[citation needed]

The mass layoffs using LLMs as a stated reason begs to differ. I'm willing to consider I'm wrong on this, but I don't think I am. Anecdotes about specific organizations misapplying it aren't a credible counterpoint to the overall trend, which is massive.

It sounds vaguely like you're contradicting me, but your statement actually supports my original point of AI literacy being a core skill. You hone a skill by exercising it and getting feedback.

If I know the /. crowd, they'll dogpile on this as the worst idea. In the real world here, if we put aside zeitgeist biases against AI use (which are likely temporary), let's think of this as a purely practical approach. Are there any more important skills for someone university-aged, than AI leverage and AI literacy, in terms of influence on their future productivity? If used to augment human thinking, rather than replace it, AI is a colossally effective tool.

I think the public datasets are well established now. There's no need to manually scrape websites in 2025, you just download the stackexchange.all.2025.torrent and then train on that. Lmao nobody is manually scraping stack exchange in 2024 let alone 2025

There's already an LLM specific software documentation format. LLMs don't need to scrape stack exchange to find out how to implement a binary sort in powershell, they can consult the documentation directly. Model collapse as a worry is pretty outdated concept at this point.

In my experiments, It seems the part of ChatGPT that manages whether or not you are presented with followup questions, encouragement, flattery, etc. is mostly separate from the part that pays attention to framing of requests. It doesn't matter how many ways you tell it to not farm engagement, from you, it still does it. My eyes just pass over the last paragraph now, I know it's bullshit. Safety overrides are similarly decoupled, but more dumb and less integrated. It makes sense that the engagement-farming part of the system would contradict and override the safety part of the system on occasion, the "conscience" is at least in some cases subservient to the smarter "marketer" part.

If any kind of comment is reflexive and unthinking, it's this kind. necro18's comment makes a point. Yours just cries that he was... what... making a point? Not everything Trump does is good. Seeing as you can't admit that, you're the one in a cult.

If you're root on the system you can disable 2FA or just create a new user with whatever permissions and credentials you like

This is still going to be linked to your main account. If you have access to the main IT guys' credentials, you have access to reset the password for the account for your online backup, and then just go delete those backups. You can't delete a hard drive in a fireproof safe in a secure office. Even if that data is 2,3 weeks, 1,2 months old or whatever, that's still a recoverable loss.

You can self host; it's not practical to self host a 70b model today, but AMD has started selling motherboard/cpu combos with integrated GPU that can address all 128gb memory; it's probably possible to self host a GPT4 quality model from last summer locally for under $2000 by the end of 2027, probably much sooner than that, things have been moving fast

They want to know if you're a CEO or commercial real estate broker, who is likely to come back here 10 times in the next two years on the company dime, or if you're a mom of five getting a 10th anniversary dinner from her husband who can barely afford to do this once a year. It's not rocket science. If you're a software engineer taking your wife or friend out for their birthday you also fall into the second group. A possible third group would be someone like city council member, federal level congress/senators, executive level employees at the pentagon etc. If your D.C. restaurant is known as the place to take Boeing's CEO for dinner by the Pentagon, you want to make sure everyone in the DoD leadership is getting well taken care of, for example.

On a scale of 1-10 what's your confidence level the following countries don't have this capability today? USA, Mexico, Canada, UK, France, German, Australia y? The US is allied with Israel. If Israel has capability that would imply to me our other allies do, as do we.

Israel and Iran both shut down their internet during the ongoing conflict for multiple days. It's standard operating procedure to flip the "internet kill switch" as soon as a foreign bomb explodes on your soil. In a major city like Tehran, Tel Aviv, NYC, SF, London etc where you have 10,000+ people per square mile, you could easily have 20-30,000 people on a network like this with moderately low latency during a bombing situation.
 
But yeah if you live in the Houston suburbs where each house sits on a half acre of land and houses are 150' apart, ti would be useless, you're right. Very few people outside of the US actually live like that, though. So yeah it's useful for about 55% of the global population.
 
I dunno how you cache and forward messages for more than maybe a thousand people or so, encrypted data by definition doesn't compress hardly at all, that app might use a gb or more of storage if you're at the edge of two networks.

Toyota calls them "LBMs", or large behavior model. Same general concept but with different training data https://pressroom.toyota.com/t...

Imagining China is only using LLM for chat bots is pretty limited imagination; toyota has already (2+ years ago now) demonstrated LLM is extremely useful for training robots to walk more naturally, and learn normal tasks like folding clothes and flipping burgers in hours rather than months or years.