AI Leaders’ Devious Plan

Is AI set to wipe out humanity? This is the (manufactured) question of the week.

The firestorm erupted after a researcher at leading AI firm Anthropic resigned, saying the big companies are “gambling with our lives”.

This was the post on X (Twitter) which set things off:

The post went crazy viral, generating 157 million views, 756,000 likes, and 155,000 retweets. It’s one of the most popular tweets of the year.

At first, I assumed this was a natural, organic thing. Good for Jacob, standing up for his beliefs.

But then the hits just kept coming.

Next up, Anthropic’s head of alignment (~safety), re-posted Jacob’s message, and threw in an outrageous estimate of the risk.

Source: X

“We really do earnestly believe AI could kill all humans! I personally think it is [greater than] 10% chance within the next decade”.

What?! He sounds so excited about it. And that’s coming from Anthropic’s current head of alignment.

This is what a sophisticated PR campaign looks like. The goal: scare the crap out of the public and pass favorable regulation. Let’s examine the evidence.

Exhibits A, B and C

Yesterday, Anthropic also released a 154-page report about all bad actors trying to use their systems for evil.

According to the company, bad guys attempted to use Anthropic’s AI model Claude to:

• Yemen’s Houthis tried to build hypersonic missiles
• China used it to track Muslim Uyghurs
• Russia hacked Ukrainian targets
• Someone attempted to genetically engineer viruses

Anthropic says they “disrupted every operation in the report.” Disrupted is an interesting word choice here. They didn’t say thwarted, or stopped. They said disrupted. Let’s file that away for the future.

The company implies that the world is only still spinning due to their robust safety measures.

This is one thing you’ll notice about these big AI companies. Only their specific company can be trusted with superintelligence. Everybody else is out to destroy the world with their reckless ways. That’s the messaging from almost all the big guys.

And THEN we got another big hit. Reuters yesterday:

OpenAI pushes for mandatory national AI safety rules

“The prospect of AI-accelerated AI development demands more than voluntary commitments. The United States needs ​mandatory, capability-based national regulation that can evolve as the technology does,” OpenAI Chief Global Affairs Officer Chris Lehane said in ​a blog post.

…OpenAI is urging ​Congress to adopt capability-based national AI safety requirements, including testing standards, independent assessments, cybersecurity protections and incident-reporting rules for the most advanced AI systems.

Now it’s coming together. Some people are probably thinking, Well, gosh that seems responsible! Who could argue against common sense safety laws?

Therein lies the problem. They won’t be common sense rules. They will be designed to crush smaller competitors like bugs, and outlaw open-source models (which cost far less). And probably eliminate online privacy, as well.

An FDA for AI

Most of the proponents of AI regulation advocate the creation of a sort of “FDA for AI”.

The government would conduct trials to make sure new models are safe. Companies would have to pay huge regulatory costs to get their models tested and monitor users. Many smaller startups won’t be able to afford that.

Unless we prevent it, these laws will hand an oligopoly to the existing big players. Primarily OpenAI, Anthropic, and Google.

The tobacco firms ran this playbook back in the 2000s. Philip Morris supported FDA regulation of the industry, because it gave them a big advantage.

Today every new tobacco product has to go through the FDA. So new competitors are few and far between.

Great news for the incumbents. This is one of the reasons I own Altria (MO), Philip Morris’ American division. The company prints money and hands out huge dividends. Competition is limited by law.

But if this happens in AI, it will be extremely bad for the country. Innovation will dry up, and foreign competitors with more competition-friendly systems will eventually win.

To Be Clear…

There may need to be some sort of regulation around AI at some point. But we must not let the leaders of this industry dictate the rules. There are countless examples of why this is a horrible idea.

They will pull the ladder up behind themselves and set up a cozy environment with limited competition. They’ll print money, and have unprecedented influence over the population, and even the government.

Any AI regulation needs to allow startups to compete with the big boys. It needs to allow the use of open-source models, so small companies aren’t stuck paying gobs of money to an oligopoly.

That’s what I’d prefer. But we may not get it. The FDA approach seems likely. Less competition and innovation, higher prices, and malign influence.

What else does the future hold? Probably a digital ID system. This is the elimination of online privacy I mentioned earlier.

To use a sophisticated AI model, you’ll need a digital government ID. It will be tied to your real identity, which will reduce potential for abuse. Eventually, to access the internet at all, you’ll need to authenticate with your biometric data (eye scan, fingerprint, etc). This is inevitable in my view. It will happen soon after the first gigantic AI-powered cyber attack.

Oh yeah, OpenAI and Anthropic are both planning an IPO soon. So the timing on this whole campaign is no accident.

Open-source AI is giving the big guys real competition, and they’d like to eliminate that risk. Will they succeed? I don’t know. But our political system is extremely vulnerable to these types of operations. And with so much money sloshing around, all the wheels will be greased.

More soon.