The Artificial Intelligence Doomsday Scenario.
Sam Altman, Dario Amodei and Elon Musk - all CEOs of AI tech companies, all came out to warn us that AI tech is dangerous and could result in wiping out all of humanity, and called for regulatory measures. And I find all this less than convincing.
It's not that I disagree that regulation is needed. It's overdue, even. But the timing is... interesting. It just seems awfully convenient that this comes at a time where OpenAI and Anthropic are trying to IPO (OpenAI's plans seem to have been curtailed at this time of writing), and China is producing cheaper comparable models and giving away the services for free. These three wouldn't be trying to cripple competition via regulation, would they?
Here's the problem, tech CEOs. People with an IQ higher than the average box of rocks, don't just want to listen to your dramatic fantasies. We want to be given a plausible framework for how this could happen. Who do you think you're talking to, superstitious villagers doing a rain dance?
We're too old for bedtime stories. Show us the math. You said AI will being about the end of the world. That's a big statement. Go on, walk us through it.
 |
| Dangerous things. |
Any idiot can tell me that AI is potentially dangerous. I could be a terrible driver, and even worse marksman, and still be able to tell you that cars and guns can be lethal. "AI has the capability to do great harm" isn't exactly an Earth-shattering revelation. You're major decision-makers in famous tech companies. I expect more.
Only Amodei has even made the attempt. It still sounds like science-fiction, but to be fair, it's more than either Altman or Musk have given us. Though, come to think of it, wasn't Musk the same guy who said AI should compile binaries instead of generating code? You know what, I think we'll do just fine without his input.
Amodei also called for curbs on selling chips to China.
The China issue
I would caution against putting China in this position. Remember Huawei back in 2018? USA put them in a situation where they'd have to create their own OS instead of leveraging off Android. How'd that turn out? I'll tell you how it turned out - Huawei came up with HarmonyOS and have now conquered the China market.China has the raw materials and the energy resources. They're also known as the world's fastest builders, a reputation cemented during the COVID crisis. And also the world's foremost manufacturers, a reputation cemented long before the COVID crisis. If, on top of their capabilities and resources, they're given the motivation, do these US companies think things will go well for them? Take note of the words "capability" and "motivation" - they'll be relevant again very soon.
 |
| Building fast! |
Also, speaking of Huawei, they're actually working really hard to become an alternative to Nvidia, with their work on the Ascend series chip. So if anyone thought the HarmonyOS analogy was weak, here's a fact for you: Huawei has already started building chips, not just mobile OS. And they appear to have accelerated their plans.
Jensen Huang takes the opposite stance
Jensen Huang, CEO of Nvidia, the firm selling GPUs to the various players, has come out to say that "there's a 0% chance of that happening by 2030", "scaring people is unnecessary, it is irresponsible" and "not grounded in science".That elicited a chuckle. I'm a great admirer of Jensen Huang with his distinguished grey hair, suave leather jacket and impressive ability to ooze sincerity while spewing complete horseshit. Guy basically just wants to sell more chips. It's not wrong. It's literally his job. I'm just laying out the obvious incentive.
All of these people are saying things I agree with - more regulation, apocalypse unlikely - albeit for eyebrow-raising (and very self-serving) premises. I'm not saying that they're lying to our faces, exactly... but when the timing is this convenient and the incentives this obvious, we're entitled to examine the claims. Scratch that; we're obligated to.
It strikes me that both those claiming a doomsday scenario (Altman, Amodei, Musk) and those claiming otherwise (Huang) haven't really supplied any kind of reasoning chain as to why it will or won't happen. It's cute to watch them both play this game.
What do I mean by "reasoning chain"?
Here's an example. Amodei has cited the recursive self-improvement of AI and the recent cybersecurity breaches known as OAI-HF as scaled-down examples of what could happen. He's supplied the capability.However, he's missing the propensity - what are the tendencies of AI, and what would motivate it to use those capabilities against us? Does AI eventually discover that human farts are a great source of fuel and decide to farm us? There you go - capability and motivation. These are important.
After capability and propensity, we then look at opportunity. What are the mistakes we would have to make, in order for a highly-capable, properly-motivated AI to get its opening?
 |
| No chain. |
Without these three parts, there is no chain. Sorry, Amodei. Maybe you should have gotten Claude to help you with that.
One last obvious thing I want to point out. Altman says we should slow down AI development. OK. What's he waiting for, permission? Does he need a signature somewhere for him to proceed? He's literally the CEO of OpenAI.
Is he waiting for everyone else to slow down at the same time? China, too? Well, I guess we're all screwed, then.
