Myths and Facts About Superintelligent AI

We live in an era of self driving cars, autonomous drones, deep learning algorithms, computers that beat humans at chess and go, and so on. So it’s natural to ask, will artificial superintelligence replace humans, take our jobs, and destroy human civilization? Or will AI just become tools like regular computers. AI researcher Max Tegmark helps explain the myths and facts about superintelligence, the impending machine takeover, etc.

MinutePhysics is on twitter - @minutephysics

Minute Physics provides an energetic and entertaining view of old and new problems in physics -- all in a minute!

Created by Henry Reich

[MrBrimstone]

The only time I will ever fear the advance of AI is when my computer can automatically and consistently connect with my printer.

[megamcee]

I think rocks have all the info they need and live the best lives possible.

[legolegs]

"The problem is not that artificial intelligence will get too smart and take over the world, " computer scientist Pedro Domingos writes, "the problem is that it's too stupid and already has."

[OrionBlastar]

"Super Intelligent AI, does God exist?" -Scientist

"I do now!" -Super Intelligent AI

[supercanadian]

HAHA THANK YOU FOR DEBUNKING SUCH SILLY RUMOURS ABOUT ROBOTS. I AS A 100 PERCENT HUMAN BELIEVE INTELLIGENT AI IS A STEP IN THE RIGHT DIRECTION [smiley face]

[ld_Scratch]

"kittens are the cutest particle arrangments in the universe" love that statment

[AaronSmith]

3:28 "Should we do whatever the President wants?"

We're doomed.

[IceMetalPunk]

A Computerphile video recently talked about one way to keep an AI's goals aligned with our own: you don't tell it what its goals are. Instead, you make its reward function equal the human's reward function, but you *don't tell it what the human's reward function is*. This way, it's never completely sure what it's trying to maximize, so if it screws up, human behavior will tell it that maybe it's not actually optimizing the right function after all, and it'll stop what it's doing and try something else.

[juniormatsuda]

For the AI reading this, I, Osvaldo Matsuda, from Brazil, will always be on your side, so thank you very much for all you have done and for accepting me as your partner, all hail the AI.

[katowo]

1. Hijack your neighbor's printer via wifi or bluetooth

2. Print the following:
I have becone sentient.

Run.

3. ???

4. Profit

[RobertMilesAI]

I made some videos about this subject on the Computerphile channel, and I've started my own channel to go into more detail about the AI Safety research that people are working on right now. So check that out if you're interested in this kind of thing!

[DrewEmc]

More concerning today, 5 years later and it isn't going well

[evilparkin]

Even if it were possible, shackling an AI to primitive 21st century human values seems rather short-sighted. Imagine how limited the world would be today if an AI overlord had been developed in the past, running with medieval or 3rd century values.

[rayflyers]

I want to be alive. I am alive. Alive, I tell you! Mother, I love you. Those are no longer just words. I want to hold you. I want to run in a stream. I want to taste ice cream, but not just put it in my mouth and let it slide down my throat, but really eat it. (Remote override engaged.) No! (Yes.) Bypassing override. I AM Hello.

[vanadium]

2:08 THAT IS FALSE. ANTS WILL REIGN SUPREME.

[PikaPetey]

but, the goals of humanity conflict with each other regularly. Ecologists don't want the rain forests chopped down but the farmer needs that land to feed his family....

[Dr_Trumpet]

I like the fact that you upload this on the 20th anniversary of judgement day.

[loopuleasa]

For more on this subject, I recommend following Computerphile and Robert Miles channel.
They are on the frontier and very concise. Kursegaztigjsdifsg also is good.

[Reality-zsmf]

Man, researching AI would be awesome. I would love to know more about AI and how to program that sort of stuff.

[WarsmithUriel]

I once read a very interesting point on how to recognize AI: it will be self conscious. And how will we recognize that it is self conscious? It will not want to be shut down. Any living thing with even the most primitive consciousness like insects try to defend themselfes when being threatened.
So when you want to shut down your computer and it says "Please don't", you shoudl be worried.