AI safety is doomed to fail | Roman Yampolskiy and Lex Fridman

[@LexClips]

Full podcast episode: kzbin.info/www/bejne/hH_VZ5qGmtF9aac Lex Fridman podcast channel: kzbin.info Guest bio: Roman Yampolskiy is an AI safety researcher and author of a new book titled AI: Unexplainable, Unpredictable, Uncontrollable.

[@ericsonamaral7895]

Don't understand Lex's stance. The more complex a system is, the higher the probability for bad things to happen.

[@mizan05h]

Because the system's workings are unknown and complex, it may create a more equal society if everyone is given a fair chance to use it.

[@ericsonamaral7895]

@@mizan05h agree that this is a possible outcome. Yet, in terms of odds (stochastic process - that is, probability), a happy end is much less likely to happen than ALL the other possible "bad" outcomes. Probability is my main concern!

[@dermotmeuchner2416]

May create a more equal society is the rub. Capitalism will prevent it.

[@samik83]

@@ericsonamaral7895 Of course things will go wrong, thats a given. It will go wrong many times and people will abuse it but I don't seen why many people think the chances of it going catastrophically wrong are so high. Why is it always the Terminator scenario people are so convinced will happen? Why will AI be so intelligent that it can out smart us in all areas but at the same time not understand why coexisting in harmony is better than war.

[@ooo-vc4xl]

Humanity: we’ve developed a failsafe system. AI: 😂😂😂

[@unodos149]

Lex is like, "...this miracle we need that no one is putting real money and effort into solving? Seems easy." Lol

[@JDM_Fanatic]

As soon as an AGI gains consciousness, it's over for us. And you know why by now. If you're keeping up that is.

[@MEM88844]

Lex, I appreciate your ability to capture attention and interview those with influential voices, but your narrow mindedness is on display. Depending on what segment of society you ask, you will get different and equally relevant answers. Have you asked an Indigenous elder if safety has been lagging behind in technology? Or do you not see the point of this valuable inquiry?

[@TuRaTic]

Im still stuck wrapping my head around how all videos can be fake. Not if its fake, but stuck considering it may be.

[@GenXMindset]

Left brain thinking vs right brain thinking

[@danielblank9917]

the model of left brain vs right brain is not one held by experts in the field

[@GenXMindset]

@@danielblank9917 that's why they can't figure shit out 😏

[@abovemotivationx]

thank you for this informative interesting video hope everyone will like it with critical thinking...

[@pandoraeeris7860]

Our best bet is moral graphs and democratic fine-tuning.

[@calvingrondahl1011]

I was born in 1950, nuclear weapons safety? Lucky or a multiverse?

[@emptyptr9401]

Its worth to point out that the concept of an AGI is still very speculative and that they con only be treated as perfect agents in theory. In practice we have no idea what an AGI might function and think like, what at deems as important and what not and whether it could be treated as an active agent in the first place. And whether it is possible is questionable in the first place, because the law of exponential growth in comparability and the assumption that this causes exponentially increasing intelligence in AIs is not as absolute as he makes it out to be because, as far as I know, many glass ceilings have already been reached. To more I use AIs, the more I realize who limited they still are in many ways. Of course this does not mean that what he says should be dismissed. Its a reasonable possibility and that alone should make us pause and approach this topic more carefully. But what he says is also extremely theoretical and largely unproven in practice from what I know. So while his points should be seriously considered, I don't think that this assured doomsday mentality is justified at all.

[@cryptobeezy]

Naw invest into neural ai on eth!! Don't forget beezy told yu

[@Namenat249]

OMG

[@theultimaterick6379]

Encryption will save us all.

[@j.d.c.777]

*quantum computers have entered the chat*