Using AI to accelerate scientific discovery - Demis Hassabis (Crick Insight Lecture Series)
Вставка
- Опубліковано 22 тра 2024
- Using AI to accelerate scientific discovery
Demis Hassabis, co-founder and CEO, DeepMind
At The Francis Crick Institute in King's Cross, London
Abstract:
The past decade has seen incredible advances in the field of Artificial Intelligence (AI). DeepMind has been in the vanguard of many of these big breakthroughs, pioneering the development of self-learning systems like AlphaGo, the first program to beat the world champion at the complex game of Go. Games have proven to be a great training ground for developing and testing AI algorithms, but the aim at DeepMind has always been to build general learning systems ultimately capable of solving important problems in the real world. Excitingly, I believe we are on the cusp of a new era in science with AI poised to be a powerful tool for accelerating scientific discovery itself. We recently demonstrated this potential with our AlphaFold system, a solution to the 50-year grand challenge of protein structure prediction, culminating in the release of the most accurate and complete picture of the human proteome.
The speaker:
Demis Hassabis is the Founder and CEO of DeepMind, the world's leading AI research company that aims to solve intelligence to advance science and benefit humanity
In 2016, DeepMind developed AlphaGo, the first program to beat a world champion at the complex game of Go. In 2020, its Alphafold program was heralded as a solution to the 50-year grand challenge of protein structure prediction and in 2021, DeepMind launched the AlphaFold Protein Structure Database, which offers the most complete and accurate picture of the human proteome to date.
A chess prodigy, Demis reached master standard aged 13, and went on to program the multi-million selling simulation game Theme Park aged 17. After graduating from Cambridge University in computer science, he founded pioneering videogames company Elixir Studios, and completed a PhD in cognitive neuroscience at University College London. Science listed his neuroscience research on imagination as one of 2007’s top ten breakthroughs, and in 2021, AlphaFold2 was selected as the Breakthrough of the Year.
He is a Fellow of the Royal Society and the Royal Academy of Engineering. In 2017 he featured in the Time 100 list of most influential people, and in 2018 he was awarded a CBE for services to science and technology. - Наука та технологія
![DeepMind x UCL RL Lecture Series - Introduction to Reinforcement Learning [1/13]](/img/n.gif)
Demis starts speaking at 5:50
Deepmind for AGI 7:15
Reinforcement learning and Go 12:10
AI for scientific discovery 24:45
Protein folding 28:22
AlphaFold2 37:21
Future of Deepmind 51:09
Q AND A ~skipped some questions
Bias in AI 54:33
Intuition and consciousness 57:07
AlphaFold3 1:03:02
Potential negative effects of AGI 1:15:15
Explainability of AI 1:22:40
Thank you
Legend
Gods work
Thank you
You did a man's job. Thank you.
Exceedingly rare to get a glimpse of the future...but here it is...Wonderful work by DeepMind...
Please post more of these presentations and consider doing an ongoing podcast interviewing members of your team!
Have you listened to the Deepmind podcast presented by Hannah Fry?
ua-cam.com/video/r_Q12UIfMlE/v-deo.html
@@ddos87 Yes. I wish Denis had more time to talk about these topics.
@@ddos87 Thank You. Her podcast is more is a general overview of AI topics. I’m looking for more in-depth technical coverage.
Thanks for the update. Always very interesting to hear from you.
That the brain is a black box and better understood than the algorithms it created is a mystery. Though funny when Demis asked "how are you coming up with these questions?"
Protect this gentlemen at all cost. What a treasure.
The Real Genius : DEMIS HASSABIS
He is undoubtably VERY smart. But this is the results of a big team of 1000 people. He just happens to be the articulate figurehead.
@@byrnemeister2008 I think there's a hard won skill in distilling the extraordinary and novel work your people do. Fashioning into a narrative that makes sense and is true to the nature of the endeavour...and delivering something so exiting and palatable to such a broad range of folk.
Maybe that is a bit genius-y ?
Love how he keeps on track of such a diverse topic with knowledge garnered from working with a machine that's working on current problems, involving something he neve studied. That will be a pretty nice application of all this, a teacher that can understand you and what you are trying to learn then formulate a way to get that knowledge in your brain in such a way you will understand. Pretty sure educational academy's can be fucked of then.
Wow the questions were great, and Demis appears to be all over his crusade. Formidable.
