Vincent Carchidi

I'd be very interested in seeing those results...

Most of it makes sense yeah, but I thought "regular use" would be significantly higher than 23% across all demographics

This is really interesting. I think the "job creation" question can be taken sort of with a grain of salt since coverage of AI and jobs is almost uniformly negative. So not very surprising to see that.

But the "regularly use" stat kind of shocks me.

question 41

q42

q43

q44

Fresh polling on #ai usage, highlights

- roughly 70% have tried with 30% regularly, if we remove retirees
- far more mistrust, surprising no one
- split on economic impact ~50/50
- job outlook heavily negative

d3nkl3psvxxpe9.cloudfront.net/documents/ec...

I think he was saying that's the effect, not the motivation

Once the end is in sight, that's about the end of the excitement

On Palantir's AIP? I wouldn't be too surprised by that, but I'm not sure. I don't think they're on GenAI mil at all.

Either way, I honestly dont envy Amodei's position with this right now. Negotiating with this DoD...

I haven't, but this is now on my reading list, thanks for sharing!

AFAIK, Palantir's platform where third-party models can be accessed is separate from GenAI dot mil. And I believe OpenAI has integrated or is integrating a version of ChatGPT into it (very recent news).

That said...I think GenAI dot mil is a very unserious thing which is basically hit or miss.

Context

bsky.app/profile/vcar...

Also, you all owe Missy Cummings an apology for acting like this wasn't obviously the case.

Ironically, the biggest losers of that designation being made would be Anthropic's competitors - they would have to effectively excise Anthropic's products from their own work - but Anthropic has basically already kicked them all off its APIs.

One of those times when my personal interests intersect with my day job, and this is nuts.

My feeling on this is that it's very complicated and some of this is happening and will continue to happen without necessarily generalizing to all cases. I can imagine a plausible future where software engineering is radically upended, while most everyone else can get by without total disruption.

All very reasonable. And I think this is the tension between two basically sound approaches; (1) AI works well and therefore will cause job losses/other losses; (2) AI does not work well enough but will still be used to displace workers/etc (and put extra burdens on others).

One interesting thing about this is that I did not find these cases useful until basically the second half of last year. Which is interesting. Lines up with many other people's experiences. Prior to that, it really was basically nil. I think much of that comes down to tooling and tuning.

What I do with those 2-3 hours does not really amount to anything, because it means I'm probably poasting on here instead of, idk, transforming my productivity or whatever.

Maybe I'm unimaginative. But for side projects, it's a similar story - useful, time-saving sometimes, but not panaceas.

Yeah. I can only speak anecdotally, but I have ChatGPT and Gemini Enterprise versions available for (non-coding) work. And I do use Gemini probably regularly, often for things that might called a fancy form of search, and it does save me time. I would say 2-3 hours per week, maybe.

Maybe. I mean, I don't see a principled reason why that couldn't be true.

But I think everything from the development of LLMs to how they actually impact the labor market is extremely un-intuitive. What you say might happen, but I think these predictions suffer from the time in which they are made.

Like, if my work was affected the way (some) coders say their work is affected, I'd probably have a sense of frustration and anxiety too!

But I would gently suggest coding is probably not as representative as some coders think. Could always be wrong.

Quite possibly. But my sense is that the sudden "things have changed and you skeptics need to pay attention" is because of coding automation, and Claude Code specifically.

Good points here, and I don't really disagree with anything you say. I guess my question is: what does this have to do with the capability denial stuff that's above in the thread? Sounds like it'll happen whether someone buys into various claims/talking points or not, no?

What's the disagreement here though?

Worthwhile thread

Amanda Askell, who is contributing to the project of displacing human beings that I oppose, is the philosopher of our time and I *will* defend her honor.

I care deeply about people losing their jobs to AI, and the effects this has on the most vulnerable. This is why I have decided to spend most of my time arguing with those people about whether Claude has feelings.

AI would be so much better culturally if this was basically the consensus

This an open call for responses

bsky.app/profile/vcar...

(And me saying this only makes sense because I lack this exposure and the anxiety that would come with it.)