| Comics | http://www.threepanelsoul.com |
Matt Boyd
@3psboyd
- 999 Followers
- 85 Following
- 4.2K Posts
Normal man, normal posts.
Putting aside the monetization, hot-swapping between models sounds like it would shock normies who have anthropomorphized LLMs into having something you could call a coherent thought. Which is probably almost all of them, because that's how they've been sold since day 1. https://typo.social/@colinmford/116613158941426952
Colin M. Ford (@[email protected])
“One approach, proposed by Google Research, is what you might call a “token auction.” In this model, advertisers don’t buy ad slots on a page. Instead, they bid, token by token, on the actual text the model generates. Each advertiser brings their own LLM, and an auction mechanism decides whose model gets to influence the next word. The output is a weighted blend of competing interests, shaped by who’s willing to pay more.”
Adrianna TanMore importantly I feel like most criticism of the H-1B program just.. doesn’t really understand how skilled visas work.
Treating skilled legal immigrants like shit and giving them no rights is kind of the hallmark of almost all major immigration policies, not just the American one.
Proposing that we get rid of it because you think it is exploitative is as bad as the ‘I am all for open borders but I have no real suggestion for how western governments can right now stop treating brown people like shit’ problem.
The no 1 thing to fix in US immigration is the country caps.
Because of racism, people born in India and China etc face years of long waits (ranging from a couple of years to hundreds of years)
All of the applicants born in big countries have to compete for the same no of green cards as people from small countries
This is the specific mechanism that companies use to trap H-1B workers in subservience, not the H-1B program itself. There is little to no easy path to permanent residency because of the country caps.
This part of the training does start with "I know we're asking everyone to use AI in their work, but ..."
Surprisingly, I think the ubiquity of LLMs might be improving the situation, since it makes specific interviewer followup questions more necessary.
I kept saying during that whole fiasco that I felt like I was playing a game where nobody told me the rules, and now, here they are.
There's a lot of things done to ostensibly reduce bias (aka the individual judgement of the reviewer) that demands you answer questions in just the right way that they can repeat to show you meet the requirements of the job and the (sigh...) values of the company.
Receiving interviewer training for my current and realizing how much of my 2.5 years of unemployment was down to not doing parts of this performance correctly.
Jason LefkowitzIf you ask the AI boosters, they will tell you that large language models write code at the level of an enthusiastic junior programmer — competent but not particularly wise.
Meanwhile in Washington, the message policymakers are getting is that AI has “reached a level of coding proficiency surpassing all but the most skilled human minds”
https://www.politico.com/news/2026/05/16/sean-cairncross-ai-mythos-expertise-00925336
