> I wonder if over time SEO spam becomes an invisible problem to us and instead the AIs have sort it out for us
There have been some good Twitter threads about this. Basically, this is the last generation of AI to be trained with inputs that were ~100% human. It probably won't take long until the bulk of the content online is generated by AI, at which point it will just be an AI feedback loop. Any new concept, idea, technology, whatever, that doesn't already have a bunch of content about it will never have a guaranteed fully-human-written corpus to start from.
Once there's a viable OpenAI competitor or two, it's probably only a matter of time until the AIs start an arms race to feed each other bad data. Tweak the model to generate subtly incorrect information, regenerate your training corpus in a way that it feeds into other AIs but ensures you ignore it, and now you're basically all the way back to having a few trusted "information gatekeepers".
Except at present we can also train classifiers that recognise GPT output, so ingressed text should be able to be filtered.
https://huggingface.co/openai-detector/ detects GPT2, but it still detects ChatGPT snippets most of the time, even though the algorithm has been improved and the training set is presumably quite different.
That said, I did just generate a complete lie https://news.ycombinator.com/item?id=33874757 using ChatGPT that scored 99.98% real on the detector, so adversaries can pre-filter too.
Also interesting is that historical corpora (every existing corpus) of data becomes more valuable, because they are untainted.
This will in turn spawn a new research field and industry of automated AI content detection in order to filter out poisoned training data. Also expect to see web browsers incorporate that technology with some sort of rating or color code to flag pages that seem to be AI generated. It's going to be a real arms race.
There have been some good Twitter threads about this. Basically, this is the last generation of AI to be trained with inputs that were ~100% human. It probably won't take long until the bulk of the content online is generated by AI, at which point it will just be an AI feedback loop. Any new concept, idea, technology, whatever, that doesn't already have a bunch of content about it will never have a guaranteed fully-human-written corpus to start from.
Once there's a viable OpenAI competitor or two, it's probably only a matter of time until the AIs start an arms race to feed each other bad data. Tweak the model to generate subtly incorrect information, regenerate your training corpus in a way that it feeds into other AIs but ensures you ignore it, and now you're basically all the way back to having a few trusted "information gatekeepers".