He can make the same criticism of Internet searches as he does of GPT: you shouldn't trust them until you validate them.
I find that GPT's answers are for the most part more reliable the searches, specifically today's searches. In the last 12 months, search results have become so spammy with AI generated pages (oh the irony), that it's hard to find reliable answers.
So like search, I look at GPT's answers with a grain of salt and validate them, but these days I use GPT all day every day and search rarely.
To be fair, I use it a lot because I have a GPT CLI that works just the way I want it to, since I wrote it :-).
https://github.com/drorm/gish
Ya, that's pretty much my workflow as well. Though for my little web app, I could give it the whole ball of html/css/js each time.
I haven't seen anyone else describing this workflow. Feed it the existing code, ask it to modify/improve/fix the code and output a new version of all the input code, review diffs.
It has downsides, because you can easily run out of context window of chatgpt-3.5-turbo. But I am getting much better code out of it versus other approaches I've tried. And it's a very efficient and natural workflow -- we're used to getting and reviewing diffs/PRs from human collaborators.
Speaking of AI generated pages, I wonder how OpenAI filter these low quality web pages out of their training set as they continue to training.
Also, I wonder how they decide what code is worth training on. Because a lot of code is written in poor style/has technical debt, it might be the case that these LLMs in the long run lead to an increase in the technical debt in our society. Plus, eventually, and this might already be happening, the LLM are going to end up training on their own outputs, so that could lead to self immolation by the model. I am not certain RLHF completely resolves this issue.
> I wonder how OpenAI filter these low quality web pages out of their training set as they continue to training.
This. The value proposition is very clearly tied to the quality of the training data, and if there's secret sauce for automatically determining information quality that's obviously huge. Google was built in part on such insights. I suspect they do have something. I'd be utterly astonished if quality sorting were an emergent property of LLMs (especially given it's iffy in humans).
The problem, of course, is that if they do have a way of privileging data for training, that information is going to be the center of the usual arms race for attention and thinking. It can't be truly public or it's dead.
yea i'm kind of shocked none of these models implement any kind of fingerprinting, something encoded in zero width spaces or other invisible unicode. It would be trivial to delete it but for the vast majority of cases, it would allow content to be flagged as model output-do not ingest
Google and others would be wise to add a date filter of "before summer 2023". Maybe a bit longer, but not much time left till AI spam really takes over.
I find that GPT's answers are for the most part more reliable the searches, specifically today's searches. In the last 12 months, search results have become so spammy with AI generated pages (oh the irony), that it's hard to find reliable answers.
So like search, I look at GPT's answers with a grain of salt and validate them, but these days I use GPT all day every day and search rarely. To be fair, I use it a lot because I have a GPT CLI that works just the way I want it to, since I wrote it :-). https://github.com/drorm/gish