Page 1 of 3

a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 12:28 pm
by Bob Kane
congratulations , to Demis the man who also solved chess!, we played for the CYMA chess club once in a lifetime long ago
Quote
The Nobel Prize
@NobelPrize
·
1h
This year’s chemistry laureates Demis Hassabis and John Jumper have developed an AI model, AlphaFold2, to solve a 50-year-old problem: predicting proteins’ complex structures.

Check out two examples of protein structures determined using AlphaFold2. First up, a bacterial enzyme
Show more

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 2:10 pm
by Christopher Kreuzer

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 2:54 pm
by Nick Burrows
Should have stuck with chess imo.

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 3:10 pm
by MJMcCready
He hasn't solved chess. Even the claims that it has been partially solved are somewhat exaggerated, so it is also claimed.

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 3:14 pm
by Matt Mackenzie
It is "partially solved" in the sense that complete tablebases of play with a certain number of pieces are now available.

This is still just scratching the surface, however.

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 3:42 pm
by MJMcCready
Yes and bears no significance to actual play since the last time that anyone ended up with a position that is now classified as solved was...well never.

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 3:48 pm
by Joey Stewart
Id have doubts in the research of someone who previously claimed to have 'solved' chess - proteins have a near unlimited number of combinations as well, it feels like Nobel prizes these days are becoming increasingly redundant as there is so little left to discover

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 4:05 pm
by Jon D'Souza-Eva
Joey Stewart wrote: ↑
Wed Oct 09, 2024 3:48 pm
proteins have a near unlimited number of combinations as well
That's the whole point - out of the incredible number of possible shapes that a protein could take, how does it "decide" so quickly on the one with the minimum potential energy? If you try a simple mathematical model, starting with the atoms in the protein in any old position and then moving each one slightly to see if the new shape has a lower potential energy, you will quickly find yourself in a local minimum and certainly not the shape that the protein takes.

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 4:32 pm
by Joey Stewart
I figured it was just because proteins are created by DNA and most DNA of terrestrial life forms is pretty similar, kind of like the chess equivalent of studying players from a school that only taught one opening and then trying to base your model on how many games of chess there can be assuming they all have to begin from that position

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 4:51 pm
by Christopher Kreuzer
I don't think trying to draw analogies between chess and protein studies will help. What would help is to go and lean more about the science.

I cannot remember if all those who receive a share of a Nobel give a speech at the awarding ceremony.

Wonder what Demis will say! And you often get some new bits of information in the biographical material prepared.

There is a transcript here of what Demis said when receiving the news:

https://www.nobelprize.org/prizes/chemi ... interview/
Well, it’s unbelievably special and, you know, it’s actually really surreal, to be honest. It hasn’t really sunk in, but it’s an incredible honour. You know, it’s the big one, really.
(It gets better than that, worth reading.)

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 5:52 pm
by Roger Lancaster
Congratulations to Demis. Any comment I might make regarding proteins would just advertise my ignorance (and that applies to much else as well as proteins) but I should like to question the assertion, earlier, that "...it feels like Nobel prizes these days are becoming increasingly redundant as there is so little left to discover". How exactly does one determine how much or little there is left to discover? Are there no more "unknown unknowns"?

Re: a nobel prize for Demis Hassabis

Posted: Wed Oct 09, 2024 8:17 pm
by John Clarke
Roger Lancaster wrote: ↑
Wed Oct 09, 2024 5:52 pm
Congratulations to Demis. Any comment I might make regarding proteins would just advertise my ignorance (and that applies to much else as well as proteins) but I should like to question the assertion, earlier, that "...it feels like Nobel prizes these days are becoming increasingly redundant as there is so little left to discover". How exactly does one determine how much or little there is left to discover? Are there no more "unknown unknowns"?
Physicists in the late 19th century were saying much the same sort of thing: all the important discoveries had been made, and there wasn't much left to accomplish apart from more and more accurate evaluations of certain constants. And then they found that the phenomenon known as "black-body radiation" couldn't be explained by any of the known laws of physics ...

Re: a nobel prize for Demis Hassabis

Posted: Thu Oct 10, 2024 5:14 am
by IM Jack Rudd
MJMcCready wrote: ↑
Wed Oct 09, 2024 3:42 pm
Yes and bears no significance to actual play since the last time that anyone ended up with a position that is now classified as solved was...well never.
It happens quite frequently, actually. (Game 6 of the Carlsen-Nepomniachtchi match, to give one high-profile example.)

Re: a nobel prize for Demis Hassabis

Posted: Thu Oct 10, 2024 11:01 am
by Roger de Coverly
IM Jack Rudd wrote: ↑
Thu Oct 10, 2024 5:14 am
It happens quite frequently, actually.
For example anytime someone reaches an ending with no more than six pieces on the board.

John Nunn did a series of books "Secrets of ... " where he used the computer generated tables to explore the positions and perhaps formulate some rules and principles for practical play.

Re: a nobel prize for Demis Hassabis

Posted: Thu Oct 10, 2024 2:32 pm
by MJMcCready
IM Jack Rudd wrote: ↑
Thu Oct 10, 2024 5:14 am
MJMcCready wrote: ↑
Wed Oct 09, 2024 3:42 pm
Yes and bears no significance to actual play since the last time that anyone ended up with a position that is now classified as solved was...well never.
It happens quite frequently, actually. (Game 6 of the Carlsen-Nepomniachtchi match, to give one high-profile example.)
That can't be right surely? If true how come they haven't memorised all those solved positions and know how to play them out. On move 115 that game was a theoretical draw and top professional players should have at the very least 350,000,000,000,000,000 positions memorised, if not much more than that. What was Nepo's excuse for losing that one then?