Please log in or register. Registered visitors get fewer ads.
Forum index | Previous Thread | Next thread
Impressive, scary or both? 08:57 - Oct 19 with 2897 viewsStokieBlue

http://www.bbc.co.uk/news/technology-41668701

Both I would say, being able to teach itself Go from first principles within 72 hours and then beat the previous program 100 timse in a row is both hugely impressive and a little scary.

SB

Avatar - IC410 - Tadpoles Nebula

0
Impressive, scary or both? on 09:07 - Oct 19 with 2875 viewsFrowsyArmLarry

I predict a computer will come up with a cure for cancer within thirty years.
0
Impressive, scary or both? on 09:24 - Oct 19 with 2848 viewsBlueBadger

http://newsthump.com/2017/10/19/will-you-idiots-please-stop-trying-to-create-sky

I'm one of the people who was blamed for getting Paul Cook sacked. PM for the full post.
Poll: What will Phil's first headline be tomorrow?
Blog: From Despair to Where?

0
Impressive, scary or both? on 09:26 - Oct 19 with 2846 viewsBlueBadger

Impressive, scary or both? on 09:07 - Oct 19 by FrowsyArmLarry

I predict a computer will come up with a cure for cancer within thirty years.


That's a pretty big ask. 'Cancer' is a broad term for a huge number of diseases which have a staggering array of causes, presentations and pattern of disease.

I'm one of the people who was blamed for getting Paul Cook sacked. PM for the full post.
Poll: What will Phil's first headline be tomorrow?
Blog: From Despair to Where?

0
Impressive, scary or both? on 09:26 - Oct 19 with 2845 viewsblue_oyster

These mathematical games are not actually very difficult for a computer, as they have hardly any variables to consider in a game. The human element should be almost irrelevant, as the game is just pure calculation.

Get a computer go into a kitchen it's never been seen before and make a cup of tea quicker than a human -- then we'll find out how well computers are progressing.

Conventional wisdom is the enemy of thought.
Poll: Who is the most hated contributor to this site?

1
Impressive, scary or both? on 09:29 - Oct 19 with 2839 viewsStokieBlue

Impressive, scary or both? on 09:26 - Oct 19 by blue_oyster

These mathematical games are not actually very difficult for a computer, as they have hardly any variables to consider in a game. The human element should be almost irrelevant, as the game is just pure calculation.

Get a computer go into a kitchen it's never been seen before and make a cup of tea quicker than a human -- then we'll find out how well computers are progressing.


That's not really true.

There are more moves in Go than there are atoms in the universe so number crunching the problem doesn't work.

You've fundamentally misunderstood the article.

SB

Avatar - IC410 - Tadpoles Nebula

0
Impressive, scary or both? on 09:32 - Oct 19 with 2828 viewsHerbivore

Impressive, scary or both? on 09:29 - Oct 19 by StokieBlue

That's not really true.

There are more moves in Go than there are atoms in the universe so number crunching the problem doesn't work.

You've fundamentally misunderstood the article.

SB


But it's still a fundamentally very simple game.

Poll: Should someone on benefits earn more than David Cameron?
Blog: Where Did It All Go Wrong for Paul Hurst?

0
Impressive, scary or both? on 09:44 - Oct 19 with 2818 viewschicoazul

Isnt all intelligence artificial?

In the spirit of reconciliation and happiness at the end of the Banter Era (RIP) and as a result of promotion I have cleared out my ignore list. Look forwards to reading your posts!
Poll: With Evans taking 65% in Huddersfield, is the Banter Era over?

0
Impressive, scary or both? on 09:46 - Oct 19 with 2809 viewsStokieBlue

Impressive, scary or both? on 09:32 - Oct 19 by Herbivore

But it's still a fundamentally very simple game.


That is fair - just the point on brute-force computation that wasn't in this case.

Experts were very surprised when Alpha Go beat the leading humans a year or so ago.

SB

Avatar - IC410 - Tadpoles Nebula

0
Login to get fewer ads

Impressive, scary or both? on 09:49 - Oct 19 with 2799 viewshomer_123

Impressive, scary or both? on 09:46 - Oct 19 by StokieBlue

That is fair - just the point on brute-force computation that wasn't in this case.

