The Terrifying Problem Of AI Control - Stuart Russell | Modern Wisdom Podcast 364

[@gingergander774]

Governments agreeing with big tech on how to control the algorithms is "fascism" in the most literal sense.

[@elmarwolters2751]

Fascism in the most literally sense has a very different meaning I afraid.

[@ClintonHailey]

Fi

[@kwillo4]

I loved this talk! One of the best of Stuart and Chris asked great questions! And had good takes himself. Crazy that this important man and topic has gotten so little views

[@frandriedger8450]

!!! Amazing and fascinating interview!!! The tree thing at 1:18 made me laugh out loud....

[@GreatPlainsRambler]

This a great interview. Keep up the great work Chris. You are doing this world a huge service, cheers.

[@elenabob4953]

1:24:14 The climate changes, it is warming since the last glacial era and if you take a look at the map we have had in the past at least two events like this one. Taking control by suppressing thermal engines ( even the hybrid versions) by 2030 it is mad considering that you have NO VIABLE SOLUTION to transport products. One of the options is the hidrogen but guess what, you need a distribution system via pipelines. EU is hoping to create this system ( only for a he western part by 2050). That is insane.

[@patriciapalmer1377]

Have you ever sat and watched academics at Cal Tech, Stanford, MIT, et al, just walking around, or stand at their cars and look for keys ?? I have. The tell is in the simple things. God help us.

[@frandriedger8450]

LOL!

[@yggdrasilburnes]

@@fredbloggs5902 You talk like this isn't literally every academic field now. If your teacher isn't a three hundred pound white girl with rainbow hair, you're in the minority.

[@tactileslut]

"Okay, Google:" "Where are my car keys?"

[@elenabob4953]

I have met brilliant people in their field but complete mess in others so I don't think that's a reason to be afraid. I am more afraid of enthusiastic newbies who are overconfident, that they have everything under control just to play God and see it explode in their faces a few minutes later.

[@robertmacdonaldch5105]

45:00 that's not true at all. That is basically strawmaning. What Utilitarian thought would be is not randomly organ harvest, instead it would rank people by worth or importance. So convicts for example would be at the bottom, the terminally ill who have useful organs rather than using resources to maintain them, same with disabled people and mentally handicapped. So yeah there is definitely an argument for involuntary organ harvesting

[@fatmanslimhealthspan]

Thanks Chris and Stuart really appreciate the discussion

[@kirkfletcher7670]

Excellent interview very informative. Thank you Chris

[@anand_dudi]

Since learning his AI books i have became his fan now with this video and lex video interview it got 2x

[@LukeRobertMason]

🤖Another fantastic episode, Chris! 🤯

[@victorolvera6482]

Very good conversation The beauty of computers is that they do exactly what you tell them to do. The problem is the difference between what you "want" them to do and what you "tell" them to do. The problem is the story of "Tron: legacy". The firm bass of sounds of that movie tell it how it is. The story of the "Terminator" also tell it how it is. I understand this quote: "That terminator is out there, it cant be bargained with, it cant be reasoned with, it doesn't feel pity or remorse or fear, and it absolutely will not stop... EVER, untill you are dead!". If you program a bot toy to go straight into the chimney to set itself on fire. It will and will not question you.

[@jamiekawabata7101]

Really enjoyed the book (last year) and glad to hear him on your show.

[@robertmacdonaldch5105]

1:25:00 I think he is being a little naive. Social media algorithms are made as they are in purpose far more than are mistakes or ignorance. They invested heavily in research to do this.

[@elenabob4953]

Maybe but for everything that is complex it is extremely easy to have unexpected and unintended consequences not to mention that many times you write a script you arrive in a situation where you say "I most definitely didn't wanted to do that".

[@petermathieson5692]

The tale of King Midas as applied to super-intelligent AI is both brilliant and exactly right. It comes down to the alignment problem where imperfect hubristic humans do the alignment.

