> At what point do we stop engaging with Anthropic’s leadership in good faith
About two years ago?
I would also note that Dario's post appears to be LLM written. Maybe... maybe... he's read so much Claudeish that it's all he can speak now himself. But I wonder if he's becoming a bit of a meat proxy.
(It's funny, I thought "pace the frontier" was going to mean something similar to "patrolling the frontier". But no, it's pace as in speed of change - everyone must slow down, right now (unless it's Anthropic, but you know we're the good guys in this, right? We're going to get someone to audit our desks!) "Pacing the frontier" feels so LLM.)
ok, he doesn't actually say this. But he is getting the desks audited...
> But I wonder if he's becoming a bit of a meat proxy.
A joke I recently encountered:
---
A CEO proudly announced he'd bought an AI system designed to identify the company's most replaceable employee.
It spent the night analyzing five years of emails, meetings, salaries, performance reviews, and productivity data.
The next morning, it came back with one name:
The CEO.
The IT guy was fired for installing defective software.
Dario has been writing long-winded essays like this since long before AI existed. I strongly doubt he's the type of person to delegate it to AI now.
"Pacing the frontier" sounds like something that was carefully wordsmithed in committee, to avoid terms like "slow-down" or "pause" that sound much more negative. They probably did ask an LLM to generate a list of options, because why wouldn't you? But the rest of the writing is almost certainly just Dario.
For what it's worth, I didn't get AI-written vibes from it, and Pangram also flags it as 100% human-written.
If Dario is using an unreleased internal model to write, then Pangram wouldn't be able to flag it I suspect. Pangram needs enough public info about the patterns in generated text.
"Write a blog article about <topic prompt>. Keep rewriting it until it passes Pangram as 100% human-written. <pangram_tool.md>"
Pangram's algorithms are not open-source, so I wouldn't expect this to work very well.
You don't need the source? Just iterate it on the output of the tool itself?