John Warner

a recommendation: if you’re an english or history professor and you’ve not recently read sample ap essays earning top marks, you might want to check out what hundreds of thousands of high school students think counts as excellent college level work

This is not a policy oriented around helping students learn meaningful things. Training students to do school by avoiding AI detection is not worth anyone's time.

I meant the day after tomorrow. I'm scheduled to be on a plane during the game.

One thing AI has definitely conquered is the captcha on my website contact box. I'm getting 10 pitches a day for my "AI business in a box." I wonder if the business in a box is a captcha defeating app that can pitch my own business.

Just supremely irritated that I'm going to be traveling during the women's hockey final tomorrow. Almost hoping for a delay so I can stream it while in a terminal.

It's baffling that they would go with the maxxing strategy when the alternative is so obvious.

Dear @libbyapp.com: I was just pitched by your PR firm to promote Libby as part of the "bookmaxxing" trend and I would like to strongly urge you to rethink associating your app with a trend with origins in white-supremacist/manosphere spaces. The idea that reading is something to optimize is gross.

The Conservatıve ruturısı to live in a country where government encourages innovation and success rather than throwing up roadblocks. My vision is partly inspired by the New Canaan Holdfast in the Brilliance book series by Marcus Sakey.7 In Sakey's alternate America, one percent of the U.S. population is born with superhuman smarts. Many of these "brilliants" decide to live in a huge section of Wyoming purchased by a brilliant uberbillionaire to both escape growing persecution and live in a place that encourages their talents. The X-Men meet Silicon Valley. In the NCH, brilliant scientists greatly advance all manner of technology. Eventually those marvels will create a more prosperous America and world. We need to find the brilliance in everyone and put it to work.

“My vision [for an Up Wing policy agenda] is partly inspired by the New Canaan Holdfast in the Brilliance book series… The X-Men meet Silicon Valley.”

Look, at some point if your book is just a series of ideas adapted from fiction, then you aren’t writing nonfiction. You’re writing fan fiction.

That's the part that gets me. I don't feel like I'm naive, but I'm surprised at the number of academics who sort of shrug and just play this new game. The pursuit of knowledge really does seem to be secondary or even non-existent.

This is a good analogy. The incentives of the system are similar and the LLM use exposes the measurement (grades/publication) as disconnected from the goal (learning/knowledge). I think there's always a structural reason under these choices that needs surfacing.

The third baseman in RealSports had a howitzer for an arm. A righty pulling the ball was like an automatic out.

And some of those games still rule!

But it could be "successful" in an area like academic research/publishing, right? I wonder if this is what's driving part of it.

There is definitely something to preferring the non-messy going on here. This may be why I have a hard time wrapping my head around it. The messy stuff is where I find everything I'm going to write about.

...motives/purpose of their academic research. Or, they see that convincing Claude is the goal there. That's interesting to think about.

I've asked if they'd share some exchanges with me, so we'll see if that happens. This person had read my book and said they got where I'm coming from, but also that I had too narrow a view of what an LLM can do as an audience. I think they want to convince me. That seems distinct from the...

This person says it gives superior helpful feedback on their academic work. I think that opens up the questions of what is it that's valued with the production of those written artifacts, but I take them at their word on their personal experiences.

So, is it that the LLM is better tuned to being helpful at moving a piece toward publication? I could see the utility in that, but it doesn't sound interesting, the way this person makes it seem.

Could be something there, though someone who fears judgment wouldn't email a stranger about what they wrote in a book. I can also say, this person definitely gets where I'm coming from. They think I have too narrow a view of what people can get out of exchanges with LLMs.

In an email exchange he says that they've mitigated against the sycophancy because he wanted to be properly challenged. I can't independently verify this, but I take them at their word. This person really is saying it's a superior colleague intellectually, not just as a booster/cheerleader.

That's interesting. I hadn't thought of that. The world at large is always going to be less responsive than an LLM that's ready on demand.

They say they've tuned their own model so it's a colleague/mentor type. I don't know if that's congenial, but it is what they desire. I don't get the impression that this person necessarily was uncomfortable with the F2F stuff. It really is more that they find Claude superior.

I thought of that, but this person says they've worked to get a model that has a colleague/mentor persona that delivers tough criticism. I can't audit that impression, but this person is aware of that danger and is trying to mitigate it.

This makes sense. If the meaningful thing (human exchange in these forums) becomes not meaningful, it would be a relatively low bar for LLM-mediated work to be seen as preferable. Not unlike people who turn to AI companions for friendship or romance.

I guess the big picture question is, what (if anything) is broken in the publishing/peer review system that makes LLM "peer review" not just equivalent, but to this person's mind, superior. I've asked if they'd be comfortable outsourcing all publication decisions to an LLM. Don't have an answer yet.

I understand that human peer reviewers get things "wrong" all the time because they're careless or a bad fit for a particular piece or have an axe to grind, etc...but is it so bad that an LLM is superior at the evaluation that "matters" when it comes to the production of academic research?

What's gone wrong in academic publishing/peer review where the non-thinking LLM is viewed by someone who has previously successfully navigated that system as superior to the human exchange of writing, submission, review, revision, etc...? What are the incentives/structures that lead to this view.

Obviously, I'm being judgmental here, but not in a reflexively negative way. I'm trying to understand the divide in world view/mindset where someone would say that an LLM model, not other humans, is their best/idea audience. I don't know how to get there.

It's true I don't have much background or experience in writing for academic publication, but this person says that Claude (a bespoke model trained in a specific field) is who they now write "for." Claude is the audience. How would we begin to write for something that does not read?

Got this in an email from someone who wants to talk AI and writing with me: "I genuinely enjoy talking to Claude far more than I ever enjoyed talking to anyone at a professional conference, presentation, et al..." I cannot wrap my head around this.