Some reflections on Anthropic's "AI pause button"
[Different from the last two posts that got a lot of traction, this is a short(-ish), stream-of-consciousness post.]
Now Anthropic is calling for a "global pause in AI." 1
I have been somewhat of a distant sympathizer of the "Pause AI" movement for reasons outlined in the previous posts, but was thinking this was an utopia by now. There's nothing we can do, the genie is out of the bottle, uh?
So this new move by Anthropic did catch my attention, but not in a good way. I have a few theories:
This is just marketing: Anthropic knows this will not happen but saying it out loud makes them look like the "good guys."
They want to pause to keep the lead: they are desperate to IPO with the lead on Enterprise mindshare/Mythos hype, etc.
This is from their "Effective Altruism" roots: they really do want to pause it for the reasons stated
Reasons #1 and #2 contradict each other. Not just because #1 implies they do not want it and #2 implies they do, but also because #1 implies this is good marketing and #2 implies this is bad marketing.
Well, we cannot read the minds of the Anthropic staff who wrote this, but I did a small search on social media and news website comments and the perception is overwhelmingly in favour of hypothesis #2 in some way (some don't mention IPO, but things like "this is just them trying to make their models look more capable than they are")
...which brings me to hypothesis #4:
- They're trying to change the public perception on movements like Pause AI and anybody saying out loud that AI advancement poses risks to society (or the job market)
I know, I know, I know. This sounds far-fetched. I also don't think this was the main goal, but I think this "side-effect" was calculated and intentionally baked in.
I'm living through this right now: comments on the web are calling my blog "suspicious", implying I'm somewhat a shill or something (lol). Gosh, is it so hard to acknowledge that some people do find AI useful/good at doing things and do think they can have negative impacts on the job market?!
Yes, I do think a lot of the Mythos hype is pure marketing. Yes, I do think (current?) agents/LLMs are no replacement for human engineers. Yes, I do hope humanity is not replaced by soulless machines. Yes, I do hate AI slop - and that's why I write all my posts by hand with minimum machine intervention (spell-checking at most) and use artwork from Unsplash to illustrate my posts.
When I wrote my first piece (on this blog), I was just venting out some of my feelings. Now I'm being accused (by some, I know) of hyping stocks for IPOs.
Anyway, whatever Anthropic is doing, it's serving a good purpose preventing members of the Anti-AI/AI-Skeptic movement from realizing (some) people who are vocalizing the dangers of AI advancement share similar goals with them. Bravo.