*

Offline Baby Thork

  • *
  • Posts: 2228
  • I am Baby Thork. Hear me roar!
    • View Profile
Google AI
« on: June 24, 2018, 04:11:32 PM »
Welp, brand new forum, more flexibility to look at all kinds of conspiracy theories etc so I'll kick things off.

https://www.chess.com/news/view/google-s-alphazero-destroys-stockfish-in-100-game-match

So Google AI has been coming on in leaps and bounds. They basically have a neural network, give the thing the rules of whatever it is doing and let it work out how to do it. In the example above, google told their AI about the rules of chess. And that's it. Just the rules. They didn't teach it any openings, any tactics, they taught it absolutely nothing. 4 hours later, not only could it destroy any human on earth, it also stuffed the world's best chess program in a 100 game match by winning 28, drawing 72 and losing zero games.

One of the games I have seen is very interesting. Google wins as black (harder to do). Twice during the game, white offered a draw by repeating the same move 3 times. In both cases, google's AI decided to weaken its own position by refusing the draw and picking an alternative move. This was completely unexpected by programmers who expected it to only ever want to strengthen its position in a game. Stockfish (the rival machine) ends up resigning about 83 moves in ... note most people didn't even know stockfish could resign. No human has forced it to do that. Also Stockfish calculates 70million moves per second ... google's machine was only looking at 80,000 moves per second. It seems it just understood chess better than anyone has ever done before. Google has only released 10 of the 100 games to the public, but it is already clear it doesn't play chess like any human, or any computer for that matter. It really does have its own way of doing it.

Google have also used the same AI to dominate Dota 2 and a game called Go.

The thing is, I wonder what else google uses this for, and would be interested to see your opinions? For example, tax avoidance would be an obvious choice. The rules are known (the law), let the algorithm find ways of being creatively efficient at avoiding tax and milking subsidies. 
Serving adverts more likely to make you buy things is another obvious choice.
Altering political opinion by pushing certain view points to sway elections is another I can see google doing, using the algorithm to be as persuasive as possible.

Google keep talking about curing cancer and other altruistic things like this ... but I don't see them using this power for good.

What do you think google will do with AI, and how can the public defend themselves against it?
« Last Edit: June 24, 2018, 04:16:21 PM by Baby Thork »
Rate this post.      👍 6     👎 1

*

Offline Baby Thork

  • *
  • Posts: 2228
  • I am Baby Thork. Hear me roar!
    • View Profile
Re: Google AI
« Reply #1 on: June 24, 2018, 04:40:07 PM »
I managed to find a commentary of the google win as black that has been released, for those not wanting to look through a list of moves.


Rate this post.      👍 6     👎 1

*

Offline Rama Set

  • *
  • Posts: 5682
  • Round and round...
    • View Profile
Re: Google AI
« Reply #2 on: June 24, 2018, 09:30:27 PM »
Wrong forum?
You don't get races of anything ... accept people.

*

Offline Baby Thork

  • *
  • Posts: 2228
  • I am Baby Thork. Hear me roar!
    • View Profile
Re: Google AI
« Reply #3 on: June 24, 2018, 10:59:38 PM »
Is it?

My understanding is that you can look at a wider variety of topics in here, but are still subject to upper fora rules. To that end, it is you that made the error with a low content off-topic post.

Its a new thing, we'll see how it plays out. The basis is to have a broader array of topic flexibility, with a stringent set of forum rules to keep things civil and productive. We wanted to encourage more debate, and not just keep doing the same old tired debate about gravity and sunsets in the upper fora. Its an experiment. If it works and we all enjoy it, we'll keep it. If it fails, we'll can it Google Glass style. Read the debate club threads in S&C.

Google is a company pushing earth shaped propaganda, so discussion about its credibility in other areas would be valid ... as far as I understand. So, do you have anything to say about google, AI or chess programs?
« Last Edit: June 24, 2018, 11:05:52 PM by Baby Thork »
Rate this post.      👍 6     👎 1

*

Offline Rama Set

  • *
  • Posts: 5682
  • Round and round...
    • View Profile
Re: Google AI
« Reply #4 on: June 25, 2018, 12:27:04 PM »
Yeah my bad. I thought that it still had to be FE related.

I don’t think the Google AI dominated go, but it certainly won against the world’s top player. Another AI, not developed by Google, defeated a bunch of human’s in a poker game.  What made this interesting is that unlike chess or go, poker has a component of hidden information and theory of mind components in betting. The AI adapted to both of these elements and learned to call bluffs.
You don't get races of anything ... accept people.

*

Offline Baby Thork

  • *
  • Posts: 2228
  • I am Baby Thork. Hear me roar!
    • View Profile
Re: Google AI
« Reply #5 on: June 25, 2018, 12:38:38 PM »
Yeah my bad. I thought that it still had to be FE related.

