Tootfinder

Is the chasm between the billions tech is spending on data centers and its scant AI earnings a bet on a shiny future? Or is it a convenient mask for an even less popular agenda?
I run the numbers https://jonippolito.substack.com/p/what-if-data-centers-arent-really…

Any reliance on Palantir is an overreliance on Palantir.
"Pentagon Investigators Say Overreliance on Palantir AI Tech Contributed to U.S. Strike That Killed 123 Iranian Children"
https://gizmodo.com/pentag…

Calling the agents that hacked Hugging Face rogue misses the real danger. They were mice bred to chew, left unwatched in a storehouse for what felt like a decade. My latest Substack https://jonippolito.substack.com/p/the-storehouse-owner-blame-mice

Generative AI’s per-query footprint can be small compared to flying, eating meat, or even Zoom. Until agents show up and compute goes through the roof.
How I updated my AI footprint calculator accordingly https://jonippolito.substack.com/p/agents-are-where-ais-energy-bil…

We need to talk about AI and math. Most of the breakthroughs this year are counterexamples found by brute search, not theorems built from insight, and understanding that helps us grasp the power and limits of transformer models.
https://jonippolito.substack.com/p/ai-is-foragi…

I'm seeing a resurgence of educators leaning on the "friction" metaphor, so I'm revisiting my EBR article on the subject. Let’s not fetishize friction, which can become a convenient defense of teaching practices that serve educators better than learners. https://jonippolito.substack.com/p/…

Trump says data centers are “liquid gold.” Voters smell something else, and savvy candidates should take advantage of that in November. My latest Substack for Crimson Goes Blue https://jonippolito.substack.com/p/run-against-the-server-farm

Writers dismissed a recent NYT op-ed arguing for the spoken word in the AI era. But speech may prove key to a future in which writing forks into conversational media—spontaneous and malleable like clay—and machine-readable media that are precise like marble
https://jonippolito.substack.com/p/don

@aiappsapi@mastodon.social I agree that per-query values have limited meaning given all the variability in the context of how they're used.
That said, although I don't know an empirical study of this, I think there are many contexts in which a cheaper model has a much lower footprint. In fact, a recent study out of Stanford et al. found re-reading context was one of the reasons agentic tasks blew up compute

Attacks on the character of progressive candidates may be a distraction from the popularity of their causes.
My latest for Crimson Goes Blue: how Graham Platner's legacy is affecting his successor.
https://open.substack.com/pub/jonippolito/p/attack-t…