TehPers

We’re comparing this to the observable behaviors of humans, though. Humanity engages in unsupervised learning without a scoring function (let alone a differentiable one) and, as a result, can not only train on the fly, but can also find its own ways to self-evaluate its learning. Humans also have hormones which change how the brain works even further.

This is also all disregarding the ability for the human mind to become faster and more accurate on recall when it has more knowledge to search in a field. LLMs can’t recall very effectively, instead becoming less effective the more information it needs to search.

Llm’s that only have one glyph per token can easily count this

[Citation needed]

There’s no difference between counting tokens containing the letter “r” and tokens solely representing the letter “r”. To begin with, this assumes the way we encode words in our brain is by tokenizing each letter, which we pretty much know to be false (we can process words without seeing evry ltr of th wrd and even think about them without it being mentioned).

yeah we train in bursts because it’s computationally cheaper, this hardly means anything important.

No, this is not how neural networks work at all.

We train in bursts because we need a ground truth/scoring function to guide the learning, which means we need a way to guide the model in the right direction. Even with unsupervised learning, the model has a way to determine whether it’s going in the right direction, like with GANs.

Neural networks are also incapable of synthesizing completely new skills from its own weights. Humans have been doing that for the entirety of human history.

What a dumb question. “Cognition” is an abstract concept that just leads to debate over its definition, except that it’s commonly understood that humans have cognition.

This entire interview also seems predicated on the belief that neural networks work fundamentally the same as human brains, which is an oversimplification of brains. To begin with, neural networks cannot train themselves at inference time (context windows aren’t training). Also, humanity has barely a concept of how the brain works and regularly learns new things about it. What we do know is that they are more complex than a bunch of connected neurons, and that we have no way of modelling something we don’t yet fully understand.

So do neural networks have cognition? If they did, then we’ve had cognitive AIs for longer than I’ve been alive.

The rest of the interview is imaginative fiction based on what they think the model is doing predicated on what Anthropic likes to advertise their models do.

As basic evidence, if I tell you that the word “strawberry” has 3 "r"s, you can remember that. You can recall that no matter how many conversations we have in between when I told you that and when I asked you again.

If I teach you how to count how many "r"s there are in the word “strawberry”, you learn a skill. In the future, if I ask how many "r"s there are in the word “strawberry”, you can count them using your new skill.

If I teach a LLM how many there are or how to count the letters, it can only use that skill or knowlege if it can effectively search for it. Its ability to execute skills (which, to begin with, isn’t it executing the skills but it asking something else to do them) is diminished the more skills it gains access to, and its “knowlege” (aka long-term storage) becomes diluted when it has access to more of it.

This is blatantly untrue of humans, who can process knowlege and skills they learn, efficiently and quickly search them, and actually get faster and more accurate as they develop related skills and learn related knowlege.

you better license it AGPL 3.0 or you’re part of the problem.

They ignore copyright. License it however you want, but they’ll still train off your content.

You can block Facebook ads by blocking the Facebook domains. This is what I do, anyway.

Even if you get through the ads, there’s nothing of value there. At best, you’re maintaining superficial relationships while wading through AI slop and ragebait. If you haven’t actually talked to the person in years, consider whether there’s even a real connection there or if you’re just inflating each other’s egos by pressing a like button every couple weeks.

Sorry, forgot about a few. Sam Altman, Sundar Pichai… Honestly, it’d probably be easier to list the white cishet billionaires in tech than it would be to list the exceptions. There’s tons of them outside of tech, probably the majority tbh, but within tech it’s far too diverse.

Don’t forget about the other big players! Jensen Hwang, Lisa Su, Lip-Bu Tan, Satya Nadella, plus the memory cartel (Samsung, SK Hynix, etc.)

This particular war is a class war.

If your life total is 20 or more and you control an effect that states “you can’t lose the game”, then an infinite loop is formed. If neither player wants to end the loop by removing one of these two cards from the battlefield, the game ends in a draw.

It’s a Self-Destruct Button!

We are somewhat similar to contractors and directly help our customers write their code. Usually this means bringing in some kind of starting template, and it’s easier to do this with a permissively licensed template than make them sign another legal agreement.

