve Omohundro. Alan Turing.

>>7901288
>somewhat educated)
lmao what's your education faggot?

>mfw CS is the most important science right now because of AI
eat shit /sci/

>>7901352
I've almost graduated primary school

>>7901288
>Geoff Hinton will probably be more influential desu.
Why? He's almost dead, and Demis already achieved more

File: backprop.gif (4KB, 410x141px) Image search: [Google]

4KB, 410x141px

>>7901358
>demis already achieved more
>Demis' most cited paper got 687 citations since 2007, so 80 citations per year
Hinton's most cited paper got 18,000 citations since 1985, so 600 citations per year.
>literally 7 times as many citations per year

Look it up if you don't believe me.

Not knocking Demis, he does god tier research and I'm sure it's possible he will achieve more than Hinton but I doubt it. Geoff invented backpropagation, momentum,restricted boltzman machines and dropout. Backprop alone will make him one of the most celebrated ML researchers of all time.

>>7901381
>measuring success in citations
did you get brainwashed by academia? you forgot to throw in an h-index and some royal society membership bullshit

demis is no longer in academia, he's only published papers during his phd and a bit with gatsby afterwards. geoff didn't invent backprop, check your facts[1]. boltzman machines is not nearly enough to keep him in history, and he did that with the guy who teaches "learning how to learn" on coursera. geoff is good, but demis has just started AND it has the media/hollywood appeal of being a chess genius and etc. expect a movie about the AI boy genius demis

[1] https://en.wikipedia.org/wiki/Backpropagation#History

>>7900398
What do you know about Hawking Radiation?

>>7901402
>Criticizing me for measuring success by citations
>Measures success by media hype and chess skills

I said Hinton would be more infulential. Citations aren't perfect but they're a great way of measuring academic influence.

>demis is no longer in academia

Private AI research and academia are still so interconnected they might as well be the same thing
>hinton didn't invent backprop

Sure, but he made it useful and popularized it, thus he was the most influential.

>demis has just started

Demis is 40. When Geoff Hinton published the backprop paper, he was 39. I doubt anything Demis has done will be as influential as the backprop paper(the second most cited paper in ml).


Perhaps we are comparing apples and oranges. I wouldn't be surprised at all if Demis is more celebrated by the mainstream media in 50 years time, but it seems unlikely he will have more influence on actual research.

>>7901418
well, if you pay attention, the fact that stirred the argument is that demis will be our einstein aka "le important science man" [1] so media hype and chess skills are likely more of a factor than number of citations

i meant demis just started researching AI. his neuroscience papers are not neural computing/neuroscience hybrids, but rather straight up neuroscience. he claims to have used AI for his games but i don't know how true it is and to what extent.

well, if alphago beats the Go champion, that'll be quite a milestone in AI. obviously not as important as creating backprop, but geoff didn't invent it so hey

[1] >>7900379

>>7901418
>Private AI research and academia are still so interconnected they might as well be the same thing
except it's not an industry based on releasing papers, like academia is

>>7901418
>(the second most cited paper in ml).
Where do you check that? Does Google scholar have a feature supporting that?

Oh so whe he does computer science hes a prodigy but when i do it im a brainlet on sci fuck this double standard you niggers.

>>7900342
That's funny because I often mentioned him on /sci/ and nobody did even know who he is.

I don't think he will be considered in the same way as we now consider Einstein; but I think that he is currently one of the 3-4 most important people on Earth regarding the future of humanity.

>>7901504
>falling for the CS meme
you are a brainlet bro

>>7901637
>often mentioned him on /sci/
>https://warosu.org/sci/?task=search&ghost=yes&search_text=demis
>he's been mentioned like 3 times before this thread
ok bro

>>7901657
Warosu is back up at last?

>>7901668
and https://4ch.be/ bailed out

>>7901657
Exactly, those posts are mine.

>>7901684
one of them is mine faggot

File: lefty.jpg (1MB, 1663x1176px) Image search: [Google]

1MB, 1663x1176px

>>7900303
>mfw IQ 193
>mfw I could go to Mars and Pluto simultaneously if I wanted to
>mfw I prefer to sit at home and watch Caribbean TV
Life is good

>>7901691
I'm super smart too its just I'm dumb.

>people falling for the demis meme
I thought he was inspiring at the beginning too, but then you realise he just knows how to market himself, and lucked out to be intelligent.

The thing is, he's all about games. He did CS cause he wanted to do games, he was a game designer, and now all the AI he's done was mostly focused on games. Furthermore, after DeepMind was bought by Google, they declared that they're now also working on recommender systems (i.e. improving Google ads), and MORE games.