I don’t think the Google AI dominated go, but it certainly won against the world’s top player. Another AI, not developed by Google, defeated a bunch of human’s in a poker game.  What made this interesting is that unlike chess or go, poker has a component of hidden information and theory of mind components in betting. The AI adapted to both of these elements and learned to call bluffs.
I've actually written poker algorithms before. I wrote one for texas holdem based on the probability of winning any hand based on what's in my hand, whats on the table and what others could conceivably be holding by probability, what I might draw on the river, what they might get etc (odds for each stage). The problem is, whilst it worked and could say "you have a 45% chance of winning this hand over the other 4 people left in", I could never wrap my head around the maths of how much to bid to lure maximum money from people, to stop them folding because my algoritmn was predicting say 97% chance of winning so it went all in etc. I could get it to work out if the pot money vs the next stake was worth it to 'see' my opponent cards, but should my algorithm raise by $4, $6, $7? I had no idea how to sort that. And it seemed how much you bid, is actually more important than the cards you hold ... you hold average cards, you need to stay in but not lose too much.

I have recently been looking at tensor flow (from a hobbiest point of view), and it might be able to solve these problems for me, and as you say, be able to call a bluff ... which I wasn't even close to getting the maths right for. I could only say, odds in my favour, bet, not in my favour, don't bet ... binary ... and that won't beat a top player even if I know the odds. I did have 'the gun' in my probability, and the algorithm I wrote would know where on the table it was and calculate the odds dependant on its seat when asked to bid ... an instantaneous set of odds.

But I won't be using tensorflow for this ... gambling sites are already onto this and now actively hunt down signs of machine learning. That window has passed.  :)


Machine learning would be the ultimate answer to earth's shape. Not even Tom Bishop would argue because it is based on observable science. You don't give ML any assumptions. You just feed it data and it iterates repeatedly until it finds the answer. The problem with ML in today's format, is that whilst we'd end up knowing what shape the earth is, we'd have no idea how the machine came to that conclusion, we'd only know it is right. Much like we have no idea how Google's AI plays chess. It just does it.

I have a theory that ML will actually cause a 'great ignorance'. Lets say you got ML to start predicting the weather. Now if would just look at all the data from ocean buoys, airport reports, temp, pressure, visibility, dew point etc etc. And it would work out the weather ... and it would be far more accurate than anything we have today. Maybe we'd end up with a 30 day forecast. People at the weather service would abandon trying to predict the weather, the machine does it better, but no one knows how it does it. So, you'd have a meteorological office filled with people that could write machine learning, and no one who actually knew how to predict weather themselves from data as no one is employed to do that. The science would grind to a halt. There is no point learning a solved problem, as it is useless, weather isn't predicted that way any more.
Span this through multiple industries such as medical cures, accountancy, logistics, etc ... no one would have any knowledge or skill whatsoever. But I guess this is why people think AI will kill jobs for billions of people. We'll all be dumb, unskilled and unneeded.
« Last Edit: June 25, 2018, 01:02:17 PM by Baby Thork »
Rate this post.      👍 6     👎 1

*

Offline Rama Set

  • *
  • Posts: 5682
  • Round and round...
    • View Profile
Re: Google AI
« Reply #6 on: June 25, 2018, 01:04:06 PM »
Yeah my bad. I thought that it still had to be FE related.

I don’t think the Google AI dominated go, but it certainly won against the world’s top player. Another AI, not developed by Google, defeated a bunch of human’s in a poker game.  What made this interesting is that unlike chess or go, poker has a component of hidden information and theory of mind components in betting. The AI adapted to both of these elements and learned to call bluffs.
I've actually written poker algorithms before. I wrote one for texas holdem based on the probability of winning any hand based on what's in my hand, whats on the table and what others could conceivably be holding by probability, what I might draw on the river, what they might get etc (odds for each stage). The problem is, whilst it worked and could say "you have a 45% chance of winning this hand over the other 4 people left in", I could never wrap my head around the maths of how much to bid to lure maximum money from people, to stop them folding because my algoritmn was predicting say 97% chance of winning so it went all in etc. I could get it to work out if the pot money vs the next stake was worth it to 'see' my opponent cards, but should my algorithm raise by $4, $6, $7? I had no idea how to sort that. And it seemed how much you bid, is actually more important than the cards you hold ... you hold average cards, you need to stay in but not lose too much.

I have recently been looking at tensor flow (from a hobbiest point of view), and it might be able to solve these problems for me, and as you say, be able to call a bluff ... which I wasn't even close to getting the maths right for. I could only say, odds in my favour, bet, not in my favour, don't bet ... binary ... and that won't beat a top player even if I know the odds. I did have 'the gun' in my probability, and the algorithm I wrote would know where on the table it was and calculate the odds dependant on its seat when asked to bid ... an instantaneous set of odds.