They own the code once we’re done, and they don’t need to make it open source (companies naturally don’t like to open source their products). Our templates are open source and well-documented, though.

We’ve managed to convince our workplace to allow us to use MIT on a few projects. We mainly share that code directly with customers, and it’s easier to do that with permissively licensed code, plus we can make it public for an extra win (both at work and for the community).

There are definitely good uses for permissive licenses. In the same way that you should consider not using one, you should also consider whether one meets your specific needs.

“Such speech may be offensive to many, but it is not Meta’s role to police offensiveness,” the company added.

I don’t really know what else to write here because the quote is already enough. Anyway, if anyone still had any doubts about Meta for some reason, there you go.

What’s yours? Pay Anthropic $2k/mo to generate fake C&D letters and send them out to random businesses?

Unless you have a plan that doesn’t rely on the service-based models, then I don’t see where you’re going with this. Sure, you can use self-hosted models, assuming you’re fine paying for the GPUs (Jensen Huang? Lisa Su? Lip-Bu Tan if you’re feeling special?) and power to run them of course. But you’ll be “behind” the big cloud-based models, endlessly chasing after them.

Or, hear me out, don’t use them and do it all yourself and you won’t have to pay these companies and make their execs richer.

With a saguaro.

I call it a car subscription, and it’s for people who make poor financial decisions. It’s better to buy a car you can afford, pay it off, and own it until it becomes unsustainable to keep it running.

Unfortunately, any tools a screen reader can use to read the page are tools a LLM can use. Without challenge-based solutions like Anubis, I’m not sure what can really be done that isn’t IP whack-a-mole.

If it comes down to losing accessibility or losing the website due to scrapers obliterating the hosting costs, then I think most website admins would rather just put a “use OCR” message at the top of their skip nav.

As much as I hate using JS for this, I wonder if there’s a JS-based solution that gets the best of both worlds while killing off scraping. Serve this obfuscated trash initially, then swap it out after completing some challenge sometime after FCP.

Well that was a dodged bullet. I’m glad we settled on a different car when we got ours recently.

Still, I think pretty much the whole industry is moving towards SaaS bullshit. Everything from subscription heated seats to ads in the infotainment. Hell, even in our new car, I can’t start the car without it begging me to connect to Amazon and setup Alexa. It’s making me miss my older car’s basic dash and infotainment honestly, but getting Android Auto/Car Play and a suspension that doesn’t feel like the Tower of Terror (or whatever it’s called now) whenever we go over a speed bump was life changing.

This shit should be illegal. At the very least, I hope that ad isn’t playing while driving.

GitHub more like ShitHub

Codeberg is free. If you need uptime, self host it.

This is undecided by courts.

From what I’ve found from some simple research, it’s possible that if it’s reproducing the same information as another user that the website is protected, possibly even if the LLM modified the exact presentation of that information.

If it’s creating new claims, then I think it’d be likely that whoever is running the service would be liable. Who knows though, since they might just claim that the new claims are from its training data and therefore they are protected.

In an ideal world, we’d just follow Germany’s example and hold the website liable for any new content generated by a LLM. Ideally, if you make shit up, you made the claim, and if you’re parroting a user, the user made the claim.

I’d imagine due to current circumstances, the US would want to shield AI companies though because politicians want money and the SCOTUS is useless.

This reminds me of Rossman’s praise for the AI overviews when it helpfully informed everyone that he has paid sponsors and his friend was a 9/11 terrorist. Except it wasn’t praise.

The AI summary at the top is like asking a college freshman to summarize the latest research in a field. It’s accurate enough to be dangerous, but not accurate enough to be reliable.

This. A charge isn’t a conviction, though.

Plausable deniability might help you here as long as you can think of a valid reason why you had that code on the note. It could be as simple as “this is the code I needed to remember for this one game” or something too. As long as you don’t tell them that the note contains the code to your phone, you can pretend the note contains a code of some other value instead later on and let them shoot themselves in the foot.

As a general rule, most lawyers would tell you if you are under arrest to say nothing more than what you are required to by law. Since this is at a border, it gets weird, but that’s what makes this case unusual as well. Under normal circumstances, they might seize the phone as evidence, but saying nothing might result in them wiping the device themselves before they ever get a warrant for the passcode.