So there’s this whole thing from Anthropic again: “let’s slow down with AI, because from now on the software basically writes itself.” It all sounds lovely, and it’s all technically true - but let me add my own footnote, since I’ve apparently become the resident bird of ill omen. Consider this a little extra note on top of what Italy’s very talented AI influencers have already said.
The part nobody says out loud
One of the fundamental reasons Anthropic keeps calling for a slowdown is simpler than the grand narrative: they didn’t invest in compute as aggressively as everyone else.
Look at OpenAI. Look at Musk, who just handed them what amounts to a giant cluster on rent. Anthropic didn’t move as fast on compute as the others did - and now they’re the ones calling for everyone to ease off the gas.
It’s a beautifully packaged message, wrapped in safety and responsibility. But strip away the eloquent gibberish and you’re left with a much more mundane story: when you’re behind on the hardware, “let’s all slow down” starts to sound a lot more appealing.