Demis is intelligent and capable, but even though he tries to market himself as someone who meticulously planned out his life/career, it's just a façade. He claims that "his maths got rusty" and tries to downplay its importance, but (and, ignoring /sci/ math meme), it is obvious that he never really studied maths properly. CS is not a mathematical rigours degree, and neuroscience even less.

His neuroscience PhD sounds good on paper, and he probably thought that it's gonna bring him an understanding of brain that he can use for developing AI, but if you take a look at his papers released, although they are somewhat important in neuroscience, they have little to do with his current research.

Basically, Demis is an intelligent gamer that dedicated more than half of his life to games: chess when he was young, "mind games olympiad" or whatever, game designer job, computer science degree, etc. Then, he decided to improve his angle with some AI, but he already wasted too much time on AI and lacks the basics (e.g. mathematical knowledge). Now that his company was bought by Google, he spends most of his time in Skype conferences with the US offices, and rest of the time with his family.

Sure, he might defeat the Go champion and whatnot, but Demis consumed himself on other stuff already, and he knows it, but creates a whole narrative to make himself look good. He's a very intelligent guy, but he won't solve general AI. Someone else will though

>>7901698
>knowing this much and still not understanding what is going on

If you take a break and look at the other people working at Deepmind, you'll see how every single one of them is a top-tier ML researcher. These people don't waste time on games and conferences: they just publish astounding papers. A lot of them.

Demis doesn't directly do research (and yeah, maybe he wouldn't even be capable to do it as good as his colleagues), but he understands it at the very least on an intuitive level, and gives good insights to the researchers on where to go next. He's also costantly thinking of possible applications of their system and of new ways to improve it. If he happens to have a a good idea, the people working for him are so damn good that they'll translate it into code at fucking light speed.

If your post was "he won't be the loner genius who solves AI" I absolutely agree with you. But it's very likely that he'll lead the team that does it. That's what he also means when he says that he intends to solve it.

>>7901714
I don't argue that DeepMind won't solve intelligence, they have enough resources to get the most motivated scientists to work on it. It's just that Demis has become a pop star now, and is too busy to be a CEO to do anything else. If this was a DeepMind thread instead of a hero worship thread, I wouldn't have posted that.

Btw, is there a list mentioning everyone employed by DeepMind? I know they have a list of publications on their website, but I'd be interested in a list of their superstars.

>>7901714
>These people don't waste time on games and conferences: they just publish astounding papers. A lot of them.

>Mastering the game of Go with Deep Neural Networks & Tree Search [view publication]
>Authors: D Silver A Huang C J Maddison A Guez L Sifre G van den Driessche J Schrittwieser I Antonoglou V Panneershelvam M Lanctot S Dieleman D Grewe J Nham N Kalchbrenner I Sutskever T Lillicrap M Leach K Kavukcuoglu T Graepel D Hassabis

By the looks of it, most of the people there are still focussing on gaems

>>7901766
Furthermore, looking at DeepMind, they have 3 main projects: AlphaGo, DQN (Atari games), and the recently added Health division. As far as I can tell, their main focus is still games and it might continue with this trajectory while Demis is the CEO

What the hell? Games provide a useful microcosm for testing intelligent agents. Just because the algorithms developed are being used in games doesn't mean they won't eventually be applied to other things. Do you seriously think you know better than these guys? Do you seriously think Google would have handed them $400m if their only interest was in developing game AI?

And can you really not see how groundbreaking their research is, despite being primarily game-focused? The Atari paper made Nature ffs.

I assume you have a better idea, Mr. Armchair?

>>7901766
>>7901774

>I don't know what I'm talking about

The alternative would be wasting time to program specific virtual environments where they can test their system on.

But wait, there are thousand of those environments that have already been programmed in the past decades. They are called games.

>>7901788
>Do you seriously think Google would have handed them $400m if their only interest was in developing game AI?
Google bought them so they can work on recommender systems in parallel (which is the main source of revenue for Alphabet Inc)

>And can you really not see how groundbreaking their research is, despite being primarily game-focused? The Atari paper made Nature ffs.
There's groundbreaking research everywhere. Have you checked some recent CMU papers? Berkeley and Stanford are also working on AI, you know? DeepMind is good, no doubt, but imo they're overhyped.

>>7901797
But wait, there are even more environments that already exist: it's called real life

File: 1454885681185.jpg (37KB, 420x418px) Image search: [Google]

37KB, 420x418px

>>7900175
seems like a cool guy

t. expert on what is and isn't cool

>>7901819
That doesn't make any sense. Are you trying to troll or are you just retarded?