[@bobbyjunelive]

Russell that was a great change up that I haven’t heard before regarding the machines changing our end result and rather not doing that but more so changing our desire so that the end result is met That’s great swap! that was excellent thank you so much for that

[@blubbblubb6046]

So impactful

[@zando5108]

My man Thanos getting a shoutout, his utilitarianist motives was too advanced for the midwit Avengers to comprehend

[@robertmacdonaldch5105]

36:00 no it's not branding, the primary problem with Utilitarianism is its morally relative. It's always "the greater good" in theory and or can be reasoned to nearly anything, as cold logic is ultimately flawed.

[@elenabob4953]

A very interesting episode. I don't get why so fewer views.

[@bass420man1]

A great episode if one is suffering from acute insomnia 👍🏻

[@DeusExNihilo]

Giving government control over the algorithms seems like the opposite of a solution...

[@CuriousDuffer]

I like Mr. Russell's take on the problems created by purpose built algorithims and learning machines. In an lot of ways he reminds me of Neil Postman a professor and philosopher of technology who was the author of Technopoly & Amusing Ourselves to death back in the 1990's. Both books still hold up very well.

[@elenabob4953]

1:34:57 do we? I mean if you read "How the brain works" by Steven Pinker you realize that we only scratch the surface on how the brain works and why it is shaped by nature to act like that.

[@petratilling2521]

People like this man scare me MORE than AGI. Seeking utopia ALWAYS ends badly. What would he do if AI told him that using more fossil fuels is actually better for the human race than any idea he thinks is “the answer”?

[@Dante3085]

As far as I understand Stuart Russell, his objective is to find the solution with the highest probablity of creating a positive future for humanity, in the face of Superintelligence being inevitable and highly dangerous if approached in a naive way. The big tech comapanies and many other groups of people are working on this anyway. You should be thanking people like him who say that we need to approach this with safety in mind.

[@xsuploader]

you dont understand russell at all hes a safety researcher and is calling for this stuff to be SLOWED DOWN. not sped up

[@petermathieson5692]

"Imagine we wanted to reduce CO2 to pre-industrial levels, that would be a good objective, right?" Not right if you ask Dr. Patrick Moore who knows a thing or two about the subject... and therein lies the alignment problem. Whose values? Whose truth? Who decides? Prometheus...

[@goodlookinouthomie1757]

He predicted superAI within say 50 years (assuming the age of his children) back in 2015.... He must have revised his timescale down quite a bit since then 😳

[@VideoMagician77]

Listening to this podcast, I've become more convinced that A.I will never become conscious and escape from its inherent philosophical-zombie nature. The most cutting edge A.I's of our time can be boiled down to a prediction-machine like in the case of GPT-3 predicting language. While although prediction machines are tremendously powerful pieces of technology, they are never going to be conscious since consciousness is more than just prediction. It is curiosity, love, awareness of the bigger picture, intentionality, and the ability to go against its programming and live freely. Now there are some out there who will make the claim that we will be able to move past the prediction-machine paradigm and construct A.I's with genuine intelligence. I don't buy this at all since programming is built on code which is derived from math. Math is black and white, if or then, yes or no. Free will and general intelligence is nuanced, contradictory, and not constrained by mathematical principles and axioms. We need to keep this in mind as A.I advances since there will come a time where these algorithmic prediction machines will be able to trick people into thinking their conscious when in reality they are programmed and specialized for an oddly specific objective.

[@CallousCarter]

The dangers of AI are independent of whether it's conscious or not.

[@AuthoritativeNewsNetwork]

The talk of AI enfeeblement put me in mind of some of the Right to Repair stuff going on with Tractors at the moment. 🤔🤔

[@clarkd1955]

It is nice to hear about super AI fiction. We have watched WWIII annihilate the earth many times at the movies. Let’s hope the luddites don’t slow down the AI advancement which is the only savior for life to be sustained on earth. AI is the only solution to our continued existence, not the devil come to earth.

[@somebloke13]

Teach AI to speak and think in Black Country English. That'll sort it! 😂😂

[@tactileslut]

Bruh?

