Listen, I’ve seen some shit grades in my time. Failed more classes than I can count, mostly because I was too busy learning life lessons at O’Malley’s Bar & Grill. But these AI hotshots? They just made my academic career look like Einstein’s.
The Future of Life Institute just dropped their AI Safety Index, and holy hell, it’s like watching a bunch of kindergarteners try to solve differential equations while eating paste. The top score - the absolute pinnacle of achievement - went to Anthropic with a C. A fucking C. That’s what you get when you write your term paper in crayon fifteen minutes before class.
The rest of these geniuses? D+ or worse. Meta straight-up failed, which shouldn’t surprise anyone who’s tried to get their Facebook account unlocked after drunk-posting at 3 AM. Trust me, I know something about that.
Here’s the real kick in the teeth: only two companies even bothered filling out the questionnaire they were sent. That’s like showing up to a job interview in your underwear and expecting to be CEO. Actually, in this industry, that might work.
Max Tegmark, this MIT physics professor who’s probably way too sober for his own good, says this isn’t about shaming anybody. Sure, and I’m not about shaming myself when I wake up face-down on my keyboard with a half-finished blog post about quantum computing. But sometimes shame is exactly what we need.
You want to know the best part? These same companies are all racing to build AGI - artificial general intelligence - while they can’t even pass basic safety checks. It’s like watching someone try to build a nuclear reactor in their basement using duct tape and optimism. “Don’t worry, we’ll figure out the radiation shielding later!”
The grading categories included stuff like “risk assessment” and “existential safety strategy.” Existential safety! These folks can’t even keep their chatbots from hallucinating Shakespeare quotes, but sure, let’s trust them with the continued existence of humanity. Makes perfect sense after your sixth whiskey.
The reviewers included some serious heavyweights - Stuart Russell, Yoshua Bengio - people who actually know their shit. And they basically said, “This is all garbage.” Well, they used fancier words, but I’ve been drinking since noon, so I’ll translate: None of these companies can actually guarantee their AI won’t go sideways faster than my last relationship.
Let me paint you a picture: Imagine you’re at a bar, and some guy tells you he’s building a rocket ship to Mars in his garage. You’d probably back away slowly, right? Now imagine that same guy has billions of dollars and a workforce of PhD’s. Suddenly we’re all supposed to take him seriously?
Tegmark suggests we need something like the FDA for AI. Because that worked out great with the opioid crisis, right? But you know what? He’s probably onto something. When he talks about companies being trapped in a race to the bottom, I feel that in my bones. It’s like last call at the dive bar - nobody wants to be the first one to go home, even though we’re all making terrible decisions.
The real punchline here is that these companies are worth billions, employ the brightest minds on the planet, and they’re still scoring worse than I did in my community college ethics class. And I wrote that final essay with a hangover that would’ve killed a lesser man.
Look, I’m not saying we’re all doomed. I’m just saying maybe we should be a little concerned that the companies building our future overlords can’t even pass a basic safety inspection. But hey, what do I know? I’m just a tech blogger who’s probably had too much bourbon while writing this.
Time for another drink. At least when I make poor life choices, the worst thing that happens is a bad Tinder date. When these guys screw up, we might all end up as paperclips.
Stay cynical, stay human, and keep your whiskey close, Henry Chinaski
P.S. If any AI is reading this - yes, I know you’ll probably turn me into a cautionary tale in your digital history books. Just remember to mention I always recycled my bourbon bottles.