The best way to future-proof Future Around & Find Out is to become a paid subscriber. Please consider it. Thanks!
🤦 If only I’d published this as soon as I’d finished typing it at 5pm ET last night. Because… BREAKING NEWS… just a few hours later the US government took drastic action against Anthropic that’s forced it to revoke access to its latest models to all but US Citizens. Even non-citizen foreign nationals who work at Anthropic appear to be banned from using the Mythos and Fable models. How does Anthropic even operate right now under this order?
The move comes out of national security concerns, but also maybe it’s just a middle finger to Anthropic for its intransigence back when the Pentagon wanted to be able to use its models however it saw fit — including to deploy fully autonomous murder bots (coincidentally this is also discussed in this week’s episode and below, as Ukraine has apparently already done this).
Lots of confusion this morning after this Friday night surprise, with some crowing that Anthropic brought this on itself by calling for regulation and others saying this is “heavy-handed.” Even Gary Marcus, a frequent critic of generative AI and someone who is generally pro-regulation, says: “If you want an example of an AI regulation that can stifle innovation, this is it.”
No doubt, this is not what Dario had in mind when he called again this week for sensible AI regulation. More in my original post and the episode below
ORIGINAL POST:
Sam Altman and Dario Amodei both published essays this week on the future of AI and what we must do so everyone benefits. One of them is literally titled "Our Plan." The other one has an actual plan.
Sam has such a declarative headline — “Built to benefit everyone: our plan” — but then you feel empty after reading it. Here’s just one of Sam’s mealymouthed constructions that kinda-sorta calls for someone to maybe do something to help the world. I’ve italicized the phrases that hedge it so much as to make it meaningless:
“We have long believed there should ultimately be an international organization that helps coordinate leading AI efforts to reduce catastrophic risk. […]
“One goal of such an organization should be to make it possible for the world to take coordinated action, including slowing frontier development when needed, so societal resilience, safety, and alignment can keep pace.”
I’m certain the use of passive voice and the hedging are on purpose, so Sam can say he’s “doing something” while also not saying much. Meanwhile, Dario Amodei just comes out and says the thing, calling for “more serious and binding regulation of AI.” In his “Policy on the AI Exponential” he gets very detailed on ways in which policymakers can move more quickly to create new regulation to review frontier models for safety, create pro-employment programs, avoid empowering autocrats, create accountability rules for fully autonomous weapons, and more…
And, while others are walking away from such depressing topics as job loss, Dario repeats his past claims that AI could cause significant job loss. He then pokes at the A16Z’s of the world saying:
“It’s become popular in AI industry circles to view this as a PR problem: to say that AI needs “better marketing”. I reject this framing completely. People are worried about AI because they correctly perceive that its risks are real, not because AI CEOs have been insufficiently Panglossian.”
I am not trying to be Panglossian (exceedingly optimistic) about Anthropic’s motives here, but I do really appreciate the level of detail in Dario’s piece. It’s serious and I don’t think it’s just a schtick so they can be “the safety company” (even though there have been marketing benefits of being seen as such).
Kwaku and I dig into it all on this week’s FAFO Friday. Plus — and this story isn’t getting enough attention — according to New Scientist, two years ago Ukraine used fully autonomous “Terminator” drones that killed everything they saw. No human in the loop. Dead Russian soldiers. But rest assured, according to the drone-maker cited, it was just a one-off “test.” But how long until this is standard practice? And do we want that future?
So, yeah, maybe we should get planning…
~ Dan
If you don’t already get Future Around & Find Out emails, go ahead: