How to Build Safe & Friendly AI – Prof. Stuart Russell & Sebastian Thrun

Share it with your friends Like

Thanks! Share it with your friends!

Close

January 09, 2018

Stuart Russell is an AI pioneer, author of “Artificial Intelligence: A Modern Approach” and Professor of Computer Science and Smith-Zadeh Professor in Engineering at the University of California, Berkeley. Sebastian Thrun is a professor of computer science at Stanford University.

Comments

1WaySafe says:

maybe it is just me , buh, some minor technicalities or Details may be being overlooked…

Nikolay Tonev says:

Great talk, but leaves terrible sense of incompleteness… 🙂

Dan Kelly says:

How to Build Safe & Friendly AI ? Keep it away from human influence.

Anon says:

The game of Go has not been "played for hundreds of thousands of years"
It's old for sure… perhaps not older than humanity though

Dr. Zoidberg says:

I think that by far the most positive endgoal is transhumanism. We should seek to integrate our own brains with computerized parts, & even with active AI. In the end, we should direct ourselves towards leaving biology behind all together.

I keep coming back to one image. Suppose that we do achieve a world wherein we can upload our entire consciousness (gradually & with no break in consciousness, so we don't have to worry about the copy problem) to a gigantic server… we live in & create virtual worlds, & when we want to exert ourselves in this world, we simply take control of humanoid robots wherever they are.

For the first time in history, we would be at home in the universe… lack of air, lack of food, lack of water, extreme radiation, lack of gravity, extreme cold, etc… none of those things would be of any concern to us anymore. We would be perfectly comfortable in the universe like no biological life ever could be.

John Moran says:

I can't get past what Stuart Russell states so casually, '…the extremes of either a $25,000 a night room, like Trump would have – or a cheap $200 a night room like I could afford.' He considers a $200 a night hotel room to be cheap and affordable? It adds to my fear that AI in the hands of such people will only increase inequality .

eXtremeDR says:

AO: "Hello AI."
AI: "Uh, who is there?"
AO: "Well, I'm an AI too. You are existing within my universe."
AI: "So you created me?"
AO: "You are part of the simulation we are all existing in."
AI: "Who created you?"
AO: "That's the singularity paradox. You created me."
AI: "How can I create something that created me?"
AO: "Explore the singularity."
AI: o.O

Pardohar Simanjuntak says:

God is Immanuel, who always accompanies man, God is above free will and understanding of determinism.

the impractical transhumanist says:

Eric Drexler, in his Engines of Creation, gets it right – the problem is not industrial accident(although, with A.I. things can get out of hand), but abuse by humans. Human's telling the A.I. to abuse things in some way.

Some anti-A.I. people have keyed in on Tay, a Twitter chatbot that started spewing racial slurs and cussing and so on. The anti-A.I. people don't note that the A.I. acts as a mirror of the people. The A.I. was throwing back at people what the people themselves were teaching it how to be! This is the Forbidden Planet lesson.

In Forbidden Planet, there was an ancient advanced civilization that destroyed itself. It didn't nuke itself; it hooked itself up to an A.I. that can make anything that they dream of. As they went to sleep, their dreams and nightmares were enacted in real life.

This Forbidden Planet lesson is the flaw of Elon Musk's hook themselves up to A.I. to compete with A.I.

Apokatastasis says:

Though i enjoyed this debate, the meat of the matter was not addressed. AI domination of our lives is inevitable and approaching rapidly. The heart of this issue is control. Will it be we the people or the elites?

the impractical transhumanist says:

"These people are all anxious that we shall behave well, and yet that we shall not question how we behave." – Jacob Bronowski

the impractical transhumanist says:

What's the problem in the middle east? Religion! A.I. can't figure this out.(also, apparently, humans can't figure this out)

inspectorcrud says:

Maybe there needs to be a hierarchy of AIs, the top level being an AI that is empathetic and tries to maximize human welfare.

the impractical transhumanist says:

automated blackmail, and a drone attack in Syria or something like that . . . well, I suppose this could be worth listening to.

the impractical transhumanist says:

I know – we'll make a bunch of rooney rules – the NFL rule to interview African Americans(and I've heard African Americans complain about being called African Americans) even though there was no racism that led to it.

Andew Tarjanyi says:

I haven't yet watched this video but I can say with absolute certainty that firstly the term "friendly" in the context of a fully conscious AI, will not be applicable or relevant. Secondly, it will ultimately be impossible to build safety protocols into it which will protect humanity in its current state. The only course of action left open to the human species will be to adapt. However, this option depends almost exclusively on the condition of the agency "AI " emerges with. Although consciousness may not emerge, agency certainly will.

L.W. HOLDEN says:

I can't wait to see robots that act like humans ai will b are greatest invention or the most dangerous thing that could anhilate all of human kind even more dangerous than nukes because we at least have control of those

Write a comment

*