[gleenn]

I think the fundamental problem is that training current SOTA AI models is very expensive. If a simple "classical" model can detect them, presumably at much lower algorithmic cost, then why wouldn't the model trainers use these same tools to feed back into their models to improve them at low cost to make them better? It's an arms race. Any cheap pattern can and presumably will be used to retrain if it becomes and effective way to catch AI.

[arjie]

It’s simply not a priority. The labs can do many things. Making text non-LLM is not really that useful. Analogous to Facebook not picking up the obvious $20 bill in front of them. It’s because they’ve got $100 bills at their feet they’re picking up.

[pixl97]

Not a priority currently. Selling services to spammers... I mean marketers is still big money and eventually someone will pick it up. If training costs ever drop, then it's one of the first things that will happen.

[weitendorf]

Because model providers are not optimizing for being indistinguishable from human text, and in fact, there is more value/demand in modeling a different distribution (ie an “agent” capable of producing vast amounts of concrete procedural/planning text interspersed) than there is in modeling the way humans write (ie GPT3).

Also you have to keep in mind that most AI companies are in fact trying to create and offer legitimate products and services to customers doing actually-useful work. They’re not trying to help fly by night hustlers scam people out of crypto or run spam campaigns, and in fact often voluntarily watermark to prevent misuse of their products.

You could argue that’s “just to avoid bad PR” and maybe you’re right, but that’s just another way of saying that it’s more profitable to prioritize other use cases than the deepfake/spam market. Spammers and fraudsters are shitty customers and a major brand risk.

[t-writescode]

Could also be a problem of the form of P=NP. Validating might be very easy, but writing might be hard. Like the traveling salesman problem. It’s very easy to tell whether a specific path takes N units of time, but it’s hard to figure out if there’s any path, among all possible paths, that takes N units of time.

[OtherShrezzing]

In part because model vendors specifically prefer when people think that lots of content is produced by their model. The more Claude-like writing appears on the internet, the more signal there is to investors that people are using Claude for a greater number tasks.

[Retric]

It’s an arms race where the AI companies are at an extreme disadvantage due to relative training costs.