Really incredible and fascinating talk, with great questions in the Q&A section as well!
great roadstep taken and the trial there!
If you think deep enough you will realise a calculator is very dumb, we made it smart by its functional value, but its not AI and nor are computers, do you understand or do you want to just fantasy your reality?
Great talk 👍
Great ❤️
Fabulous talk.
Thank you so much
I believe Demis should win a Nobel prize in biology for AlphaFold. I would love to meet him one day, would be a great honor and a fascinating person to listen to and have a conversation with
hello. Thanks for the lectures. How can one access the slides.
In regards of the question of intuition, actually the model itself is intuition, because you don’t know how the model came up with the correct results, you need to investigate the how in the aftermath. Congrats for your great work!
Massive respect for Lee sedol who beat alphago ....even if it was a single time
After seeing this, I just wanna work with you guys!!. This is freaking awesome!!.
its just computer programs doing what WE TELL IT TO DO,. GROW UP! NO AI
Is still awesome no matter what you say
@@autisticsimon12 aren't we all humans the same in low level?
@@autisticsimon12 No. You are not right. This programs can learn by themselves. And after learning they can do what people did not wright in the code
@@Sol-En I see what your saying but i dont agree it is AI. Take fractal patterns, they seem to be created in a way that's similar to what you are saying, does that mean fractals are AI?
Is there going to be Alphafold zero?
Whenever I think about protein folding, I'm amazed that life at some level is based in geometry.
Damn.
Demis Hassabis be like “… but first let me tell you the entire history of chess and go.”
So AlphaZero can play any of the two player games on demand? i.e. if it learns Chess it will not then forget how to play Go? Amazing if true - how has catastrophic forgetting been solved?
It's not necessary for one trained net to be used. Also why do that when multiple nets can be bundled together?
Who is the guy introducing him?
Great talk. But during the Q&A talk can the host not breath heavy or talk while the mic is on?
Instead of white please make the background black. It’s better for the eyes and oled screens. Put the logos on a different color background if they are really necessary.
Demis was working on the game syndicate??? As if there wasn't enough reason to like/be impressed by the guy...
There is common thing in tier and chappals.friction
There is common thing in flying in birds and aero plane "shape and pulling& pushing of air to back."
The common thing in intelligence is efficiency
Use AI to give the best job in the "game" of real life to all the people in need, solve problems of unemployment, low paid jobs, professionals in the wrong jobs, professionals unemployed because they have no previous experience, adults ignored because they are over 35, families that do not have enough income or have reached the limit, etc.
Computer already can see, hear and walk (boston dynamics for example) like animal. So we already have some modules of the brain. It seems to be, that we just need to create more different modules of the brain and connect them together in a right way and that's it. General AI will be done
How these modules are combined will be the final task a software engineer does.
The question is whether the composition of the module imitates the "whole brain" or does not imitate the brain.
In Japan, where I live, there are companies that are trying to build an intelligent system with an emphasis on the neocortex.
@@alph4966 Do you work in AI field ?
Nah, it doesn't really work like that. You don't have "modules of the brain"; you either have a brain or you don't. The fact that boston dynamics robots give you the zoomorphic illusion means nothing in the context of general AI. They don't work any differently than animals in video games, yet would you call those "general AI"? It's not about faking it, but actually making it.
May be intelligence is property of living cells.
That's cells which replicate , differentiate to different types of cells and integration of those different cell to new modules and drop out of useless cells.
@@Adhil_parammel I do not believe that biological cells are critical for consciousness existence. No reasons for that, because biological cells are also machines. Nano chemical mechanisms. Cells consist of the same atoms as computer processor
mass extinction?, World Peace ?
Macalo
, abiceli
Alpha zero kramniq papper didn't find out anything new learned by alpha zero.
It checked things,what present in alphazero that already we know in stock fish.
Alpha Zero's modus operandi came before Stockfish adopted it several years later.
Just have to point out! Programming using a computer or a robot controlled by a computer can't be a standalone intelligence. It's made and programmed or produced by intelligence human. Also, it has nothing to do with natural sciense, just technology!
A program or technological device that has an ability to examine data and postulate new ways of doing that is 'sort of learning' surely ? Whatever it 'is' its certainly able to modulate and modify what it is. Given its ability with data ( reading hundreds of books in a millisecond ) and its burgeoning plasticity and complexity I feel its more of an exiting ..or scary, prospect than most of us can wrap our minds around .