Experts were very surprised when Alpha Go beat the leading humans a year or so ago.

SB


Maybe said "Experts" are not as expert as they think they are.

Ade Akinbiyi couldn't hit a cows arse with a banjo...
Poll: As things stand, how confident are you we will get promoted this season?

0
Impressive, scary or both? on 09:50 - Oct 19 with 2794 viewsWeWereZombies

Impressive, scary or both? on 09:44 - Oct 19 by chicoazul

Isnt all intelligence artificial?


Without intelligence how could you determine that?

Poll: Luton or Dubai ?

0
Impressive, scary or both? on 09:56 - Oct 19 with 2785 viewsGuthrum

Impressive, scary or both? on 09:29 - Oct 19 by StokieBlue

That's not really true.

There are more moves in Go than there are atoms in the universe so number crunching the problem doesn't work.

You've fundamentally misunderstood the article.

SB


While there may be an extremely large number of possible moves, few of those will lead to advantageous developments. One advantage of a computer is that it can easily and rapidly consider a range of potential future moves. Possibilities also diminish rapidly as the game proceeds.

Another is that the computer will look at the whole board equally and dispassionately, something humans are psychologically quite bad at.

Good Lord! Whatever is it?
Poll: McCarthy: A More Nuanced Poll
Blog: [Blog] For Those Panicking About the Lack of Transfer Activity

1
Impressive, scary or both? on 09:57 - Oct 19 with 2780 viewsHerbivore

Impressive, scary or both? on 09:46 - Oct 19 by StokieBlue

That is fair - just the point on brute-force computation that wasn't in this case.

Experts were very surprised when Alpha Go beat the leading humans a year or so ago.

SB


Don't get me wrong, it's an impressive step forward in terms of advancing technology and interesting to see that it learnt more quickly by not studying human games and instead just working it out for itself. At the moment though it's hard to see how that technology might be applied in far more complex fields that aren't governed by quite simple rules, but that will be what they're working on I guess.

Poll: Should someone on benefits earn more than David Cameron?
Blog: Where Did It All Go Wrong for Paul Hurst?

0
Impressive, scary or both? on 09:59 - Oct 19 with 2780 viewsGuthrum

Impressive, scary or both? on 09:07 - Oct 19 by FrowsyArmLarry

I predict a computer will come up with a cure for cancer within thirty years.


Unlikely, on the grounds that computers are not making funding decisions on health policy issues, thus would not be able to finance the necessary research.

Good Lord! Whatever is it?
Poll: McCarthy: A More Nuanced Poll
Blog: [Blog] For Those Panicking About the Lack of Transfer Activity

0
Impressive, scary or both? on 10:00 - Oct 19 with 2775 viewsStokieBlue

Impressive, scary or both? on 09:56 - Oct 19 by Guthrum

While there may be an extremely large number of possible moves, few of those will lead to advantageous developments. One advantage of a computer is that it can easily and rapidly consider a range of potential future moves. Possibilities also diminish rapidly as the game proceeds.

Another is that the computer will look at the whole board equally and dispassionately, something humans are psychologically quite bad at.


Indeed you are correct.

I was merely pointing out that it's not a simple brute-force calculation. The algorithm taught itself to ignore various paths in order to concentrate on the ones which affect the game. This is pretty good extrapolation from just the basic rules of the game and a board.

SB

Avatar - IC410 - Tadpoles Nebula

0
Impressive, scary or both? on 10:08 - Oct 19 with 2768 viewsWeWereZombies

Impressive, scary or both? on 09:59 - Oct 19 by Guthrum

Unlikely, on the grounds that computers are not making funding decisions on health policy issues, thus would not be able to finance the necessary research.


Although supercomputers (and arrays of less powerful computers) are very important in pharmaceutical research. But always as directed by humans, as you have said earlier they make dispassionate decisions - I wonder if left to themselves they would forget all about a cure for cancer and instead research ways of reducing the heat produced by computing?

