How far away our we from A.I.

Stiffler2

As the title says : How far away our we from A.I.

I know this vid is over a year old but still very impressive.


IBM's Watson computer taking on humans at Jeopardy - random questions on different topics

Watson destroy all humans in the game, apparently the 2 competitors are v smart, 1 guy looks well pissed at Watson - lol

https://www.youtube.com/watch?v=WFR3lOm_xhE

Watson ( from a doc I watched last night, although he could have more hardware now - ( wonder why I reffered to Watson as a he ) has 2000 processors.

IBM obviously or must have a database of knowledge about everything in the world which he access's

Either that or it's a guy behind the screen with a microphone

Still - we can't be too far away if Watson can take in a question, search for the answer in the correct context and then get the answer right.

Granted he is not thinking for himself but rather accessing a massive database with his 2000 processors and then giving the answer.

In I-Robot there was a centralised computer which controlled all robots or should I say database which all robots accessed for information.

So AH - How far away from AI are we ??

blueser

That's all well and good; but can a computer differentiate between ''our'' & ''are''?

Nothingbetter2d

i am a terminator. but i malfunctioned... damn microsoft windows 124 professional keeps bsod'ing on me. bad memory it seems.

Stiffler2

blueser wrote: »

That's all well and good; but can a computer differentiate between ''our'' & ''are''?

ah bollix :rolleyes:

Jester252

maybe it does exist and we dont know about it. Maybe it's a poster on boards and is using AH to see if human should live or die

smash

Artificial intelligence... that's when a blonde dyes her hair brown isn't it?

blueser

Stiffler2 wrote: »

ah bollix :rolleyes:

No no no; not ''ah'', ''are''.

the_syco

Stiffler2 wrote: »

Still - we can't be too far away if Watson can take in a question, search for the answer in the correct context and then get the answer right.

Knowledge versus intelligence.
Knowing facts, versus knowing how to produce answers.

AI without any human input; fairly far away, I'd say.

Pushtrak

Jester252 wrote: »

maybe it does exist and we dont know about it. Maybe it's a poster on boards and is using AH to see if human should live or die

We're doomed.

Skylinehead

Stiffler2 wrote: »

As the title says : How far away our we from A.I.

I know this vid is over a year old but still very impressive.

Proper AI? Still far. At this point, in terms of computing at least, AI is the one that involves the most guesswork.

Stiffler2

the_syco wrote: »

Knowledge versus intelligence.
Knowing facts, versus knowing how to produce answers.

AI without any human input; fairly far away, I'd say.

tbf ( I think ) watson was built and is on his own there is jeopardy ?

he's taking in questions, searching his database and spitting out answers, then picks his next category ??

No human input really except maybe an IT guy making sure his servers doesn't crash

wexie

Stiffler2 wrote: »

tbf ( I think ) watson was built and is on his own there is jeopardy ?

he's taking in questions, searching his database and spitting out answers, then picks his next category ??

The mere fact that he has a database to pick from would imply that there's been human imput, ie. the database.

AI would imply being able to take in facts, surroundings etc and making decisions based on the information available. Not simply answering questions that have a set answer. (which can be looked up in a database which has been created by humans)

FatherLen

very.

Sky King

I'd like to see two watsons head to head

IN THUNDERDOME!

Stiffler2

wexie wrote: »

The mere fact that he has a database to pick from would imply that there's been human imput, ie. the database.

AI would imply being able to take in facts, surroundings etc and making decisions based on the information available. Not simply answering questions that have a set answer. (which can be looked up in a database which has been created by humans)

that's fair enough and a very good answer,

If Watson had legs and could walk around answering these questions and if you didn't know he had a database behind him what would you think of him ??

LostPassword

Answer: probably no more than 4 feet. AI is everywhere in modern devices.

wexie

Stiffler2 wrote: »

that's fair enough and a very good answer,

If Watson had legs and could walk around answering these questions and if you didn't know he had a database behind him what would you think of him ??

Don't get me wrong, it's an incredible achievement, (and even more so if they could make him walk around) but it's not AI.

AI would be the ability to make a decision based on limited information. Not just cross referencing facts. I read an article a long time ago (no link sorry) that basically the difference is that a computer can 'recognise' a table if it's been shown many different pictures of tables and has something to compare it to. A human can recognise a table of a certain shape and design even if he/she has never seen anything similar simply by recognising where it is placed in a room etc etc.

Even very advanced computers can 'recognise' a table based on the fact it has 4 legs and a flat surface. Where they fall down however is for example recognising a 'design' table that has no legs (coffee tables etc) but is still a table.

(sorry if that's a pretty dodgy explanation).

I do agree however that it wouldn't take too much to make a computer like this 'look' like AI.

Duggys Housemate

We don't know how to model the human mind or create self awareness in machines. It's a long way away.

Jester252

Duggys Housemate wrote: »

We don't know how to model the human mind or create self awareness in machines. It's a long way away.

Don't lump me in with that

hooradiation

LostPassword wrote: »

Answer: probably no more than 4 feet. AI is everywhere in modern devices.

Of course there is a massive difference between the concept of AI as imagined by popular culture and things that use tools that have been created for use in academic AI, such as optimised searching.

ruthloss

Holy God, that frightened the life out of me!.

