Is The A.I. Singularity Coming And If So When?

Is The A.I. Singularity Coming And If So When?

Author
Discussion

AshVX220

5,929 posts

189 months

Wednesday 23rd March 2016
quotequote all
mondeoman said:
http://qz.com/636637/the-beginning-of-the-end-goog...

Wasn't the original site, but it'll do.
Very interesting piece, thanks. beer

Guvernator

13,104 posts

164 months

Wednesday 23rd March 2016
quotequote all
AshVX220 said:
mondeoman said:
http://qz.com/636637/the-beginning-of-the-end-goog...

Wasn't the original site, but it'll do.
Very interesting piece, thanks. beer
Yes it IS an interesting piece in what they've achieved but again why with the very unnecessary Rise of the Machines hyperbole? It really is annoying that any time anyone mentions AI, the first conclusion we jump to is it's going to wipe us out. I find this a very negative view on what could be one of the most important areas of human research ever.

RobDickinson

31,343 posts

253 months

Wednesday 23rd March 2016
quotequote all
The Hollywood effect, because Hollywood is so accurate on everything else too...

mondeoman

11,430 posts

265 months

Wednesday 23rd March 2016
quotequote all
I don't see it as Hollywood, more of a dealing with a complete unknown...

We just don't know how an AI is going to evolve, and the more I read, the more I get that feeling.


glazbagun

14,259 posts

196 months

Wednesday 23rd March 2016
quotequote all
I'm sure the guys who're at the cutting edge of this stuff are already running into philisophical problems with the "should we do this because we can do it" thing. Like Big Data and personal information only became an issue for legislation by the time it was already too late. (How quaint does it look in 2016 to think that only ten years ago everyone was up in arms about ID cards!).

At the moment we can tell when an neural net doesn't understand that a dumbbell is seperate from an arm, but how will we be able to spot its errors when it's controlling a fusion reaction faster than we ever can, or if it tries to assimilate a new data set which is full of garbage.

I don't see it as a malevolent force, but if we can't get a Windows OS to work well, I can imagine a buggy AI causing some damage because it doesn't understand what it's doing, like HAL 9000.

Moonhawk

10,730 posts

218 months

Wednesday 23rd March 2016
quotequote all
glazbagun said:
I don't see it as a malevolent force, but if we can't get a Windows OS to work well, I can imagine a buggy AI causing some damage because it doesn't understand what it's doing, like HAL 9000.
A system can only be as good as it's weakest link (i.e. humans).

AshVX220

5,929 posts

189 months

Thursday 24th March 2016
quotequote all
Guvernator said:
Yes it IS an interesting piece in what they've achieved but again why with the very unnecessary Rise of the Machines hyperbole? It really is annoying that any time anyone mentions AI, the first conclusion we jump to is it's going to wipe us out. I find this a very negative view on what could be one of the most important areas of human research ever.
It's the nature of the media etc really, the best news is bad news. I think the waitbutwhy piece touches on too very good points, firstly, the nature of any AI will depend on who's first to develop one. The thinking that once the real singularity happens, then that is likely to forever more be the only AI in existence because it will learn so much faster, so who develops it and how they code it will be key. If the US/UK/Google or some nation that we think is "good" make it first, then happy days. But, if some utter bd somewhere, some how manages to get to the singularity first, then all bets are off.
The second most interesting point for me, is just how quickly any Real AI will learn and improve, the thinking being that it will be as intelligent as a human for minutes, before it learns at such an amazing pace that before we know, it could have cracked most of sciences problems before we even recognise there's a problem. Leading onto the ability for an high-intelligence AI to properly crack nano-tech and the ability to manipulate material at the atomic level, which in turn with lead to either our immortality/salvation or could also end in our complete annihilation (I personally think that the boffins working on this will very quickly put safetey mechanisms in place to ensure it's the former). It's an amazing and growing field and I wish I were smart enough to be involved somehow, I think the next 20-30 years could see the first true AI, then the sky's the limit.

AshVX220

5,929 posts

189 months

Thursday 24th March 2016
quotequote all
Whoops, this is what happens when you let the public get involved, much in the same way we have a Polar Exploration vessel that people want to call Boaty McBoatface. People seem in-capable of being remotely sensible about things would are [IMHO] important. This particular AI could have learnt so much about communication.

http://www.telegraph.co.uk/technology/2016/03/24/m...

Quite sad IMO.

warp9

1,583 posts

196 months

Thursday 24th March 2016
quotequote all
Light AI twitter chat bot goes Hitler sex crazy...
http://www.telegraph.co.uk/technology/2016/03/24/m...

Edit, beaten to it! And I think it probably has learned a lot about communication.

Edited by warp9 on Thursday 24th March 12:35

IainT

10,040 posts

237 months

Thursday 24th March 2016
quotequote all
Guvernator said:
Evolution is a VERY slow process. A creature that is clever enough will be able to adapt itself or it's tools to slowly changing environment. It's what humans do now after all.
AI will, I think, be far more capable of self-adaption than we are and will have less immediate drive to adapt environment. AI iterations will be of decreasing duration - evolutionary pace will be incredibly quick.

