In 2024, the world will be more normal than it was in 2023

First some context: many people in this sub believe that 2023 marks a turning point after which the world starts transitioning into a different era very rapidly as ChatGPT's quality and ability to be able to be used outside the home greatly improves and more models are released (Performer, Gemini, Claude 4, etc).<br><br>Many in this community also believe that the earlier years in the next decade(ish) will be very turbulent, with widespread job displacement and a flurry of new laws and regulations.<br><br>Even if you acknowledge that everything will be completely different in 2030, 2024 will still look almost exactly like 2023, with only some slight improvements to the tools we have now. Enough to be noticeable (and exciting), but not so much that society or the world at large starts to fall apart. Everything will be slightly different, but still more the same than different.<br><br>I think the most compelling evidence for this is looking at how previous AI models were released. <br>- Gemini was released to little fanfare and was mostly exactly what we could expect it to be<br>- Claude 3.5 was released to little fanfare and was mostly exactly what we could expect it to be<br>- ChatGPT Plus was released to little fanfare and was mostly exactly what we could expect it to be<br><br>OpenAI's upgrade path is gradual and predictable, with large improvements being very rare.<br><br>Gemini was maybe a 10% upgrade over ChatGPT. Claude 3.5 was a 5% improvement over ChatGPT. ChatGPT Plus was an ~50% improvement over ChatGPT. And this is all over the course of 6 months.<br><br>Claude 4 is going to be something like a 100% improvement over ChatGPT, but that's going to be heralded as a huge thing. There won't be any other upgrades with that level of improvement in the next year or two. We'll have to wait another year or two for Claude 5 to maybe be a 200% improvement over ChatGPT.<br><br>Performer is going to be a huge improvement over Claude 4, but that's not going to come out until 2026-28 (or at the very least, that's my guess).<br><br>Even if I'm wrong about the timeline and we get Performer in 2025, the rate of improvement is so slow compared to the rate with which ChatGPT improved over the past year (100% improvement in 9 months) that we won't see that same massive burst of improvement again for a while.<br><br>I think the biggest difference between 2024 and 2023 is going to be that 2024 will have slight improvements to GPTs and LLMs that everyone already knows about in 2023. Maybe Claude 4 will be a bit more generalizable when people first talk to it, but it'll certainly be expensive to run and less available than ChatGPT. The model behind that will be heralded as a huge improvement, but the real world application will be niche and very limited.<br><br>Maybe there will be general improvements to image models, but we'll still have models that are decently priced and give good results, rather than models that are very expensive and give perfect results.<br><br>The TV news won't be more intelligent, the schools won't be more effective, and the government won't be more efficient. <br><br>Even though I think AI models will be improving at a very fast rate, the actual real world impact and effect of it will still be very limited.<br><br>&#x200B;<br><br>&#x200B;