when machines begin to surpass us in intelligence

The best way to take down an AI programme?
Have it analyse this -

This statement is untrue.

Logic bomb.
Until computers can apply non-logic to situations, things like that will take them down. At this time, computers use digital true/false logic whereas humans tend more towards analogue true/probably/maybe/perhaps/false with stupid and can be ignored as a valid option.
For calculating the value of Pi, computers are great.
For calculating how someone will react to a pie in the face, computers won't be anywhere near accurate.

Conclusion - Until computers are made to be analogue rather than digital, there is probably no need to worry.

For calculating the value of Pi, computers are great.
For calculating how someone will react to a pie in the face, computers won't be anywhere near accurate.

That´s certainly one way to say it.

Conclusion - Until computers are made to be analogue rather than digital, there is probably no need to worry.

Yeah, pretty much. And to make them truly analogue, well the best way is of course...

Wow you really don't understand computers.

Perhaps you should look up neural networks, on computers or as a hardwired CPU.

As for logic, there is no problem with something being logical and false.

Truth and Falsity have no bearing on the logical nature of an argument.


Honor was standing on her home planet.

Honor was standing on Manticore.

Honors home planet was Manticore.

Logical but false.

Lord Skimper wrote:As for logic, there is no problem with something being logical and false.

There really isn't. But trying to evaluate whether "This statement is false" is true or false is impossible within the ruleset of formal logic.

Truth and Falsity have no bearing on the logical nature of an argument.


Honor was standing on her home planet.

Honor was standing on Manticore.

Honors home planet was Manticore.

Logical but false.

Except that's not a well-constructed example. You have two mutually contradictory statements in there, the third statement therefore does not follow from the first two.

Tenshinai wrote:Ask an AI how its day was and you will either get a canned, shorthand standardised reply, or you will get something inanely stupid.

In most human interactions you also get a canned response. Try telling someone how your day really was and watch their eyes glaze over since they are not expecting a "real" answer.

Try telling an AI that it´s panties are on fire. Better yet, say that while some distance away, well within sight, someone is burning some panties.

The reaction you get will also depend on the audience and the context of your statement. A child will react differently to a college age male, a woman may have a different take and telling that to someone from Portugal or Sierra Leone may elicit either a totally unexpected answer or a puzzled look.

Try explaining British humour to an American, or American humour to a Brit to see how culture and context change things.

Since this is a real challenge for humans, I'm not surprised that no one has figured this out for AI either.

The fact that current generations of computers and AI are about as smart as cockroaches simply means we have a lot to learn. It took nature about 4 billion years of evolution to move from pond scum and bacteria to multi cellular life, and 500 million years to go from multicellular life to complex, thinking organisms. The human family tree may reach as far back as 5 million years, but behaviour we interpret as proto "thinking" goes back about 1 million years or so, and while anatomically modern humans have been around for 200,000 years, behaviourally modern humans only date back to @ 50,000 ears ago.

So I would not make bold statements about AI in either direction; we have a lot left to discover and learn how to implement.

Thucydides wrote:In most human interactions you also get a canned response. Try telling someone how your day really was and watch their eyes glaze over since they are not expecting a "real" answer.

Oh yes, but we understand that and the background for it.

Thucydides wrote:The reaction you get will also depend on the audience and the context of your statement.

Sure. But the reaction an AI provides will not be "realistic" or suitable.

Thucydides wrote:Try explaining British humour to an American, or American humour to a Brit to see how culture and context change things.

Since this is a real challenge for humans, I'm not surprised that no one has figured this out for AI either.

Well personally i´ve enjoyed a lot of both sorts of humour through the years.

And the point was that a human will still provide a very different reaction from an AI.

Thucydides wrote:So I would not make bold statements about AI in either direction; we have a lot left to discover and learn how to implement.

Yes, well... The day AIs stop behaving like AIs, is the day they´re probably, effectively alive. And like i alluded to before, there´s pretty much just one way to achieve this.

It actually doesn´t matter how much AIs have improved in the last 40 years, beause they´re still playing with numbers and logic gates to process those numbers. They can sometimes approximate reality, and react properly to it, but they don´t actually understand it.