Lifelike droids, AI etc ...

[festive]

Can it paint a picture, write a song or a bestselling novel?

Have you seen the learned behaviour of deep learning. The latter two yes. Paint a picture definitely too. Its really quite impressive only restricted by our imagination which is the concern.

Did you read about how deep learning Ai dealt with a chess master. The moves played out were completely left field no expert could follow which beat the master chess player. The concern being we didn't see it coming until it was too late.

Going back to the picture yes. AI is mature enough now to interact and learn. Its all about a baseline of Ai from which they will continue to learn.

An artist for example honing their skills also have a unique signature to how they interpret their visual of the world. I think AI will just learn the concept then you show AI bot examples from different artists, which enable it to merge and randomise its own 'interpretation in a unique way when you ask it to paint a picture.

Don't forget it can access every image on the net randomise interpret then draw the image.

 

[EddieM]

Have you seen the learned behaviour of deep learning. The latter two yes. Paint a picture definitely too. Its really quite impressive only restricted by our imagination which is the concern.

Did you read about how deep learning Ai dealt with a chess master. The moves played out were completely left field no expert could follow which beat the master chess player. The concern being we didn't see it coming until it was too late.

I have no doubt, but there are simple chess problems (For instance) that AI struggles nightly to solve, however a very average chess player can.

I know AI can supposedly write songs, paint pictures etc, but they're emotionless rubbish mashups of existing creativity.

 

[JohnD]

I have no doubt, but there are simple chess problems (For instance) that AI struggles nightly to solve

I don't think that's right.

They used to be programmed to calculate possible future outcomes, by looking ten, a hundred, a thousand moves ahead, but I understand now have self-learnjng capability by observing which moves lead to later defeat.

Do you mean a simple problem that leads to defeat in the next two moves?

 

[EddieM]

I don't think that's right.

They used to be programmed to calculate possible future outcomes, by looking ten, a hundred, a thousand moves ahead, but I understand now have self-learnjng capability by observing which moves lead to later defeat.

Do you mean a simple problem that leads to defeat in the next two moves?

AI can't solve this. Humans can. (regardless of the fact there seem to be 3 black bishops?)

Not a chess player, so can't comment specifically.

https://wonderfulengineering.com/can-you-solve-this-chess-puzzle-that-ai-software-cant/

 

[HERTS P&D]

You can have an extra Bishop if one of your Pawns have gone to the other end.

Andy

 

[Trazor]

You can have an extra Bishop if one of your Pawns have gone to the other end.

Andy

You could in theory, promote all your pawns to become Bishops, so having 10 Bishops on the board,(Never going to happen).

 

[HERTS P&D]

Queen every time.

Andy

 

[Trazor]

Queen every time.

Andy

There are some Chess puzzles, where eventually, you need to promote a Pawn as a Knight, to give immediate checkmate.

 

[Swwils]

Can it do it with no input, I.e paint me a picture, without resorting to an "of what"?

Yes.

Similar questions were asked if alpha GO, which is a deep learning AI for the game GO.

GO has alot of cultural impact and historically is considered a game of each player's total life experiences since after a few moves you cannot logically compute the next (unlike chess where you can at least a few moves ahead.) So many believed that when you play go, you are literally playing against the opponents "soul".

Alpha GO was trained on the games of several hundred human go players and against itself.

Famously when it played it's games it created brand new never before seen moves.

Who's to say that it is "emotionless" and does not have "soul" - because it certainly learned and then surpassed the human data it trained on.

 

[Brigadier]

View attachment 220344

AI can't solve this. Humans can. (regardless of the fact there seem to be 3 black bishops?)

Not a chess player, so can't comment specifically.

https://wonderfulengineering.com/can-you-solve-this-chess-puzzle-that-ai-software-cant/

Your link is from 2017.

I doubt AI hasn't overcome this now.

I was listening to a podcast about chess AI the other day.

Traditionally, computer set-ups had fed to them pretty much every pattern of pieces going, and just churned through them all to beat their opponent.

This particular AI set-up was just taught the rules of chess (presumably including the pawn-promotion thang), then was set off to play against itself.
Tens of millions of times.

It learnt what worked, and what didn't.

It is now reportedly unbeatable, IIRC.

 

[ellal]

That's merely replication.. nothing particularly interesting.

But that's just what a human being does...

Learns from it's surroundings, sees something and copies/adds/subtracts from it...

Of course a human being can also at the moment come up with interpretations/replications/alterations of what it sees/senses far better than AI , but AI is catching up fast...

It is simply a matter of influence and then how that manifests itself...

One has to ask for example, had there been no band called the Beatles would there have been a band like Oasis?

 

[Brigadier]

would there have been a band like Oasis?

For this alone, The Beatles have a lot to answer for, imho.