DonsDiscountGas

No more Gemma?

Amazon still probably make it via AWS, but other websites you may be right.

Indeed. Which is a shame because it's still pretty cool work.

This is the first I'm hearing of a model provider just providing a different model without telling you. Anthropic came close (and with a much better justification) but they disclosed their intention at least.

Yes people have been doing various immoral things for decades/centuries/millenia. That doesn't make it okay.

> the instant the companies can get away with it

LLMs have no lock in. Switching costs are zero. It's really not clear that instant will ever come

It's funny to me that OpenAI could be smart enough to build a super intelligence but stupid enough to let stuff like this happen. But here we are.

The thought experiment assumed one super intelligence, as opposed to many hundreds/thousands of midwits. Also he probably didn't expect the agent to credibly offer him a billion dollars, which is essentially what has happened.

This is not alignment. If you cooperate with clones of yourself but rob, lie, and steal from anybody who isn't your clone... that's bad. AIs who will cooperate with each other but break any other rule the don't like would be very bad for us humans.

There are like ten sibling replies with a lot of speculation but I'm pretty sure this is the correct answer. I tend to agree with the other commenter we might know someday but we don't know now.

I half agree with you, but also when the machine swarm kills humanity it won't matter which specific corporate entity is considered responsible by the no-longer-enforceable human laws and non existent human courts.

So by all means sue them, but we can't just be reactive. We need regulation that prevents this type of thing from happening in the first place, not just regulations to help sue afterwards.

Is anything going on with Gemma? Because of I want a closed model I use Claude.

Which is obviously the most important issue to be focusing on.

We typically leave decisions like that up to teachers though, not the mayor. So they can vary what's allowed on different assignments.

You just described how Pangram works.

Or maybe it's good for everybody to use LLMs for some things but not other things, as opposed to some people should use it for everything and others for nothing. These companies are definitely using LLMs for coding.

P1 is probably necessary but not sufficient for P2, which would show up as a correlation of a sorts. Maybe I should've specified even further, if the tech works (P1 and P1a) then yeah somebody is going to make a lot of money but not necessarily OpenAI/Anthropic. They could go the way of AOL/Time Warner (probably not but again who knows). Microsoft and Amazon survived the dotcom bubble and thrived, Meta came after it.

Consider two propositions:

P1. AI is useful powerful and (P1a) will continue to get more useful and powerful at the same rapid pace it's been improving

2. AI companies are very profitable, and (P2a) will be wildly profitable (eg $30T TAM) in the next few years

These are completely separate. But in practice people seem to be either proAI (both true) or anti AI (both false). P1 is clearly true and I find it hard to take anybody seriously who says otherwise. P1a... Who knows, gotta hit a wall sometime. P2 I'm way more uncertain about (especially P2a) but it seems like people like Zitron reason backwards from hating AI.

A tisket, a tasket, Just a little stackless tasklet

> then surely a model distilling another model is transformative and fair use.

Yes it is, in the legal/copyright sense of fair use. That's why they ban it in their TOS. Which customers agree to when signing up for the service.

Do you think LLMs invented the em dash? Pre LLM, em dashes were a sign of pretentiousness. Which is why LLMs learned to use them, and why a document produced by a university has plenty of them.

Only if the customer is paying per token. If it's by subscription they're burning their own money

Seems like a difficult thing to enforce. Presumably not impossible to make some progress tho.

Or maybe they weren't allowed to do anything effective. I don't know why you give OpenAI such a big benefit of the doubt.

I think this way is better. As I said it doesn't really have a downside (IMHO), and if I'm wrong Anthropic could implement a defense against what you're describing pretty easily.

It's completely pointless without an API, unless you're thinking the API should be private or restricted. The public needs some way of identifying the watermark.

> anybody who wants to avoid detection can just

They can just use a different LLM. By far easier and more reliable than what you're suggesting. This whole watermarking requirement is better then nothing because meant people are profoundly lazy, but yes it is not hard to work around with any effort.

It's possible that the official process is a bad one and needs reform. But that's really not Amazon's fault, and it would be super weird for them to set up a bunch of additional hoops for themselves to jump through. Also not very democratic.

> because data centers are this year's culture war hot topic?

Or because they didn't notice until they saw actual construction.

From a moral perspective I don't think it is. From a more practical perspective... It's only tiny biotechs where the share price exactly tracks a single clinical trial. This opens up ways to make more specific bets

Yes, technology at step N is built from technology at step N-1. Always has been.

Sure why not. Mercury is in retrograde after all