>>7901841
k tard

>>7901819
>>7901841

They are reportedly working on robotics too.

>>7901698

>the demis meme

what the fuck man this needs to end

>>7902789
stfu kid

>>7900183
>would of

CS is a meme

sage

File: 20050222_LionGreekArmy_mar.jpg (215KB, 1024x818px) Image search: [Google]

215KB, 1024x818px

>>7900175

might run into crisis when his networks will successfully beat complex videogames.

and then nothing happens, that's just what they'll do, beat videogames.

>>7901698

>hurr his math is weak
>durr games are for kids
>urff intelligence must be about math and chess ONLY

>>7900175
I think he's an ugly virgin with a cuckold fetish.

>>7904217
are you trying to say something over there, anon?

File: demis2.jpg (169KB, 794x633px) Image search: [Google]

169KB, 794x633px

>>7902789
What the fuck did you just fucking say about me, you little psych major? I'll have you know I graduated top of my class in Cambridge, have been involved in numerous secret AI research projects and I have over 300 confirmed kaggle wins. I am trained in advanced machine learning and I'm the director of research at Google Deepmind. You are nothing to me but just another undergrad. I will beat your validation error% with precision the likes of which has never been seen before in academia, mark my fucking words. You think you can get away with calling me a meme on the internet? Think again, fucker. As we speak I am running my neural turing machine trained to deanonymise 4chan posters and your IP is being traced right now, so you better prepare for the storm, maggot. The storm that wipes away the pathetic thing you call your STEM career. You're fucking done, kid. My AI can be anywhere, anytime and can beat you at atari games in over seven hundred ways, and that's without even using an ensemble. Not only am I extensively trained in reinforcement learning, but I have access to the entire GPU cluster of google research and I will use it to beat your miserable ass 5-0 at go, you little shit. If only you could known what complete automation of your STEM job your little 'clever' post was about to bring down on you, maybe you wouldn't have called me a meme. But you couldn't, you didn't and now you're paying the price, you goddamn undergad. I will minimise a loss function with you in it. You're fucking done, kiddo.

>>7906052
>can beat you at atari games
kek

>>7906052
>psych major
>proceeds talking about STEM job/career
i don't think so demis

>>7900175
He look like a faggot

>>7906052
Could be refined but really awesome overall

>no one mentions Google X

https://en.wikipedia.org/wiki/X_(company)

https://www.fastcompany.com/3028156/united-states-of-innovation/the-google-x-factor

>Mostly, X seeks out people who want to build stuff, and who won't get easily daunted. Inside the lab, now more than 250 employees strong, I met an idiosyncratic troupe of former park rangers, sculptors, philosophers, and machinists; one X scientist has won two Academy Awards for special effects. Teller himself has written a novel, worked in finance, and earned a PhD in artificial intelligence. One recent hire spent five years of his evenings and weekends building a helicopter in his garage. It actually works, and he flew it regularly, which seems insane to me. But his technology skills alone did not get him the job. The helicopter did. "The classic definition of an expert is someone who knows more and more about less and less until they know everything about nothing," says DeVaul. "And people like that can be extremely useful in a very focused way. But these are really not X people. What we want, in a sense, are people who know less and less about more and more."

>>7900175
>CS major
>smart

>>7907295
>i memed well today XD

>>7907281
hey i know less and less about more and more! in fact i'm so advanced i know nothing about anything

>>7901426
>if alphago beats the Go champion, that'll be quite a milestone in AI.

It would be a bigger milestone than Deep Blue. A far better measure of actual artificial intelligence than IBM's marketing coup.

The match is starting in just a few days! It will be five full-day games between March 9 and 15.

>>7900407
No.

>>7907314
>actual intelligence

>>7907325
Yeah, deep learning networks are far closer to the only example of intelligence (brains) than Deep Blue's gang of chess chips. And the IBM effort turned out to be an AI dead end. It was dismantled immediately after the match and its techniques were not carried forward to other areas, or even other chess programs.

>>7907341
that doesn't make it actual intelligence

Just tested my IQ on http://www.free-iqtest.net and I got 151

Am I smarter than demis?

>>7907371
you're a lot more stupid because you fell for a meme internet IQ test

>>7907384
Are you perhaps mad you scored less than me?

>>7907384
Why would anyone fall for such low quality bait?

>>7907395
baited hard

>>7907341
Minimax with heuristics isn't a dead end. It just isn't suited for every kind of problem.

File: 1401050524487.jpg (41KB, 356x437px) Image search: [Google]

41KB, 356x437px

>>7906052
>If only you could known what complete automation of your STEM job your little 'clever' post was about to bring down on you, maybe you wouldn't have called me a meme.