FaceDeer

“Clinicallydepressedpoochie” is in the running.

Elon Musk decided they absolutely would not use lidar, years ago when lidar was expensive enough that a decision like that made economic sense to at least try making work. Nowadays lidar is a lot cheaper but for whatever reason Musk has drawn a line in the sand and refuses to back down on it.

Unlike many people online these days I don’t believe that Musk is some kind of sheer-luck bought-his-way-into-success grifter, he has been genuinely involved in many of the decisions that made his companies grow. But this is one of the downsides of that (Cybertruck is another). He’s forced through ideas that turned out to be amazing, but he’s also forced through ideas that sucked. He seems to be increasingly having trouble distinguishing them.

They believe that Jesus will come back when Armageddon happens, and Armageddon is supposed to happen in Israel.

They can make whatever excuses they want, as long as the result is the same. They need to narrow their focus to serve the browser and this is a step in that direction.

Pocket used to be a separate extension. Firefox bought it out and integrated it. It could go back to being a separate extension again, I suppose.

Personally, I think this is a good thing. Firefox is looking at some rocky financial times ahead and I’m happy to see Mozilla narrowing its focus to making just a damn browser please. As long as the browser has a good solid framework for extensions then all this extra functionality can be provided by others just as easily.

You’re still setting a high standard here. What counts as a “well trained” human and how many SO commenters count as that? Also “easier to teach” is complicated. It takes decades for a human to become well trained, an LLM can be trained in weeks. And an individual computer that’ll be running the LLM is “trained” in minutes, it just needs to load the model into memory. Once you have an LLM you can run as many instances of it as you want to spend money on.

There’s no guarantee LLM will get reliably better at everything

Never said they would. I said they’re as bad as they’re ever going to be, which allows for the possibility that they don’t get any better.

Even if they don’t, though, they’re still good enough to have killed Stack Overflow.

It still makes some mistakes today that it did when introduced and nobody knows how to fix that yet

And humans also make mistakes. Do we know how to fix that yet?

If they aren’t comfortable with their Discord messages being public, perhaps they shouldn’t have posted those messages in a public forum that the public can access.

Good thing human teachers never have hidden biases.

How does this play out when you hold a human contributor to the same standards? They also often fail to summarize information accurately or bring up the wrong thing. Lots of answers on Stack Overflow are just plain wrong, or focus on the wrong thing, or don’t reference the correct sources (when they reference anything at all). The most common criticism of Stack Overflow I’m seeing is how its human contributors direct people to other threads and declare that the question is “already answered” there when it isn’t really.

LLMs can do a decent job. And right now they are as bad as they’re ever going to be.

That’s the neat thing, you don’t.

LLM training is primarily about getting the LLM to understand concepts. When you need it to be factual, or are working with it to solve novel problems, you can put a bunch of relevant information into the LLM’s context and it can use that even if it wasn’t explicitly trained on it. It’s called RAG, retrieval-augmented generation. Most of the general-purpose LLMs on the net these days do that, when you ask Copilot or Gemini about stuff it’ll often have footnotes in the response that point to the stuff that it searched up in the background and used as context.

So for a future Stack Overflow LLM replacement, I’d expect the LLM to be backed up by being able to search through relevant documentation and source code.

What comes to mind when people talk about how mediocre AI is.

As I said above:

mobs of angry people ignorant of both the technical details and legal issues involved in it.

Emphasis added.

They do not “steal” anything when they train an AI off of something. They don’t even violate copyright when they train an AI off of something, which is what I assume you actually meant when you sloppily and emotively used the word “steal.”

In order to violate copyright you need to distribute a copy of something. Training isn’t doing that. Models don’t “contain” the training material, and neither do the outputs they produce (unless you try really hard to get it to match something specific, in which case you might as well accuse a photocopier manufacturer of being a thief).

Training an AI model involves analyzing information. People are free to analyze information using whatever tools they want to. There is no legal restriction that an author can apply to prevent their work from being analyzed. Similarly, “style” cannot be copyrighted.

A world in which a copyright holder could prohibit you from analyzing their work, or could prohibit you from learning and mimicking their style, would be nothing short of a hellish corporate dystopia. I would say it baffles me how many people are clamoring for this supposedly in the name of “the little guy”, but sadly, it doesn’t. I know how people can be selfish and short-sighted, imagining that they’re owed for their hard work of shitposting on social media (that they did at the time for free and for fun) now that someone else is making money off of it. There are a bunch of lawsuits currently churning through courts in various jurisdictions claiming otherwise, but let us hope that they all get thrown out like the garbage they are because the implications of them succeeding are terrible.