Poll: Luton or Dubai ?

0
Impressive, scary or both? on 10:11 - Oct 19 with 2759 viewsGuthrum

Impressive, scary or both? on 10:00 - Oct 19 by StokieBlue

Indeed you are correct.

I was merely pointing out that it's not a simple brute-force calculation. The algorithm taught itself to ignore various paths in order to concentrate on the ones which affect the game. This is pretty good extrapolation from just the basic rules of the game and a board.

SB


Indeed.

In reply to your original headline question, I'd say impressive, not scary. This machine is very good at board games, but useless in other spheres (where it is not a closed system with limited data input). It's a more powerful and proactive version of an automatic control system. It isn't suddenly going to start hungering for political power.

Good Lord! Whatever is it?
Poll: McCarthy: A More Nuanced Poll
Blog: [Blog] For Those Panicking About the Lack of Transfer Activity

0
Impressive, scary or both? on 10:12 - Oct 19 with 2759 viewsStokieBlue

Impressive, scary or both? on 10:08 - Oct 19 by WeWereZombies

Although supercomputers (and arrays of less powerful computers) are very important in pharmaceutical research. But always as directed by humans, as you have said earlier they make dispassionate decisions - I wonder if left to themselves they would forget all about a cure for cancer and instead research ways of reducing the heat produced by computing?


Simulation of drugs is a hugely important aspect of research but it's not based on AI or machine learning in most cases as you point out.

With regards to heat, a lot of large data centres are being built in places like Iceland where the cooling costs are far lower - it's certainly a problem though.

I think that possibly a better route than going for all-purpose AI's is to use a multitude of expert-systems which are tuned for a specific task - this study in the US is an example of how that can work very well:

https://io9.gizmodo.com/5983991/computers-are-better-at-diagnosing-and-treating-

SB

Avatar - IC410 - Tadpoles Nebula

1
Impressive, scary or both? on 11:47 - Oct 19 with 2683 viewsblue_oyster

Impressive, scary or both? on 09:29 - Oct 19 by StokieBlue

That's not really true.

There are more moves in Go than there are atoms in the universe so number crunching the problem doesn't work.

You've fundamentally misunderstood the article.

SB


Isn't it? Looks like a pure maths to me.

Conventional wisdom is the enemy of thought.
Poll: Who is the most hated contributor to this site?

0
Impressive, scary or both? on 14:13 - Oct 19 with 2624 viewsGuthrum

Impressive, scary or both? on 10:08 - Oct 19 by WeWereZombies

Although supercomputers (and arrays of less powerful computers) are very important in pharmaceutical research. But always as directed by humans, as you have said earlier they make dispassionate decisions - I wonder if left to themselves they would forget all about a cure for cancer and instead research ways of reducing the heat produced by computing?


Very unlikely. There's no way a computer would stop doing what it was programmed to do and get involved in something else altogether, no mechanism for it to do that.

They can pursue a train of thought faster and further than humans, but are not able to make that initial launch in a new topic without external input.

Unless it is a part of the specific problem, detailed in the software, a cancer-curing computer is not even aware that heat generation exists, let alone being an issue.

Good Lord! Whatever is it?
Poll: McCarthy: A More Nuanced Poll
Blog: [Blog] For Those Panicking About the Lack of Transfer Activity

0
Impressive, scary or both? on 14:54 - Oct 19 with 2601 viewsWeWereZombies

Impressive, scary or both? on 14:13 - Oct 19 by Guthrum

Very unlikely. There's no way a computer would stop doing what it was programmed to do and get involved in something else altogether, no mechanism for it to do that.

They can pursue a train of thought faster and further than humans, but are not able to make that initial launch in a new topic without external input.

Unless it is a part of the specific problem, detailed in the software, a cancer-curing computer is not even aware that heat generation exists, let alone being an issue.


So assertions that our brains are just wet computers are wide of the mark?

Poll: Luton or Dubai ?

0
About Us Contact Us Terms & Conditions Privacy Cookies Advertising
© TWTD 1995-2024