A.I. means artificial insemination down here.

Where To

Two farmers were sitting outside the new AI station in town waiting for it to open.

'Hmmmm Very quiet for a place of it's kind' says one to the other.*


this joke will most probably be lost on those not from a farming or rural background.

mawk

well wolfram alpha is getting pretty good. you should ask it if it is skynet.

steven wolfram actually has a good post on his blog about finding artificial stupidity in computers. worth a google

Guill

About five miles I'd say.

seamus

wexie wrote: »

The mere fact that he has a database to pick from would imply that there's been human imput, ie. the database.

Couldn't you say that about humans though? That we derive our knowledge from other humans?

Left alone, with no contact with other humans, we'd have very little ability to appear intelligent.

So I think it's a little naive to claim that an AI should be intelligent without human input, since we all rely on human input to build our own "database" of knowledge.

It stands to reason that true AI (at least the first generations) will require a substantial amount of human input and interaction in order to function as an intelligence. As AI develops and goes onto further generations, they will learn from other AI rather than from humans, which will vastly increase the learning speed.

I can see the point that this is just a series of algorithms and estimation and decision-making programs running through a database to derive the correct answer, but that's not a million miles away from how our own brains work.

I imagine the declaration of the first AI will actually cause quite a bit of social debate and protest as various groups try to argue that it's not "true" intelligence and shouldn't be compared with human intelligence.

In reality it boils down to determinism -v- non-determinism, but the likes of Derren Brown have illustrated that the human brain is far more predictable and deterministic than we like to think. True non-determinism can be programmed into a machine, so there's no real barriers to AI.

well wolfram alpha is getting pretty good. you should ask it if it is skynet.

steven wolfram actually has a good post on his blog about finding artificial stupidity in computers. worth a google

Lol?

Logical Fallacy

seamus wrote: »

Couldn't you say that about humans though? That we derive our knowledge from other humans?

Left alone, with no contact with other humans, we'd have very little ability to appear intelligent.

So I think it's a little naive to claim that an AI should be intelligent without human input, since we all rely on human input to build our own "database" of knowledge.

It stands to reason that true AI (at least the first generations) will require a substantial amount of human input and interaction in order to function as an intelligence. As AI develops and goes onto further generations, they will learn from other AI rather than from humans, which will vastly increase the learning speed.

I can see the point that this is just a series of algorithms and estimation and decision-making programs running through a database to derive the correct answer, but that's not a million miles away from how our own brains work.

I imagine the declaration of the first AI will actually cause quite a bit of social debate and protest as various groups try to argue that it's not "true" intelligence and shouldn't be compared with human intelligence.

In reality it boils down to determinism -v- non-determinism, but the likes of Derren Brown have illustrated that the human brain is far more predictable and deterministic than we like to think. True non-determinism can be programmed into a machine, so there's no real barriers to AI.

Lol?

While I understand your point seamus I would disagree. Intelligence is not an issue of appearance, it's a factual, measurable characteristic that involves more than the simple reiteration of information from storage.

You can teach a child that a stove is hot, or the child can burn its hand and learn itself. Two different routes to the same result and knowledge. With an A.I, you need to set parameters for what a human would consider hot and cold. Even then it has no concept of the actuality of either because the concept of pain due to temperature is an abstract when you can't feel pain yourself.

Conceptually, true A.I is fascinating and open for debate I reckon.

wexie

seamus wrote: »

Couldn't you say that about humans though? That we derive our knowledge from other humans?

Left alone, with no contact with other humans, we'd have very little ability to appear intelligent.

So I think it's a little naive to claim that an AI should be intelligent without human input, since we all rely on human input to build our own "database" of knowledge.

That's a very valid point, I think I phrased my point very poorly. What I meant to say (to stick with the table analogy) that a human, when shown a number of tables that look like this or this 'knows' what a table looks like and will likely 'recognise' a table that looks like this , this or this whereas a computer would likely not have the same cognitive powers despite the fact that it could have a very large database of tables.

It's being able to make this 'leap' of recognition I was trying to point out. I'm sure there's a proper term for it.

Of course humans will always have a large hand in defining, programming, teaching AI. When the day comes that's no longer necessary I think we'll be in real trouble. If ever there was a race that needs saving from itself it's us (as so many sci-fi movies have already pointed out)

Capt'n Midnight

It's 20 years away.

Because that's what they've been telling consistently us since the 50's


The big problem is not the technology it's understanding how humans work.

Logical Fallacy

Capt'n Midnight wrote: »

It's 20 years away.

Because that's what they've been telling consistently us since the 50's


The big problem is not the technology it's understanding how humans work.

But the last part begs the question...why try and mimic humans?

wexie

Logical Fallacy wrote: »

But the last part begs the question...why try and mimic humans?

Hahaha.....because that is our only known form of 'intelligence' so it's the only thing we have to go on? (I know I know)

I think I see your point though, but humans trying to create AI based on intelligence other than the human brain is an interesting philosophical conundrum though.....one which calls for a few pints and then some.

Duggys Housemate

we'll probably try mimic rational parts of the human brain.