TehPers @ TehPers @beehaw.org

Posts

0

Comments

681

Joined

2 yr. ago

The sooner, the better. It's so painful when I use Google these days. Why is it that smaller people can do seemingly obvious features like custom user-controlled site rankings, but the big players are completely incapable of that?

Not just marketing, that's the term it's always been called. Plug a bunch of parameters into a non-deterministic model and you've got an AI, at least by what seems to be the common definition of the term.

This is a guess since I'm not a lawyer, but since users license their content to Twitter when posting it, Bright Data might have to prove fair use. I don't think that question has been answered yet in relation to AI model training, but search engines have been doing this for decades for what it's worth, so I don't know.

Do they give up the copyright, or license it to the website? They still created the content, and I don't have a Twitter account, but after briefly reading the ToS, it says they license it to Twitter (which is pretty standard from the other services I use that I've read the ToS for).

I think Steve makes a great case against Asus in the ROG vs ROC debate. In fact, it raises the question of who is stealing from who here. Clearly Asus took the ROG name from roc. They had to have started with the ROG acronym and went backwards because, let's be real, who would have thought "Republic of Gamers" was a good name if they weren't forced into using that acronym from the beginning? The ROC branding is clearly a reference to the same bird, just taken more literally.

The Win11 ads has me looking at good Linux distros now. I didn't buy a (discounted) $200 license so i can look at ads.

I managed to get ollama running through Docker easily. It's by far the least painful of the options I tried, and I just make requests to the API it exposes. You can also give it GPU resources through Docker if you want to, and there's a CLI tool for a quick chat interface if you want to play with that. I can get LLAMA 3 (8B) running on my 3070 without issues.

Training a LLM is very difficult and expensive. I don't think it's a good place for anyone to start. Many of the popular models (LLAMA, GPT, etc) are astronomically expensive to train and require and ungodly number of resources.

Genuine question, but is this bill even constitutional? Article says no, but doesn't explain why. Still, a similar bill could easily say that people are not allowed to join any other kind of organization (except presumably a religious one), right?

If the ICC has no jurisdiction, why are they all worried about the investigation? Let them do their thing, then just nod and say "OK" afterwards.

Also, I love how the US refuses to allow itself or certain close allies to be held responsible for war crimes. At least we can criticize our own government though, could be a lot worse I suppose. Still, maybe don't do those? That'd be great.

I've used GitLab and Azure DevOps professionally, but there are a lot of services out there which host Git repositories. GitLab can also be self-hosted which is nice. They all fundamentally work the same though from my experience - code viewer, issue tracker, pull requests, some way of doing CI/CD, and various collaborative and documentation features (wikis, discussion areas, permission management, etc).

It may be good to understand also where the separation lies between features that are part of Git vs those which are part of the service you're using (like GitHub). For example, branches are Git, while pull requests and wikis are GitHub.

You're right about that. She's a real catch.

"Thieves break into a car in SF" is somehow less surprising than "the grass is green in SF".

Speaking from experience, in the past year, I've used 3 different hosting providers for git repositories at work. Only one of them is GitHub. It's good to keep your options open - git isn't locked to any particular provider, after all.

I've been writing Lua off and on for probably close to a decade, and I can't remember the lack of a round function being an issue. I may have needed it at some point, but it's not exactly a complicated function to write up in a minute.

To me, the biggest appeal of Lua is actually the lack of an overbearing standard library. It has just enough to be usable as a scripting language within a larger application, and the larger application can always include its own helper library that gets loaded into the interpreter automatically on initialization. Feature-wise, there is enough to define your own OOP helpers (but no language built-in specific OOP stuff beyond metatables basically), there is enough to build your own async/await and generators using coroutines, etc.

Not having a huge built-in standard lib comes with the benefit of not needing to distribute a huge standard lib with your larger application.

You're not wrong. All religious extremists are, in fact, air breathers. They're also generally water drinkers.

As someone who doesn't drink, this has reinforced my preference to wait until things improve (hopefully they will eventually...) before considering a trip there.

See, it's democratic as long as people vote for their candidate of choice. Only one candidate? Well, you voted for them at least.

Also, I'm curious why this hasn't been an issue in the past. Is it Ohio changing the deadline, or the DNC being moved later in the year?

Maybe the difference can affect some people, but housing/rent prices in and around SF are astronomical, and I know of several people who can only afford where they live because of their commute. They'd love to bike to work, trust me.

Better bike infrastructure by itself won't solve the problem. It wouldn't hurt, but the core of the issue is the cost of living around where people work pushing them to live far away from their workplace. And no, the people I know can't just look for closer jobs. I've asked. The jobs don't pay as well.

They don't need to be in person, but there's no reason a fully-local team (like ours) couldn't have in person meetings now and then. The main appeal is lunch though (which you could do over a call as well, but if you can do it in person and everyone's fine with that, then why not I guess?).

I don't think there's anything wrong with fully remote, but I don't see any reason local teams can't have in person events if they want to either. We mostly meet in person for the meetings/events, otherwise it's "work where you want" basically.

Bikes, unfortunately, won't help many people with their commutes. The drive alone can be over and hour, often several hours. I'm sure some people could bike that, but most people would rather not do that twice a day.

Better public transportation as a whole could be helpful though. BART has been improving over time, but it's hard to say that it's enough (or that it ever will be at this rate).

Wow, imagine listening to your employees. I know there are people who like to work in the office, and there will always be a need for one to be there. I work better at home though. There are fewer distractions, I get my own office, and everything I need is within a short walking distance (food, water, etc). I don't need to chat with my coworkers every 5m. I don't need to see their faces to discuss system design, nor do I care much about how they're dressed or really at all what they look like. If my manager wants to see if I'm productive, he can see what tasks I've finished, which is what he should have been doing to begin with anyway (he was, at least in my case) because faking productivity is easy af.

We do meet in the office once or twice a week though to get through all our team meetings. I find that to be a good compromise, because those meetings are often better in-person since we get lunch together afterwards and such.

You don't need a LLM to see if the output was the exact, non-cyphered system prompt (you can do a simple text similarity check). For cyphers, you may be able to use the prompt/history embeddings to see how similar it is to a set of known kinds of attacks, but it probably won't be even close to perfect.