The world is not all about money. Art is not all about money. It’s disappointing how quickly and easily masses of people started calling for their rights to be taken away in exchange for the sliver of a fraction of a penny that they think they can now somehow extract. The offense they claim to feel over someone else making something valuable out of something that is free. How dare they.

And don’t even get me started about the performative environmental ignorance around the “they’re disintegrating all the water!” And “each image generation could power billions of homes!” Nonsense.

It’s a great new technology that unfortunately has become the subject of baying mobs of angry people ignorant of both the technical details and legal issues involved in it.

It has drawn some unwarranted hype, sure. It’s also drawn unwarranted hate. The common refrain of “it’s stealing from artists!” Is particularly annoying, it’s just another verse in the never-ending march to further monetize and control every possible scrap of peoples’ thoughts and ideas.

I’m eager to see all the new applications for it unfold, and I hope that the people demanding it to be restricted with draconian new varieties of intellectual property law or to be solely under the control of gigantic megacorporations won’t prevail (these two groups are the same group of people, they often don’t realize this).

This is an area where synthetic data can be useful. For example, you could scrape the documentation and source code for a Python library and then use an existing LLM to generate questions and answers about the content to train future coding assistants on. As long as the training data gets well curated for quality it’s perfectly useful for this kind of thing, no need for an actual forum.

AI companies have a lot of clever people working for them, they’re aware of these problems.

I’m a fan of the Machete Order.

There may be some spoilers in that blog post, it’s been a while since I read it, so here it is in summary:

• A New Hope (4)
• Empire Strikes Back (5)
• Attack of the Clones (2)
• Revenge of the Sith (3)
• Return of the Jedi (5)

Phantom Menace is omitted because it’s the weakest of the prequel trilogy and everything that happens in it is summarized at the beginning of Attack of the Clones anyway. If you want to be a completionist then watch it between Empire Strikes Back and Attack of the Clones.

There’s good reasons for following this order, but it’s hard to describe them without spoiling anything. Basically, Lucas assumed you’d watched the original trilogy when he made the prequels, so it’s got a bunch of spoilers in it that the Machete Order preserves quite nicely.

That’s why I blocked her.

There will eventually be enough public domain content that AI will be at the quality it is today with public materials alone.

So, AI will always be ~95 years behind the times?

Except the AIs produced by Disney et al, of course. And those produced by Chinese companies with the CCP stamp of approval. They’ll be up to date.

Many people with positive sentiments towards AI also want that.

If you think death is the answer the polite thing is to not force everyone to go along with you.

So they’re still feeding LLMs their own slop, got it.

No, you don’t “got it.” You’re clinging hard to an inaccurate understanding of how LLM training works because you really want it to work that way, because you think it means that LLMs are “doomed” somehow.

It’s not the case. The curation and synthetic data generation steps don’t work the way you appear to think they work. Curation of training data has nothing to do with Yahoo’s directories. I have no idea why you would think that’s a bad thing even if it was like that, aside from the notion that “Yahoo failed therefore if LLM trainers are doing something similar to Yahoo then they will also fail.”

I mean that they’re discontinuing search engines in favour of LLM generated slop.

No they’re not. Bing is discontinuing an API for their search engine, but Copilot still uses it under the hood. Go ahead and ask Copilot to tell you about something, it’ll have footnotes linking to other websites showing the search results it’s summarizing. Similarly with Google, you say it yourself right here that their search results have AI summaries in them.

No there’s not, that’s not how LLMs work, you have to retrain the whole model to get any new patterns into it.

The problem with your understanding of this situation is that Google’s search summary is not solely from the LLM. What happens is Google does the search, finds the relevant pages, then puts the content of those pages into their LLM’s context and asks the LLM to create a summary of that information relevant to the search that was used to find it. So the LLM doesn’t actually need to have that information trained into it, it’s provided as part of the context of the prompt,

You can experiment a bit with this yourself if you want. Google has a service called NotebookLM, https://notebooklm.google.com/, where you can upload a document and then ask an LLM questions about the documents’ contents. Go ahead and upload something that hasn’t been in any LLM training sets and ask it some questions. Not only will it give you answers, it’ll include links that point to the sections of the source documents where it got those answers from.