Google AI wins second Go game against top player

AI is getting better, but there is a huge difference between beating humans in games—where the rules are well established and contained—and the singularity.


L. Spiro

Just started reading Nick Bostrom's Super Intelligence. Not sure whether to be worried or excited but I am certainly intrigued.

AI is getting better, but there is a huge difference between beating humans in games—where the rules are well established and contained—and the singularity.

So much this.

Right now the engines are just doing direct number crunching, searching trees, and and looking up terms with fuzzy logic from a massive database.

I'd be far more concerned if they were in a tight feedback loop of a distributed system for generating code from rules, compiling the code, and using the results to rebuild the rules. That's a system nobody should ever build.

The human player should be spending his evenings coming up with a radical new strategy, he would certainly win then even if its a suboptimal strategy vs a human. I think back to when Riker rescued Picard, the Borg lost because the experience to deal with the unorthodox strategy didn't exist.

The human player should be spending his evenings coming up with a radical new strategy, he would certainly win then even if its a suboptimal strategy vs a human. I think back to when Riker rescued Picard, the Borg lost because the experience to deal with the unorthodox strategy didn't exist.

This just isn’t realistic. It takes years upon years, and invariably consumes all of your primary growth years (the years when you learn best), to get just where you are against humans.
The best player can’t just retrain all of that.
And by the time a new human could be “engineered” through focused training, the AI will have gotten better to the point of being humanly out-of-reach.


L. Spiro

It's no secret that dumb (non-conscious) AIs are going to keep advancing and moving into industry, much like how industrial revolution decimated whole classes of jobs that used to exist.
Roles from chauffeur, to chef, to journalist to lawyer are all going to be automated. They'll be niche to begin with and seen as curiosities, but the Automobile was seen as a niche fad that wouldn't last too :p
The first personality constructs - digital clones of human minds are also being which is really going to shake the foundations of everything... If they appear to be conscious, what then? It's been argued that as consciousness is impossible to observe, there's no difference between the appearance of consciousness and actual consciousness (meaning The Chinese Room would be conscious under that definition, even though we know it's not alive)... However modern neuroscience has discovered ways to measure the consciousness-potential of a lump of matter (be it wet tissue or not), and has diagnosed brain-dead people and people with locked-in-syndrome. What if we one-up the philosophers and actually provide proof of a ghost in a machine?
When these bodiless people start taking our jobs, we'll have to coin a new xenophobia-esque term to describe the social backlash.

The human player should be spending his evenings coming up with a radical new strategy, he would certainly win then even if its a suboptimal strategy vs a human. I think back to when Riker rescued Picard, the Borg lost because the experience to deal with the unorthodox strategy didn't exist.

That is also what these ANN's are doing. Often they come up with strategies that have never occurred to humans, and pick up on (you could say, intuit based on) clues and patterns that are too subtle for us to notice. When they train against themselves, they also learn to beat them and are forced to invent even more esoteric strategies.

I'd be far more concerned if they were in a tight feedback loop of a distributed system for generating code from rules, compiling the code, and using the results to rebuild the rules. That's a system nobody should ever build.

Then nobody should have children as that's what they are :lol:

You could argue that if we're capable of creating a superior being, it would be immoral not to.
If you look at all earthly life as a single species (and conscious AI as part of life), then all that matters is getting life off this rock to survive the eventual destruction of earth. Non-physical life is going to be better at that goal than we are. If the ancestors of homo-sapians could've forseen their downfall and killed off their evolving offspring, we'd be even further from that goal still.

The human player should be spending his evenings coming up with a radical new strategy, he would certainly win then even if its a suboptimal strategy vs a human. I think back to when Riker rescued Picard, the Borg lost because the experience to deal with the unorthodox strategy didn't exist.

This just isn’t realistic. It takes years upon years, and invariably consumes all of your primary growth years (the years when you learn best), to get just where you are against humans.
The best player can’t just retrain all of that.
And by the time a new human could be “engineered” through focused training, the AI will have gotten better to the point of being humanly out-of-reach.


L. Spiro

Ok, maybe a bit of stretch to guarantee victory but he certainly wont win with raw logic, he needs to think outside the box, its his only advantage.

RivieraKid, on 11 Mar 2016 - 07:33 AM, said:
The human player should be spending his evenings coming up with a radical new strategy, he would certainly win then even if its a suboptimal strategy vs a human. I think back to when Riker rescued Picard, the Borg lost because the experience to deal with the unorthodox strategy didn't exist.
That is also what these ANN's are doing. Often they come up with strategies that have never occurred to humans, and pick up on (you could say, intuit based on) clues and patterns that are too subtle for us to notice. When they train against themselves, they also learn to beat them and are forced to invent even more esoteric strategies.

Isn't that kind of how deep blue beat Kasparov?
It sacrificed a piece that was completely unexpected which distracted Kasparov because he didn't think a logically thinking machine would do such a thing.

Humans Need Not Apply is a fairly short little video that is well worth watching and considering the long term and economic impact of advancing technology.