nphardon

Im trying to disentangle from the llms and get back to writing code and thinking for myself again, and so i am finding myself back on stack overflow and its like a ghost town.

This is so good. Here's one stand-out line (of many) to me:

> And as for whether it replaces or disrupts anything good that already exists, I will only say that I have watched it, in a very short time, begin to replace the patient work of learning, the slow formation of judgment, the friendship of teacher and student, the labor by which a writer finds out what she thinks. These are things that, once lost, are not quickly recovered.

theory: there are two groups that a lot of hackernews commenters fall into. And these two groups clash on every Ai post every day. One camp only uses Ai to write code and thinks of them strictly as LLMs with a clearly understood design and purpose but that show some surprising behavior, they've watched youtube videos that show how everything ties together computationally and mathematically, from neural nets to RL, etc. and dont see them as capital A, Ai. It's another tool in the toolbox, and thinking about them in the way you describe, as something closer to capital Ai would cause severe cognitive dissonance, akin to the idea that a hammer could hop out of the toolbox and start building a house. Even though what you've done is just take one small step of critical thinking, it would be a leap to the other camp members, who are themselves, quite technical thinkers.

You're spot on. I have a young kid this is exactly my experience.

Essentially, I try to catch movies in the theatre. I don't think I was watching a ton of streaming only stuff, because I haven't been aware of myself missing out on anything, but maybe that just wasn't part of my routine. But we did drop from like, disney, hbo, netflix, to nothing, and I just dont miss anything about them. Oddly enough, the other members of my fam who were hesitant also are happy about the transition at this point.

I quit all streaming services about 6 months ago and am strictly on CD's DVD's and records. I recommend it. The library has a ton of stuff, and if enough demands comes back maybe dvd shops will open back up. Its really nice, no ads, no data mining, no supporting awful streaming services, no subs, finite choice. Again, I recommend it.

that's all i see as well: self surveillance. when I read other comments like "the family in this ad is exactly like mine" or "oh if it had app store access then itd be great" it makes me think of that scott alexander essay "i can tolerate anything except the outgroup" where he makes the analogy of living side by side with people who exist in entirely separate and impenetrable social bubbles, like light and dark matter.

Is a special hardware setup required to run open weight models? It's not clear to me from the article. Also, what does paying for inference mean? I thought the inference step happened during the training of the models, and that by the time I am interfacing with an LLM all the inference has been long done.

Makes me wonder what is going on in the Dilley Detention Center (and others) and if we will find out of atrocities only after they are shut down and the people detained are released back into the population.

A full translation appears in the comment section.

Does Tao profit off Ai labs?

"OpenAi *virus* found on Wikimedia projects" ?

I'm wondering how this plays out in the interview process though, just a year or two ago it was still leetcode and hours upon hours of technical interview. Has that changed in the context of LLMs? If you were hiring, how would you go about it? Does knwoing python really matter anymore? In my shop it 100% does not, and I'm guessing that in the 75% of posts in Who's Hiring that are for Ai startups it also does not, just curious to find out if thats true.