GPT4 is better than its predecessor and quite impressively so. The (probable) param count reflects that. It’s not just param count, because not all large model

GPT4 is better than its predecessor and quite impressively so. The (probable) param count reflects that. It’s not just param count, because not all large models are good, but it clearly is part of the equation. I always wondered why Altman said going bigger is a dead end. If it were true, saying it would needlessly inform your competition. What’s the use in that? If it were false it might dissuade them from going down that path.. I think I got my answer.