3 ms·

but thats the same as in Chess, Go or even StarCraft: the AI is absolutly impossible to beat for human players unless it has explicit flaws built-in. The gener

but thats the same as in Chess, Go or even StarCraft: the AI is absolutly impossible to beat for human players unless it has explicit flaws built-in. The general point I want to make is that it doesn't make sense trying to compare humans with AI (and somewhat soon) robots, because the will outperform everyone. Handycapping them is not a solution either. The "amazing versatility" of humans also will not persist for too long anymore I'd say, therefore the only way to have a real and fair comparison/pit is against other humans in the future.

necovek 3y ago
Well, let's agree to disagree about versatility: I don't see non-specialized AI outperforming humans in a number of years (a human will easily beat chess AI at Go). While I am somewhat impressed by the conversational acumen of recent LLMs, the fact they can produce outright garbage tells me we are as far away from synthesising multiple types of intelligences as ever: humans simply need far smaller input data set to start recognizing patterns and rules (as witnessed by kids learning to speak). I am not yet convinced we'll live to see something like actual self-driving cars with as much capability as an average human driver: I believe augmented environments are needed (communicating roads, signals and cars) to get to self-driving cars. The one thing they've got going for them is consistent focus, whereas practical ability in humans significantly depends on the mood, tiredness, level of multitasking...
robotresearcher 3y ago
> the will outperform everyone I'll try to make the same point land a bit harder: Robots will not compete any time 'somewhat soon' in many/most physical tasks, because their energy storage capability, motor efficiency, strength to weight ratio, and many other mechanical and sensing properties are not good at all compared with humans and other animals, and can't be fixed by software. Everyday intuition underestimates how large this gap is. Brains are only part of the requirement for competence in the real world.

necovek 3y ago

Well, let's agree to disagree about versatility: I don't see non-specialized AI outperforming humans in a number of years (a human will easily beat chess AI at Go). While I am somewhat impressed by the conversational acumen of recent LLMs, the fact they can produce outright garbage tells me we are as far away from synthesising multiple types of intelligences as ever: humans simply need far smaller input data set to start recognizing patterns and rules (as witnessed by kids learning to speak). I am not yet convinced we'll live to see something like actual self-driving cars with as much capability as an average human driver: I believe augmented environments are needed (communicating roads, signals and cars) to get to self-driving cars. The one thing they've got going for them is consistent focus, whereas practical ability in humans significantly depends on the mood, tiredness, level of multitasking...

robotresearcher 3y ago

> the will outperform everyone I'll try to make the same point land a bit harder: Robots will not compete any time 'somewhat soon' in many/most physical tasks, because their energy storage capability, motor efficiency, strength to weight ratio, and many other mechanical and sensing properties are not good at all compared with humans and other animals, and can't be fixed by software. Everyday intuition underestimates how large this gap is. Brains are only part of the requirement for competence in the real world.