But I won't be using tensorflow for this ... gambling sites are already onto this and now actively hunt down signs of machine learning. That window has passed.  :)


Machine learning would be the ultimate answer to earth's shape. Not even Tom Bishop would argue because it is based on observable science. You don't give ML any assumptions. You just feed it data and it iterates repeatedly until it finds the answer. The problem with ML in today's format, is that whilst we'd end up knowing what shape the earth is, we'd have no idea how the machine came to that conclusion, we'd only know it is right. Much like we have no idea how Google's AI plays chess. It just does it.

There recently was a formal debate between an AI and humans, and the AI provided substantive data to back up its conclusions.  If you borrowed elements from this you could also get data on how the AI came to the conclusion.
You don't get races of anything ... accept people.

*

Offline Baby Thork

  • *
  • Posts: 2228
  • I am Baby Thork. Hear me roar!
    • View Profile
Re: Google AI
« Reply #7 on: June 25, 2018, 01:06:17 PM »
There recently was a formal debate between an AI and humans, and the AI provided substantive data to back up its conclusions.  If you borrowed elements from this you could also get data on how the AI came to the conclusion.
I hadn't seen that. Its an obvious thing to want to know how the answer is what it is. It needs building into all AI.

I'll bet that is how they'll make money. ML programs like tensorflow are free. So you can find the answer to anything you want. But if you want to know how the machine arrived at that answer so you can understand it too, that's the premium module. That's what I'd do.  ;)

I don't look forward to the first AI bot joining this forum to debate earth's shape.  :(
« Last Edit: June 25, 2018, 01:12:27 PM by Baby Thork »
Rate this post.      👍 6     👎 1

*

Offline Rama Set

  • *
  • Posts: 5682
  • Round and round...
    • View Profile
Re: Google AI
« Reply #8 on: June 25, 2018, 06:22:54 PM »
There recently was a formal debate between an AI and humans, and the AI provided substantive data to back up its conclusions.  If you borrowed elements from this you could also get data on how the AI came to the conclusion.
I hadn't seen that. Its an obvious thing to want to know how the answer is what it is. It needs building into all AI.

I'll bet that is how they'll make money. ML programs like tensorflow are free. So you can find the answer to anything you want. But if you want to know how the machine arrived at that answer so you can understand it too, that's the premium module. That's what I'd do.  ;)

I don't look forward to the first AI bot joining this forum to debate earth's shape.  :(

Tom's been here for years  ??? Jokes aside, AI is one of the things that makes me nervous about the future and not because of Skynet, but because it encourages humans to reduce the agency in their affairs.
You don't get races of anything ... accept people.

Re: Google AI
« Reply #9 on: June 26, 2018, 03:30:17 AM »

Machine learning would be the ultimate answer to earth's shape. Not even Tom Bishop would argue because it is based on observable science. You don't give ML any assumptions. You just feed it data and it iterates repeatedly until it finds the answer. The problem with ML in today's format, is that whilst we'd end up knowing what shape the earth is, we'd have no idea how the machine came to that conclusion, we'd only know it is right. Much like we have no idea how Google's AI plays chess. It just does it.


Why would you think this? When humans look at the evidence and show proof that the Earth is round, you discount the evidence. When AI comes along, are you going to withhold from it pictures from space? Will you tell it that there are no power lines over Lake Pontchartrain? That the sun doesn't move across the sky at a constant rate of 15 degrees per hour?

I see no reason why flat Earth adherents couldn't either insist on omitting large portions of relevant evidence, or just dismissing outright any conclusions that differ from their preconceived flat Earth model.

*

Offline Baby Thork

  • *
  • Posts: 2228
  • I am Baby Thork. Hear me roar!
    • View Profile
Re: Google AI
« Reply #10 on: June 26, 2018, 01:38:31 PM »
You'd just need to agree on a dataset. Something everyone knows to be true and can prove themselves. Then let the computer iterate away, extrapolate and turn into a defined shape.
Rate this post.      👍 6     👎 1

Max_Almond

Re: Google AI
« Reply #11 on: June 26, 2018, 01:49:01 PM »
There recently was a formal debate between an AI and humans, and the AI provided substantive data to back up its conclusions.  If you borrowed elements from this you could also get data on how the AI came to the conclusion.

Substantive data, you say?

