How Long Before Robots Can Think Like Us?

The Telegraph:

This summer, a computer almost passed the Turing intelligence test. Dan Falk examines the narrowing gap between humans and machines.

Read the whole story at The Telegraph

From our partners

Comments

[FrostyDog]

01:45 AM on 08/23/2012

Perhaps humans started out as somebody's robots and are just now getting self-intelligent . . .

[Stoopid American]

11:53 PM on 08/21/2012

"How long before robots think like us?"

After three decades writing software, I am thoroughly convinced it will never happen, unless someone can invent a computer that can harness indeterminism. Maybe quantum computing will make a difference. With deterministic computing? Nope, ain't gonna happen.

[rtgmath]

08:14 PM on 08/22/2012

Deterministic computing can be used, although another kind of algorithm will have to be utilized. Instead of true/false we would have true/false/maybe scenarios.

Many programs require a series of criteria to be met in order to initiate a particular action. That is true/false. All the criteria are met? Then initiate. Not all, don't. The maybe option recognizes that some criteria may be met and not others, and make a determination -- perhaps with a random algorithm -- whether to initiate an action. Post-decision analysis can log whether the decision was correct or not. That would go into a database linked to the criteria. Wrong decisions to act based on limited information (on say, one of the criteria) would add weight against acting, but right decisions based on limited information would add weight for acting.

This kind of heuristic provides learning -- decision-response-review. Other heuristics can be written for stimulus-response, essentially allowing the robot to write its own internal rules of behavior.

Deterministic computing can provide the base, while learned decisions encoded in databases (memory) and randomized responses handle overt behaviors.

[Stoopid American]

10:40 PM on 08/22/2012

To me, thinking is this incredibly mysterious process that involves formal logic but also what I can only conceive of as 'harnessed chaos'. It's indeterministic, but in an ordered way, if that makes any sense. For given inputs, you usually get similar outputs, but rarely identical outputs. It's very hard to define in rigorous terms.

Fuzzy logic is probably in the right direction, but probabilities are too simplistic and randomness is too ... random. :-) I really think that it will be impossible to map such an order-from-chaos kind process like intelligent thought to something that inherently thinks only in 1's and 0's. Assuming we ever understand thought well enough to model it in the first place, of course. This is why I'm wondering if something truly different from binary computers, like quantum computing, might make the difference.

I agree with you that we can create emulations of thought that are increasingly realistic. Maybe we can even create an emulation so good it passes the Turing test. But it is still only an emulation, not true thought.

Anyway, it's all good food for thought ... :-)

[FlipFlopMitt]

11:38 PM on 08/21/2012

Is that Mitt Romney?

[FrostyDog]

01:47 AM on 08/23/2012

The Romney "Cha Cha" is starting to catch on:

First you make a statement, Cha, Cha, Cha

Then you walk in back, Cha, Cha, Cha

Make a statement, Cha, Cha, Cha

Then walk in back, Cha, Cha, Cha

Kind of catchy, but never goes anywhere. Perfect for those who favor a candidate that dances around the issues . . .

[Enea]

09:55 PM on 08/21/2012

We have already become robots, programmed from morning till night.

[TaiJi2]

01:45 PM on 08/21/2012

"How long before robots can think like us?"

"We" think? Who knew!

[mysaltydog]

10:35 AM on 08/21/2012

We already have a robot, named Willard, who is thinking and speaking like a TeaPerson.