Seth Largo retweeted
Replying to @moultano
Reading some histories of the dust bowl. All the cowboys are like "Idiots, only this native grass will grow here, do not rip it up," and the farmers are like, "plow go brrrrr"
Found another guy confused about why puzzler enjoyers wouldn't enjoy a machine that just puts the puzzle together for you.
Good Cliff's Notes version of "Day of the Locust," which literally describes Hollywood sets as a "dead sea" of Western Civilization's washed up dregs.
I am advisor for a Film Club and one of the kids went all out decorating the theater lobby for our annual horror double feature.
AI is amazing, don't get me wrong, but the metaphor should still be "Excel spreadsheet."
Replying to @elidourado
Per this later part in the article, that might be Cade Metz' garbled understanding of something that @soniajoseph_ told him about lower dimensional manifolds
Seth Largo retweeted
Kinda miss guerrilla advertising of the 00s. In hindsight living inside our own individually sorted and prioritized algorithms is much more sinister than the Cartoon Network catching charges for ad campaigns run like terrorism cells.
I taught my 6 year old to play the Charlie Brown theme song on the piano in a week, but it's a parlor trick. Good luck to the Alpha people, but unless the kids develop a genuine interest in *continuing* this level of quick-hit training into the teen years, it fades very quickly. (My kid already forgot how to play the Charlie Brown theme song after another week.)
I am much more excited about what @NielsHoven is doing with early reading, because reading is in comparison a low-effort and enjoyable thing to commit to once you "get it." If we pop out more kids, I will subscribe them to Mentava.
I knew that last-minute "replicant uprising" thing in the sequel would lead to such slop. "Blade Runner" and most of "Blade Runner 2049" operate on a Miltonian creator/created dynamic, not a master/slave dynamic. Degradation of the franchise.
Every diaper they changed is subtracted from your account. Time to repay.
Seth Largo retweeted
I’ve spent three years engaged in long-running and surprisingly civil arguments on here over whether AI is a cause of civilisational collapse or merely a symptom.
My view is that, just as AI prose and art could only possibly flourish in the context of a mass culture that was completely sclerotic and reduced to smashing tropes together, the sort of person who’d get an LLM to tell them three nice things about their own child or make a podcast about what their child had been up to to listen to on the way to work was fundamentally inhuman long before AI came along.
Nothing fails the Bechdel Test harder than all-female podcasts. Even the true crime ones.
Something I’ve noticed about “femosphere” podcasts is that literally all they talk about is men and relationships.
Joe Rogan, the biggest male coded podcast, is known for talking about drugs and wild animals and random shit-talk with the odd bit of spicy political stuff.
Women are ingesting absolute poison on a level equivalent with the very worst of the “Manosphere” and the asymmetry in the way each is received in polite society is so vast to be a chasm.
Both of these are 100% making you stupider.
This is basically true. Two experiments.
Experiment 1:
> Dictate a draft in GPT with some clear words, some misspelled ones, and some xxx, where xxx are few placeholder words for AI to fill in.
> Ask AI to only fix spelling and fill in xxx with few words only (ideally your dictation will only leave little room)
> Check Pangram
It'll almost surely be 100% human.
Experiment 2:
> Ask GPT to write few sentences
> Lightly paraphrase it while keeping the flow/sentence structure the same
> Check Pangram
It'll almost surely be 100% AI.
Seth Largo retweeted
Striking detail in the big NYT piece on Taiwanese families adjusting to working at the huge new TSMC chip plant in Arizona: their surprise at the lack of rigor in US schools.
Seth Largo retweeted
Replying to @tracewoodgrains
Grammarly is terrible... basically a cheating app.
It not only nudges students into avoiding doing their own writing and thinking, but now has integrated AI detection and humanizer—tools that serve no purpose other than to deceive readers.





