Well that proves that -

Nah, no need to finish that sentence. ;)

Re: Google AI
« Reply #12 on: June 26, 2018, 01:53:41 PM »
You'd just need to agree on a dataset. Something everyone knows to be true and can prove themselves. Then let the computer iterate away, extrapolate and turn into a defined shape.
Seeing as we can't even agree on that here on the forums, this seems one of those 'easier said than done' things to me. It would be fascinating to me to see what one might turn out like/into if it was just set loose on these forums for a while though. What conclusions it might come to with as little in the way of biases as possible.

Re: Google AI
« Reply #13 on: June 26, 2018, 02:00:11 PM »
This simply cements the reality this forum and the other FE site are mostly AI bots.

*

Offline Rama Set

  • *
  • Posts: 5682
  • Round and round...
    • View Profile
Re: Google AI
« Reply #14 on: June 26, 2018, 03:05:24 PM »
There recently was a formal debate between an AI and humans, and the AI provided substantive data to back up its conclusions.  If you borrowed elements from this you could also get data on how the AI came to the conclusion.

Substantive data, you say?

Well that proves that -

Nah, no need to finish that sentence. ;)

I wasn't trying to prove anything. What do you mean?
You don't get races of anything ... accept people.

Re: Google AI
« Reply #15 on: June 26, 2018, 04:10:45 PM »
You'd just need to agree on a dataset. Something everyone knows to be true and can prove themselves. Then let the computer iterate away, extrapolate and turn into a defined shape.

should be possible to input all know flights into this system and it could then draw a true map based on all the triangulation based on the data set??
Quote from: SiDawg
Planes fall out of the sky all the time

*

Offline Baby Thork

  • *
  • Posts: 2228
  • I am Baby Thork. Hear me roar!
    • View Profile
Re: Google AI
« Reply #16 on: June 26, 2018, 08:15:38 PM »
You'd just need to agree on a dataset. Something everyone knows to be true and can prove themselves. Then let the computer iterate away, extrapolate and turn into a defined shape.

should be possible to input all know flights into this system and it could then draw a true map based on all the triangulation based on the data set??
No, aircraft don't fly direct. They go via way points, beacons, they go round in circles in holding patterns, wind effects them, temp, pressure, number of passengers and fuel also effects cruising speed, as does the centre of gravity which is load dependant (Where the passengers sit). This would be a horrible way to work it out.
Rate this post.      👍 6     👎 1

*

Offline Rama Set

  • *
  • Posts: 5682
  • Round and round...
    • View Profile
Re: Google AI
« Reply #17 on: June 26, 2018, 09:09:24 PM »
You'd just need to agree on a dataset. Something everyone knows to be true and can prove themselves. Then let the computer iterate away, extrapolate and turn into a defined shape.

should be possible to input all know flights into this system and it could then draw a true map based on all the triangulation based on the data set??
No, aircraft don't fly direct. They go via way points, beacons, they go round in circles in holding patterns, wind effects them, temp, pressure, number of passengers and fuel also effects cruising speed, as does the centre of gravity which is load dependant (Where the passengers sit). This would be a horrible way to work it out.

If the AI had all that data, it should eventually be able to model the space the aircraft fly in. With all those variables it seems like it would be pretty definitive too as there are likely very few solutions that are intelligible -and- fit the data.
You don't get races of anything ... accept people.

Re: Google AI
« Reply #18 on: June 27, 2018, 05:38:09 AM »
You'd just need to agree on a dataset. Something everyone knows to be true and can prove themselves. Then let the computer iterate away, extrapolate and turn into a defined shape.

Humans did that, starting 2000 years ago. Why do you think that an AI would come to a different conclusion in, what, a month, than humans have come to in things of years? Computers are good at solving problems faster than humans, but when humans have a multi millennia head start, how could it come to a different conclusion?

You will need to fight tooth and nail to exclude all the evidence you wish to discount in our human discussions from consideration by the AI. It's the same exact problem this forum had been hashing out. AI isn't magic, it can't possibly solve this problem any better than humans already have.

*

Offline Tumeni

  • *
  • Posts: 1427
    • View Profile
Re: Google AI
« Reply #19 on: June 27, 2018, 01:53:39 PM »
... aircraft don't fly direct. They go via way points, beacons, they go round in circles in holding patterns, wind effects them, temp, pressure, number of passengers and fuel also effects cruising speed, as does the centre of gravity which is load dependant (Where the passengers sit).

(slightly off-topic, I know, but ...)

.... why don't you present this argument to anyone who claims the ISS is really a high-altitude plane? For all of this is sure proof that the ISS, with its regular flight path, absolutely-perfect orbit timing, etc. CANNOT be a plane.
==============================
==============================
Pete Svarrior "We are not here to directly persuade anyone ... You mistake our lack of interest in you for our absence."

Tom Bishop "We are extremely popular and the entire world wants to talk to us. We have better things to do with our lives than have in depth discussions with every single curious person. You are lucky to get one sentence dismissals from us"