@@bobjary9382 surley indeed I agree about that Sir,,,,, A technology , as you said: "I feel its more of an exiting ..or scary, prospect than most of us can wrap our minds around ."
You are thinking of old school AI I think in which the behavior of the AI is explicitly coded. That is not the case for neural net technology.
I wonder how all this learning will fit into an android.
its just computer programs doing what WE TELL IT TO DO,. GROW UP! NO AI
@@autisticsimon12
Your scenario Simon is like training an entity, an automaton to slavishly follow a set of commands .
AI is like taking that entity and nurturing the ability it does have so it can refashion itself and test itself against vast data again and again , each time improving what it does based on the testing. Not only improving what it does but the way it does it .
Incidentally the best education functions like this
Not only teaching the child how to do something but teaching it how to be flexible in approaching learning. Learning in new ways and developing new understanding of what the ramifications may be once a new learning is manifested.
Zlm
wth is Zlm?
Gotta love how the first question after a lecture on protein folding is a fucking Twitter SJW crying about "societal biases".
Imagine one day Ai can milk cows automatically!
Imagine one day AI can go to space and build cities on other planets for humanity. Or use some terraforming methods.
Where is the rat level AI
Linda 🌉
Normoul
Part of the reset
There is no AI only stupidf intellects redesigning their stupidity.
Humans program computers computer programs do what we tell them, result, remove computer and you have AI
@@autisticsimon12 No the machine learning algorithms learn from the data and from training and not the programmer. They learn things that the programmer didn’t know. That is the point.
@@byrnemeister2008 I understand what your saying, but i have to disagree. People are selling this like some sort of artificial brain when its in the end more closer to "Hey Google" or "Hey Siri" than some sort of lightning strike created computer life form like Hollywood cutesy creation Johnny 5. 'Number 5 is alive.’
@@autisticsimon12 Don't watch so much movies. Read some scipub articles. Don't be so cringe 🤦♂🤦♂
Would be nice if Mr Hassabis actually worked for people rather than owners?
what?
@@BattousaiHBr I don't believe I 'stuttered'?
Erm, you don’t think cracking protein structure and publishing millions of proteins in a public databases. Accelerating drug discovery across the globe. Working out the protein structure of obscure and tropical diseases. Then giving that away for free. If you don’t think that is working for people then I don’t know what to say!
@@byrnemeister2008 Do you think that's the only thing Deep Mind works on? Besides, How convenient would it be to be able to slip an untraceable "protein folding" pill into your adversary's (political or otherwise) drink? These so -called "advances" in science aren't pursued (nor intended) in order to help the common man. Though I'll concede that their side-effects may do so from time to time. These types of programs are always for the "Owner" class, always.
@@merlepatterson ok, so you don't understand what proteins are and is just a lunatic conspiracy theorist, got it.
This is the problem I have with most public speakers. They go off on too many tangents and too many “asides” and lose traction on the primary topic.
A lot of science reporting is like this too, which is a shame. Say you see an article about some new battery type. I want to read about how this battery type works, how it is better, how quickly it'll charge, what its downsides are etc. But too many of these trashy science news sites/magazines, will go off on a tangent about how the lead scientist conducts his day "Every day Mr Scientist gets up for his morning cup of Joe. He likes watching the birds fly for ". Then it goes on "Mr Scientist wanted to change the world since he was 13", and how he "got inspiration from how a cricket stores energy in its legs". Then the obligatory "He hopes one day this discovery will change the world" (and of course you never hear of it again). The article then goes on to be 80% nonsense about the scientists personal life, hopes and dreams, and only 5% will be about the actual technology or discovery itself.
Too many youtube videos and science articles are like this now. Trashy low common denominator stuff but sadly people seem to fall for it all the time, you'll see posts saying "wow Mr Scientist is such an inspiration", I guess these posters haven't realised that pretty much 70% of every new discovery science article is pretty much exactly the same now.
Not saying this video is like this but a huge amount of science journalism is like this now. If I hear of a new discovery e.g. a way of understanding how the eye works, I know if I google for more information on it I'll mostly just find trash.
@@mikesully110 so frustrating.
welcome to your replacement ;evil
Tell them about the satellite, no really can’t mention that nobody would believe this comment anyway. Just wanted to say, “ I was here “
"This is the start of normality.." Really?
What's wrong with the host breathing into his mic? disgusting.