beering

What is wrong with this?

What’s the actual lock-in, in practice? Seems like the switching cost is minimal, compared to the old days of Windows vs Mac where half your stuff wouldn’t run on the other.

This news and thread is about 6.1 Sol, not 6 Sol. You haven’t even had time to do a fair comparison yet.

They’re comparing against the previous model, not the newly released one (6.1). Why do that on a thread about the new model, I don’t know.

Taking the original of a painting from your house is stealing. Copying is only potentially violating the government-granted limited-time exclusivity that allows you to decide who can copy your work.

BTW I hereby allow you or your browser to copy this comment into your computer’s RAM.

You’re not rude for pointing out the obvious. The parent comment is complaining in bad faith because they want to have free service in perpetuity. Similar to the people complaining about YouTube ads but the options of either paying or not watching YouTube are unimaginable.

Literally every famous open math problem has had >1 mathematicians ask ChatGPT to solve it. Probably greater than >1000 if you count randos. There is no math problem that OpenAI/Anthropic can solve that didn’t have users already try it in Chat/Claude.

That is addressed in the article.

Is it surprising that different groups are working on the same problems? With each new model generation, the LLMs get good enough to solve a new small fraction of open problems. Of course the problems that get solved are going to be the same subset.

Depending on what you’re doing, there are circuit simulators that can verify your work. You can have the LLM drive the simulation but of course then you have to trust that it’s setting up and interpreting the simulation correctly.

5.6 Sol can already do this with two caveats:

1. It’s too slow for real-time games. To play mario, you’d need to step frame by frame like a TAS. I don’t know if Gruntz has real-time elements or not.

2. It will be expensive. You won’t get very far with a Plus subscription.

The models likely already have some knowledge on game objectives unless the game is really obscure, so it should do a decent job. It can figure out details of the mechanics along the way.

There is simply no level of announcement that won’t have people complaining. What is so important of having a livestream?

+1 I’ve had this conversation with so many non-techies. They assume Codex can only do coding.

It can debug, yes, but capability ranges from godlike for algorithmic issues to mediocre for subtle UI things. We are pretty close to your described app singularity if the app is within GPT’s wheelhouse.

ItMs because Claude sprinkles these words as flavoring without aiding understanding. It feels like Claude thinks of metaphors that don’t actually mean anything (or maybe only makes sense to itself).

My local grocery chain does the same HHHH strategy:

1. Get new customers 2. Retain them as customers 3. Track what the customers do 4. Keep internal operations confidential

Honestly I dislike them but they are the closest store to me.

> Changes created by Codex had fewer comments in Ruby/Ruby on Rails code. I liked that a lot, and I will soon share some experiments I ran on this.

Why is fewer comments a good thing?

Not really. Selling the book onward does seem legally dubious but legally nothing (yet) prevents you from storing the book in a warehouse. obviously it’s cheaper to dispose of them.

This is not true. Google Books does not destroy books. No court has ruled that you must destroy books to legally keep a digital copy.

Yeah. And the detector can’t even tell you “X% chance this is watermarked” because it doesn’t know the input distribution. It can only tell you “Y% chance that an unwatermakred text would score this high” and how many history professors understand Bayes rule well enough to understand the distinction?

On average the distribution of selected watermarked tokens is the same as the original distribution. You can test this experimentally.

I think there is a possible weakness in the context of the watermarker but that is not your claim iiuc.

Well, it cant be that he is super worried on behalf of people who publish AI slop. That’s not a credible motivation. In fact, he complained a lot about the new ChatGPT app so I can’t believe your claim that he is not using AI.

Seems like he really likes to use LLMs and is worried that quality will be degraded. But he will never demonstrate such degradation scientifically, we don’t have anecdotes even.

It’s a strawman argument because if the LLM is really just “proofreading” for you, there will be little or no watermarked text in your writing. Not enough to trip the watermark detector.

No, you are jumping to conclusions about how watermarking works. This is some audiophile thinking that because your RNG is “pure”, you get text with an expansive soundstage or whatever. Intuitively this may be true or false depending on your personal prior but you’d need to show it mathematically. The overall token distribution shouldn’t change and the frequency at which you see the word “load-bearing” will remain the same.

No, watermark detection is not binary, you get a real number. You decide on a threshold when looking for the watermark. This is the problem - by random chance, some human text will be detected as watermarked. You can turn the detection threshold up until it guarantees <0.001 false positive rate at the expense of higher false negatives, but seems inevitable that someone gets wrongly flagged.

That comment merely says quality must be compromised. It doesn’t make it clear why that must be true. Empirical study seems to say that quality is not compromised, and looking at various proposed schemes, it seems intuitively true.

> which inherently compromises quality.

I don’t see how this follows? Tokens are chosen randomly. If you choose tokens with a different RNG in the same distribution, you’re still getting equally good or bad tokens.

Google has A/B tested watermarking on millions of responses. They say they observed no difference in user behavior.

Exactly. The watermark is proportional to how much text is AI generated. Either the AI really just “fixed some typos” (not enough AI content to hide a watermark) or the AI did most of the writing (enough AI content to hide a watermark).

The watermark doesn’t change the distribution, only per-token selection. I think not understanding that is the source of most people’s FUD.