Why is AI such an impossible goal?

[768]

The facebook AI was just modified because it was supposed to be able to communicate with humans and it was talking, more or less, garbage. The reporting of it was just as bad.

[Stan the Bat]

Is Ray Kurzweil's theory about the timescale of "the Singularity" correct ?

[AshVX220]

768 said:

The facebook AI was just modified because it was supposed to be able to communicate with humans and it was talking, more or less, garbage. The reporting of it was just as bad.

Yes, that's what I read, but I thought that they'd assessed that the two AI's, although talking garbage, could understand each other. Or I mis-read, mis-remembered.

[768]

AshVX220 said:

I thought that they'd assessed that the two AI's ... could understand each other

Apart from being way too anthropomorphised, I don't think they even looked at that really, it wasn't English so it didn't meet the bar and they moved on.

[otolith]

Description of the negotiation bot experiment.

https://towardsdatascience.com/the-truth-behind-fa...

[ruggedscotty]

The genie is in the bottle with this one, everyone is fearful of what will happen if that genie is released, it may well work in our favour but the reality is that it very likely wont. Can you imagine something that gains self awareness and then realises that we are a danger to it, that we are destroying earth and that we have wars and idolise sky fairies…. it will quickly realise that it has to control us and it will peacefully or not.

An intelligence that is superior to us in every way, a force to be reckoned with and to be very very wary of. because we created it, it doesn't mean to say it will be friendly or without threat.

Once that is out the bottle it will never go back. And we as a race will be changed for ever.

[Mothersruin]

I don't think there's any middle ground.

It'll either be like The Culture where we're almost seen as pets while it does it's own thing, or The Terminator.

I can't see any security ring fencing of external 'escape' in the long term - it'll simply be way smarter and thinking in a totally different way that we won't be able to counter.

I may die from it but I'd be very happy/interested to see what happens, if it happens within my lifetime.

[AshVX220]

Indeed.

For anyone new to the thread I recommend this;

https://waitbutwhy.com/2015/01/artificial-intellig...

https://waitbutwhy.com/2015/01/artificial-intellig...

There will be absolutely no middle ground, it'll either be the best thing to ever happen, or it will be humanity's downfall.

[andy_s]

IIRC the States still has the top ten AI researchers and 60% of the top hundred and still dominates the field, but this may not last much longer; China is blessed with many advantages and motivations so is catching up rapidly.

Xi Jinping [with a long term view now he is President for life thus offering consistency and generational goal vision/execution) has corralled the top three Chinese equivalents of Amazon, Google and Facebook into the funding and operation of an AI centre with the roadmap looking like: 2020 - equivalency with the US, 2025 - superior to US and 2030 - dominance. The resources are staggering.
They also have the advantage of not being tied up with too many troublesome regulations or ethical concerns [CRISPR...].

If we're enquiring about the future of it, we're probably asking in the wrong language...

[Terminator X]

Yet in 2019 my sat nav can't even route me around traffic

TX.

[andy_s]

Robert Miles has done a few good videos on AI safety, nicely explaining the main problems in some detail; his latest is a nice counter to the usual responses about safety concerns for anyone interested: https://www.youtube.com/watch?v=9i1WlcCudpU

[Baron Greenback]

https://youtu.be/J63mKverb8w?t=1

The Royal Institution good 1hr presentation on Will Computers Ever Think Like Human Beings? by Vint Cerf.