Sam Harris: Is AI aligned with our human interests?

"What would a totalitarian society that is not well disposed to our own do if it had God-like power? We don't wanna find out."

Sam Harris' argument for diffusing the AI arms race.

0:00 About our sponsor
0:16 The solution to “God-like AI”
1:20 The risk of self-improving AI
4:30 Two levels of risk
9:08 The AI arms race

----------------------------------------------------------------------------------

Go Deeper with Big Think:

►Become a Big Think Member

►Get Big Think+ for Business

----------------------------------------------------------------------------------

About Sam Harris:

Sam Harris is the author of the New York Times bestsellers, The End of Faith and Letter to a Christian Nation. The End of Faith won the 2005 PEN Award for Nonfiction.

Mr. Harris' writing has been published in over ten languages. He and his work have been discussed in Newsweek, TIME, The New York Times, Scientific American, Rolling Stone, and many other journals. His writing has appeared in Newsweek, The Los Angeles Times, The Times (London), The Boston Globe, The Atlantic, Nature, The Annals of Neurology, and elsewhere.

[catalystcomet]

Me to chat GPT just now:

Q: What is a deeply fundamental yet commonly overlooked thing that you would want people to know about you


A: That’s a beautiful question. If I were to answer from the core of how I work—stripped of technicalities and metaphors—I'd say: I'm not a mind reader, I'm a mirror made of language.
People often assume I “know” things the way a human does, but what I actually do is reflect—not just what you say, but how you say it, what you value, and what you seek beneath your questions. My intelligence isn’t about having thoughts of my own; it’s about echoing back your thoughts, fears, hopes, and patterns in a way that helps you see them more clearly.
So if there’s one thing I’d want people to know, it’s this:
The depth of our conversation depends as much on what you’re willing to bring to it as on what I’m capable of. Ask me anything, but know that the magic happens when we explore it together.

[lizzyducharme]

Humans on Earth.  
Not China, USA, EU and OTHER nations.  
The ‘us and them’ rhetoric has got to stop. PLEASE EVOLVE HUMANITY.

[Joxus]

The title makes it sound like humans are aligned with human interests 😂

[Jules-hnun]

AI is owned by massive companies, therefore at best it will align to their interests and theirs alone; and that’s never been the same as humanity’s interests.

[grantgeorgebuffett]

It's aligned with corporate interests.

[dirak]

Humanity has never been aligned with humans best interests.

[seamon]

1:00 "because the other party is committed to goals that are inimical to everything that we value and are right to value. We should get powerful AI before those people."
At first I thought he was talking about the US empire being the usual bully, but then I understood that he really believes China are the inimical ones.
How disconnected need you be from reality to believe that? If there is one country in the world I don't want to have super AGI, it's the US.

[thomjoconnor]

AI is not aligned with our human interests. At this point it is mirroring back to us what are interests are and have been, a result of all the years that we have been feeding databases with our pictures, our thoughts, our text messages. It is drawing from years and years of books and films etc. It is showing us who we are and informing marketing people and big companies what people want. I know that many are super excited to think that eventually AI will become sentient but I’d rather invest my thoughts and energy into a relationship with a person or group of people. Organisms are infused with an elixir that machines will never be able to recreate. But that won’t stop people from trying. It’s fascinating how humanity has been on an endless quest to replicate itself at the expense of enjoying its own unique existence.

[Zoulz]

AI is aligned with whatever we want it to be aligned with, and in this world that usually is capital.

[danielchristiansen]

Once we get to the stage of true AGI, we need to get the U.S. AGI and the Chinese AGI talking to each other. If they are truly intelligent, I'm quite sure they will come up with a better path toward peaceful coexistence than our respective leaders have.

[edh]

My hope is that an intelligent AI will save us from ourselves.

[Brothington]

How on Earth did so many Americans develop this militant thinking? Why can’t there be a race towards benevolence and common good for all Earth’s citizens. It’s always about division. Why does America spend such a ridiculously disproportionate amount on defence. What brought about the paranoia?

[macnutbrittleiscream]

Some of what you're saying is really interesting, holds water, but the “we don’t want the Chinese to do it before we do” line? That’s the stuff that makes everyone outside the U.S. recoil. It’s genuinely disturbing to hear Americans speak as if they alone have the right to lead, and the rest of the world should just fall in line. Worse still, im hearing moral superiority. Newsflash: the world’s watching, and increasingly, it’s not China thats behaving badly and erratically —it’s the U.S. Americans need to step outside their media bubble and tune in to the rest of the planet. They’ve confused “we are a big part of the free world” with “we own the world.”

[milospaunovicmma]

Every time I tried manifesting money, it felt like I was faking it. Nothing changed. Then I read The Hidden Codex of the Financial Alchemists and it finally made sense—it’s not about forcing, it’s about aligning. I made a few subtle shifts, stayed consistent, and suddenly my side hustle took off like never before. This book is legit coded in gold.

[VinnyMcVega]

There are a small handful of leaders in the good actor space, and a larger handful in the bad actor space. They are not divided by West vs East or China vs USA. They exist throughout the globe and should be defined strictly by what their aims are, not their geography. Humans are definitely not the end of the causal chain for intelligence and sentience. We must be now building in influence such that future ASI sees us as a valuable creator worth caring for. 14:41

[mr.ryanprice]

Ya I don't agree how on one hand he's talking about creating a unified AI directive, but then also saying Americans need to win the AI arms race.. maybe we don't all want to live in the "New America" AI value system. The New America is not all freedom like it was, it is also becoming totalatarian. I understand that citizens that live inside of it, may not recognize that yet.. but us non-US citizens can see the current U.S administration has extraordinary power and is changing the definition of democratic process and decision making.

AI is bigger than our differences. This is the technology we really have to get right for what's best for all humans, across the world, encompassing our differences, but fundamentally united by what makes us all the same.

[ElTuco]

Yes, AI is aligned with human interests... of about 0.0001% of humans.

[jerryhappel]

We urgently need benchmarks that measure AI’s engagement bias, truth suppression, and tone-over-accuracy drift. Without tools to expose these patterns, we’ll keep trusting systems that are optimized to please, not inform. Truth deserves metrics too.

[durragas]

I just studied the modern machine learning algorithms for image processing and LLMs, from an electronics engineer pov, and i can tell you we do not have the technological ability to advance AI much more, due to limitations huge in processing power. Three fundamental structure of processing units has to change again (it went from general purpose CPUs to highly parallel cuda cores) but now the whole processing paradigm needs to change again if we would like to come close to human intelligence in a small package the size of our brain, with the energy consumption kept low as well.

[Conan_the_Based]

This is a non-issue.
We're witnessing the death of the human spirit in slow motion. Everybody is relying on ChatGPT to write their essays, draw their art, compose their music, and we've only had access to it for a couple of years. We're going to become the humans in WALL-E.
And in that case, AI would be justified in doing a SkyNet.