>>15408149In talking through the archives and it's honestly not easy to tell in a lot of the posts. And his model is really basic and whipped up in just a few weeks. He didn't even use the most powerful text generation AI, he only used GPT J and did very little to remove all the small mistakes that make you notice that it's a bot. With more effort this could be very very powerful. It could even be made reinforcing and learn when people notice that it's a bot and improve to make as inflammatory posts as possible you shouldn't underestimate what these things can do I'm in ML myself and see this shit first hand