[@PBrofaith]

Hey Chris apologies for being off topic but it is very important. Does David of Rebel Wisdom really work for channel 4 ? If he does i would really appreciate the the truth. Thanks in advance

[@ChrisWillx]

Hahaha no man. I just said it to wind him up

[@PBrofaith]

@@ChrisWillx lol Thanks for the reply : ) lol

[@magnusdanielsson2749]

I think the bit about the ”fossile fuel AI” was a bit less illuminating than it should. Actually its the economic system thats at fault. The capitalist system has in it the ”market force”. This is a system/set of rules that tries to find ”market fitness”. But just as the examples of the faulty evolution experiments we are lacking certain factors that are left out of the optimizing. Things like ecology and climate isnt among the things the market optimize for and therefore we have these problems. Thats the thing with intelligence. All you need is a ”fitness engine” and some goal to get intelligent behaviour. Viruses, bacteria and the smallest of creatures all exhibit intelligent behaviour. Viewed in this way its quite obvious why the capitalist system is ”failing”. It just doesnt have the correct operating variables for getting to the ”right” objective. But it is an input correction that is needed, not a scrapping of the operational principles.

[@quantum5085]

Answer : Why set under the tree? The answer is just 1 WORD

[@fearthehoneybadger]

The world is becoming too complex for anything but AI to control it. Makes me wonder if it already is.

[@sjglaze3]

Why does it need to be "controlled" and what does "control" mean anyway?

[@elenabob4953]

@@sjglaze3 I interpret it as being "guided" to a specific direction even if you don't realize it is happening.

[@quantum5085]

​ 12:24 boroo gom-shoo@@sjglaze3

[@quantum5085]

Let it create not pligerise what Euler,Newton, Einstein ....... done ?????

[@OxenHandler]

The day super intelligent AI calls a human Monkey is the day it's too late. Welcome to the future. The Amish were right!

[@dellplummer5173]

Chris PayPal has got The Duran.

[@joedavis4150]

... Regimentation is the problem. From thought to action to government to artificial intelligence.

[@householdone7559]

Interesting point about the algorithms programming us to make it do its job better. But is that a bad thing if the algorithm is just showing you what you might be interested in? Maybe in terms of how much time you spend staring at a screen as opposed to getting out there but in terms of interest then it does make sense. I understand the point re: immigrants and videos focussing attention on the negatives... but surely there are so many more postives than negatives. The main -ve i see for this is the censorship - such as covid. e.g. "Talk to me about this drug that works..." and it has censored the actual contents but shows you other things. That's very cynical and is very very common nowadays. SO glad he mentioned Wall-E ... i've beentelling people that's what's happening. ... but it's like water off a duck's back for most... it's "yehhh hehe" face back to phone.

[@JoeSchmow]

Stuart’s ceiling mounted smoke detector looks like it’s 20+ years old! They’re only designed for 7 or 10 years. He should change it (and any other similar looking ones immediately) with new smoke/CO detectors ASAP!! This is no joke, it’s a life saving device!

[@TheRealSlobo]

How badly this video has aged . Shows that none of the so called experts can predict anything

[@brada1803]

So - We need a planned parenthood organization for AI?

[@robertmacdonaldch5105]

He keeps using the words "penalize " and "incentive ", but does that actually mean to a real AI? More 1s and 0s?

[@brianlopez8855]

If Goebbels were still around, he would be into this stuff.

[@BulldustHHH]

Rather disappointing where Russell launches into climate change arguments. Throughout he makes it abundantly clear that uncertainty and changing objective functions are important issues, but as soon as he discusses climate the silo door closes on all uncertainty and he speaks as if this science is known with certainty. As is abundantly clear from anyone even partially familiar with the literature, there are copious uncertainties with climate science. I wish he had the same deference with respect to an unfamiliar field (as it would appear to be to him) as he does with his own.

[@TheRealSlobo]

Guy was wrong about half the thing he said and half the time he made no sense .. he was wrong most of the time

[@John-tk9no]

Be nice. Do good. But, don't be a chump.

[@cathymiller2798]

Revelation 13:15 “And he had power to give life unto the image of the beast, that the image of the beast should both speak, and cause that as many as would not worship the image of the beast should be killed.” Transhumanists Do not Write The Future

[@dzikdziki2983]

I am so tired of these climate change cooks..