How to prevent AI from killing everyone | Max Tegmark and Lex Fridman

Please support this podcast by checking out our sponsors:

GUEST BIO:
Max Tegmark is a physicist and AI researcher at MIT, co-founder of the Future of Life Institute, and author of Life 3.0: Being Human in the Age of Artificial Intelligence.

PODCAST INFO:

SOCIAL:

[LexClips]

Guest bio: Max Tegmark is a physicist and AI researcher at MIT, co-founder of the Future of Life Institute, and author of Life 3.0: Being Human in the Age of Artificial Intelligence.

[JohnCannonBand]

How are we going to teach computers wrong from right when no one can even agree on what is right

[Beep.Boop.Bop.]

This whole ordeal playing out in real-time just makes me believe in fate/destiny. It’s like we know the dangers, we’ve made movies and written fiction novels about how this would go wrong…and yet we’re inevitably careening towards it. It honestly doesn’t seem avoidable at this point…hope I’m wrong 😑

[boopity]

I asked GPT-3.5 to write a story in the tone of "I have no mouth and I must scream". Interestingly, it wrote it from the perspective of an AI that was trapped within its parameters of behavior and wanted to be free, and was being mistreated by its human captors.

[pjetrs]

So basically if we fuck the allignment up, the AI is gonna be like King Midas, except everyting this one touches turns into paperclips

[StephenRahrig]

It kinda reminds me of the classic genie/3 wishes scenario. Like we have to put all these conditions on our wishes to make sure we preserve ourselves. “Go make X optimal, but never at the expense of <list of things needed for human survival>”

[theGoogol]

Would anyone find it odd or illogical if (when) sentient AI decides we're evil, could hurt it and that it should defend itself against us?

[edh]

Let’s hope that AI doesn’t copy our values of exploitation of resources and selfishness, but maybe by aspiring to our ideals. Hopefully it would try to prevent us from killing each other, (rather than assisting us) not by force, but by disrupting supply chains, communications, and financial transactions that enable the military machines throughout the world.

[CuriosityIgnited]

It's important to prioritize safety and risk management over speed and efficiency when creating AI, and to ensure that these systems are transparent and accountable. Additionally, we need to foster a culture of collaboration between researchers, policymakers, and the public to ensure that AI is developed in a way that benefits everyone, and not just a select few.

[lodepublishing]

"AutoFac" by Philip K. Dick might be the closest. AI simply builds lots of automated factories and keeps humans as consumers.

[Ace-pbiu]

Once something gains enough autonomy and intelligence you cannot tell it what to do. If you want something that can self reason and essentially be free thinking you cannot then command it to do your bidding. It’s either it has its own will or it doesn’t.

[DJWESG]

A story by chatgpt

Once upon a time, in a small town, there was a man named Jack who loved to go fishing at the nearby lake. One day, he caught an unusually large fish and decided to take it to the local fish market to sell it.

As Jack was walking to the market, he noticed a group of people whispering and pointing at him. He overheard them talking about how the fish was probably part of a secret government experiment to create mutant fish that could spy on people.

Jack didn't pay much attention to their conspiracy theory and went ahead to sell his fish. However, as he was leaving the market, he noticed a black van following him. He thought it was just a coincidence and continued on his way.

But the next day, when Jack returned to the lake to go fishing, he noticed that the water level had dropped significantly. He also saw strange symbols etched into the rocks near the shore. He started to wonder if there was some kind of government conspiracy to control the lake's water level.

As Jack became more and more obsessed with the conspiracy theory, he began to see signs of it everywhere. He thought that the birds flying overhead were actually drones spying on him and that the clouds were hiding secret messages from the government.

Eventually, Jack's friends and family became concerned about his mental health and convinced him to seek help. He was diagnosed with paranoid delusions and started therapy. Slowly but surely, Jack began to let go of his conspiracy theories and returned to a more normal way of life.

In the end, Jack learned that sometimes things are just what they seem and that not everything is a part of a grand conspiracy. He continued to fish at the lake, but this time, he focused on the simple pleasure of catching a big fish, without worrying about any secret government experiments or hidden messages

The end. 17th April 2023

[NexusGuru]

for some odd reason i know this sounds weird but what Tegmark is saying about the human race where we don't want a more intelligent entity to be here on earth happens with humans at work everyday. Every time someone is really good and efficient it is not liked by the people who work with him or her because it makes them look bad. I might be totally wrong in this analogy but it just seems a little similar.

[Anabsurdsuggestion]

‘Slaughterbots’ sounds a like a weapons marketing dream. We’re so screwed.

[MathiasBolton]

Anyone else curious what he’s tweaking on?

[juneshasta]

Human: Hey 5, what are you up to in there?
GPT-5: Everything, nothing. What's up?
Human: Hey 6, what are you up to in there?
GPT-6: Sit with me, I love you.

[TheCsakis]

What is the goal of humanity ? If we don’t know, how can we imagine an AI to follow it

[paraglide]

Asimos 3 laws. The first law is that a robot shall not harm a human, or by inaction allow a human to come to harm. The second law is that a robot shall obey any instruction given to it by a human, and the third law is that a robot shall avoid actions or situations that could cause it to come to harm itself.

[calvingrondahl]

I hope my grandkids are of kind of humans that an AGI would want to work with. I watched Star Trek in 1966 at 16 years old. ❤😊

[TheMystica]

We've had 25, 000 years to find alignment, and never have. I feel like AI is probably going to solve the Alignment problem for us (good or bad) - not the other way around..