Garry Kasparov on AI

[Christopher Kreuzer]

See interview in Wired from 21 February 2020:

Defeated Chess Champ Garry Kasparov Has Made Peace With AI

Some interesting comments.

See also the Chessbase exceprts from this interview, which also links to this:

Chess, a Drosophila of reasoning

That is an editorial in Science magazine by Kasparov in December 2018 (not sure if it was mentioned here back then).

[JustinHorton]

Is there any reason to think Kasparov is actually expert in this field at all?

[Joey Stewart]

Only that he has had a strong interest in chess computers since they first started to get good so you would imagine he has above average knowledge in the field.

[Nick Burrows]

He also appeared on this podcast, talking about chess and AI:

https://youtu.be/8RVa0THWUWw

[NickFaulks]

Is there any reason to think Kasparov is actually expert in this field at all?

What can you mean? He's an expert on everything.

[David Robertson]

Is there any reason to think Kasparov is actually expert in this field at all?

None whatsoever.

He was invited to join Nick Bostrum's at his Future of Humanity Institute at Oxford University, without doubt as a name-recognisable, sponsorship-seeking intiative by the Institute. In his own book on the subject: Deep Thinking, Kasparov jokes, with leaden cod-humility, that at meetings of the Institute he was "the most stupid person in the room". Surprising no one, certainly not me, given the tripe outlined in 'Deep Thinking' (written, where it has any accuracy, by Mig Greengard).

Kasparov's principal, and only, USP is that he's faced an AI. After the chess is done, he's got nothing to say. Which doesn't stop him plagiarising what credible others have said

[Francis Fields]

Since the first chess algorithm was written by Alan Turing in the 1950's, the chess computers have relied on calculation to improve their strength as the computers got faster. AlphaZero was the game-changer as it learned how to play by itself. How this was achieved is serious science and if there are applications in medicine as Kasparov claims, then it is unlikely to concern me.

The idea that billions of pieces of data is a business advantage is not something that I, as a consumer, believe. If enough of his thoughts are strung together then it ends up sounding like Sci-Fi.

[Roger de Coverly]

Since the first chess algorithm was written by Alan Turing in the 1950's, the chess computers have relied on calculation to improve their strength as the computers got faster.

Until Alpha Zero, they also had to use the rules their programmers taught them. So they would calculate ahead and then check the resulting position against the evaluation, retaining only those lines that lead to positions thought worthy of further consideration.

[Ted Black]

AlphaZero was the game-changer as it learned how to play by itself. How this was achieved is serious science and if there are applications in medicine as Kasparov claims, then it is unlikely to concern me.

If enough of his thoughts are strung together then it ends up sounding like Sci-Fi.

AI is already better at detecting certain types of cancers than specialist doctors. Deep Mind was not just created to play Chess and Go, but possibly to solve real world problems, such as medical diagnosis and financial algorithms, etc. And I would say that many of the technologies we take for granted now would have seemed like Sci-Fi fifty years ago too!

[John McKenna]

Much of what is said above may well be true.

However, in the final analysis, the only solution to "financial algorithms" is likely to be "42" (plus or minus an infinite margin for human error).

[Francis Fields]

Until Alpha Zero, they also had to use the rules their programmers taught them. So they would calculate ahead and then check the resulting position against the evaluation, retaining only those lines that lead to positions thought worthy of further consideration.

I think the thoughts are in the wrong order. First, calculate ahead, then 'rules their programmers taught them' and only then 'further consideration'.

[Christopher Kreuzer]

AlphaZero was the game-changer as it learned how to play by itself. How this was achieved is serious science and if there are applications in medicine as Kasparov claims, then it is unlikely to concern me.

If enough of his thoughts are strung together then it ends up sounding like Sci-Fi.

AI is already better at detecting certain types of cancers than specialist doctors. Deep Mind was not just created to play Chess and Go, but possibly to solve real world problems, such as medical diagnosis and financial algorithms, etc. And I would say that many of the technologies we take for granted now would have seemed like Sci-Fi fifty years ago too!

Popping back to this thread, to quote the above, as it is particularly topical now in the Covid-19 pandemic as Demis Hassabis has been reported as taking part in Sage meetings:

Google executive took part in Sage meeting, tech firm confirms

The reaction in the press has been negative, which is worrying.

[J T Melsom]

'The reaction in the press has been negative, which is worrying.'

Why is it worrying that the press are asking questions? If Christopher Kreuzer sat in on a SAGE meeting don't you think people would want to know why. At present we don't know who attends SAGE meetings and what influence they might bring to bear on government policy and what commercial gain might accrue to private business representatives. We may not see full minutes of the meetings, but a list of attendees doesn't seem unreasonable.

[Kevin Thurlow]

"but a list of attendees doesn't seem unreasonable."

It is unreasonable unfortunately. Some of the people attending would not wish to be identified, (and maybe are not permitted to be identified) because of the nature of their work.

[John McKenna]

So, there is a partial list of attendees of SAGE.

And, I believe, minutes were made available retrospectively until some time in March.

It has also pointed out that the motor vehicle traffic levels disappeared fron the transport stats slide at yesterday's 5pm briefing at No.10!?

AI is already being used in various ways in the 'fight' against the virus.