We already have the ability to evolve ourselves that is held in check currently with all sorts of ethical objections AI would likely be unfettered from changing its own 'DNA' and we may see it jump, human assisted, from traditional silicone to something as yet unknown.

I'm really interested in the transhuman potential for AI - ranging from fixing damage to enhancing things.

otolith

55,899 posts

203 months

Thursday 24th March 2016
quotequote all
AshVX220 said:
Whoops, this is what happens when you let the public get involved, much in the same way we have a Polar Exploration vessel that people want to call Boaty McBoatface. People seem in-capable of being remotely sensible about things would are [IMHO] important. This particular AI could have learnt so much about communication.

http://www.telegraph.co.uk/technology/2016/03/24/m...

Quite sad IMO.
Alternative plot for Idiocracy style movie.

AshVX220

5,929 posts

189 months

Thursday 24th March 2016
quotequote all
otolith said:
Alternative plot for Idiocracy style movie.
Indeed.

burritoNinja

690 posts

99 months

Sunday 27th March 2016
quotequote all
I studied a bit of AI and Robotics as part of my computer science degree and on what I have studied, we are many many years away from a computer system that can even come close to the consciousness and dynamic intelligence that we as humans experience. It is barely even possible at this stage within computer science to even create AI's on a rat size brain.
I believe it will happen someday way down the line. Though will it ever really have it's own reality? Computers are dumb systems. They only do what we program them to do. Love a T-1000 as a toy.

Edited by burritoNinja on Sunday 27th March 08:57


Edited by burritoNinja on Sunday 27th March 08:58

popeyewhite

19,622 posts

119 months

Sunday 27th March 2016
quotequote all
Guvernator said:
Evolution is a VERY slow process. A creature that is clever enough will be able to adapt itself or it's tools to slowly changing environment. It's what humans do now after all.
If you understand evolution to mean adaptation to environment then humans and all animals within the human domain have just about stopped evolving. Head to an isolated volcanic island, undersea vent or South American tapui and you may have a slim chance to witness ongoing evolution.

davepoth

29,395 posts

198 months

Sunday 27th March 2016
quotequote all
popeyewhite said:
If you understand evolution to mean adaptation to environment then humans and all animals within the human domain have just about stopped evolving. Head to an isolated volcanic island, undersea vent or South American tapui and you may have a slim chance to witness ongoing evolution.
Not strictly true. Darwinian evolution perhaps, but if you look to see what breeding has done to certain pedigree dog breeds you can certainly see it in action.

In terms of AI, a true intelligence is probably a fair time away. The question really is how many jobs need actual intelligence? The answer is very few. Most jobs involve being given a set of rules and following them. Computers are very good at that.

popeyewhite

19,622 posts

119 months

Sunday 27th March 2016
quotequote all
davepoth said:
popeyewhite said:
If you understand evolution to mean adaptation to environment then humans and all animals within the human domain have just about stopped evolving.
Not strictly true. Darwinian evolution perhaps,
I think I explained that in "If you understand evolution to mean adaptation to environment"!

burritoNinja

690 posts

99 months

Sunday 27th March 2016
quotequote all
Don't know if any of you people seen the recent documentary on AI. I think it is still available on BBC iPlayer. Horizon show I believe. All about will computer system ever be able to think like us.

warp9

1,583 posts

196 months

Wednesday 30th March 2016
quotequote all
Interesting article here on MS vision of AI bots and a PR response to the Tay Hitler monkey incident. http://www.bloomberg.com/features/2016-microsoft-f...

So we're all going to have a KITT and businesses are going to be able to develop their own specific bots. I can see this hitting the service industry first.

otolith

55,899 posts

203 months

Wednesday 30th March 2016
quotequote all
popeyewhite said:
If you understand evolution to mean adaptation to environment then humans and all animals within the human domain have just about stopped evolving.
I think that's unlikely. Evolution is changing gene frequencies. Even if you narrow it to changes which are adaptive, there is still likely to be change. Sexual selection remains as potent as ever (other people are environment) and we still have differential survival and reproductive success. Especially when you consider that a minority of the global population of humans live in the first world.

AshVX220

5,929 posts

189 months

Tuesday 5th April 2016
quotequote all
Thinking about this further. As a safeguard against any AI inadvertently destroying us, surely a line of code could be put into any AI that states it will not do anything in the real physical world, or not without a "man-in-loop"? This would mean that the AI could be used by us purely for advancing our own knowledge and could then be used to do things only on the authorisation of a person.

The general thinking from what I've seen is that once we hit the singularity, any AI at that point will surpass our intelligence at an amazing rate and could therefore open up parts of science very quickly, producing theories, designing experiments in the virtual world for us to replicate on the real world. And eventually being able to create things on it's own, with that rule that it cannot effect the physical world.

It's probably a lot more complicated than that, but you get the idea (or I'm talking bks)! laugh