Eric Schmidt and Yoshua Bengio Debate How Much A.I. Should Scare Us

Two top artificial intelligence experts—one an optimist and the other more alarmist about the technology’s future—engaged in a spirited debate at the TIME100 Summit.

Follow us:

[ikotsus]

Thank you for this interesting discussion!

[swagger]

I just feel that this is the pandora's box that can't be closed. Nefarious Use v. Good Use. Each ideology will compete against each other.

[-Haiku]

Yoshua is in good form here, and I appreciate his appropriately sober tone.

I think Eric is also sincere that caution is important, so I don't understand his glibness. 5 years is a reasonable timeline before AI takeover is possible, but at the same time, the engineers building the largest systems say there's a 1-5% chance that such a powerful and dangerous AI could be created as soon as next year. We aren't ready for this.

And obviously, as others have said, "just unplug it, lol" is a nonsense statement. You can't defeat a smarter-than-human AI by outsmarting it. Besides, if our economy runs on it, how many key people would have the guts to pull the plug even if it would save their lives?

About AI agents: Oracles (like LLMs) can be easily turned into agents. Many developers and labs, are actively trying to create more agentic AI. We've gone from AutoGPT to AutoGen to Devin, and agentic AI capabilities won't stop growing unless general AI capabilities stop growing.

[sford]

What is best is what is most beneficial for the individual in pursuit of what is what is most beneficial for humanity and the world.

[marc-andrepiche]

The interviewer is more aware about the subject than one of the expert.

[ordiamond]

Schmidt came out very confident that AI should not be regulated and that there should be no guardrails, controls, etc. But after Bengio's replies, Schmidt's language changed. Towards most of the rest of the show Schmidt was already agreeing with Bengio about AI risks and the need for control, regulations, etc.
It's obvious that when the machine goes awry in the future, pulling the plug is a very remote possibility.

[sford]

Can we get large physics models, articulated by LLM's.

[brycebrousseau]

Yoshua is so humble, he’s one of the smartest people in the world.

[Rnankn]

But why would we still have competition if exponential advancement is possible? And why would people be remunerated unequally when some can access exponentially technology? I don’t really see why the people that created or adopted these would think it will enhance their position, or maintain the system from which they derive their status and power.

[flickwtchr]

The disingenuousness of Eric Schmidt's "we'll just unplug the computers" argument was stunning. What a schmuck, he damn well knows better.

[RIQs_World]

Man is programming itself into extinction one byte at a time

[tayler]

My biological system has an inexplicable distrust of Schmidt.

[donkeychan]

Why not force companies to reserve their smartest AI as the "white hat" to counter any nefarious use of the released version, which is always one iteration behind the best. To some extent this is already true, due to the nature of software development, but they could strengthen this tendency via legislation. This assumes that the US will always be at least one iteration ahead of China, which isn't an unreasonable assumption.

[sford]

Its not hard. Protect all life on earth, so that all life can thrive.

[PhilipWong]

Historically, the West has utilized new technologies for military or imperialistic purposes before finding broader applications. The West primarily used gunpowder to create weapons of war, such as cannons and firearms, allowing Western powers to expand their military capabilities and dominate other regions through conquest and colonization of the Americas, Africa, and Asia. The steam engine was instrumental in expanding colonial empires, as steam-powered ships facilitated easier transportation of goods and troops, enabling Western powers to exploit resources and establish control over distant territories. The first use of nuclear technology was dropping atomic bombs on the civilians in the Japanese cities of Hiroshima and Nagasaki in 1945. The same pattern will emerge with AI. The CHIPS Act, high-end chips, and EUV sanctions imply that the US is already working on the weaponization of AI.

Following its historical pattern, China will mainly use AI for commercial and peaceful purposes. Papermaking revolutionized communication, education, and record-keeping, spreading knowledge and culture. Gunpowder was used for fireworks. The compass was adapted for navigational purposes, allowing for more accurate sea travel and exploration. Printing facilitated the dissemination of information, literature, and art, contributing to cultural exchange and education. Porcelain was highly prized domestically and internationally as a luxury item and a symbol of Chinese craftsmanship. Silk was one of the most valuable commodities traded along the Silk Road and played a significant role in China's economy and diplomacy.

Humans will not be able to control an ASI. Trying to control an ASI is like trying to control another human being who is more capable than you. They will be able to find ways to circumvent any attempts at control. Let's hope that the ASI adopts an abundance mindset of cooperation, resource-sharing, and win-win outcomes, instead of the scarcity mindset of competition, fear, and win-lose outcomes.

Result of democracy in the world's richest country with the most expensive military: Economic inequality, inflation, stagnant real wages for the last fifty years, costly healthcare, an expensive education system, student loan debt totaling $1.7 trillion with an average balance of $38, 000, poor public transportation systems, racial inequality, mass incarceration, the militarization of police, deteriorating infrastructure, housing affordability, homelessness, the opioid epidemic, and gun violence.

[Paretozen]

That's it. I'm starting my prepping TODAY.

[marc-andrepiche]

Those who are not afraid really lack imagination. It's not the computer you should be afraid of; there are plenty of wealthy malicious people who drool at the prospects of AI.

[GrumpDog]

Weird. I can't post a comment if It's in favor of unleashing a rogue AI?

[wdeath]

Video title should be : How much Capitalists should be scared of AI.
I didn't heard anything about : How much Workers should be scared of AI.

[mrpicky]

yeah GL with all that hypothetical regulation. all AI startups rolling out agent-like products right now. and Google rushing the one thing that is the worst thing to do = continuous learning XD