I’ve fucked around a bit with ChatGPT and while, yeah, it frequently says wrong or weird stuff, it’s usually fairly subtle shit, like crap I actually had to look up to verify it was wrong.
Now I’m seeing Google telling people to put glue on pizza. That a bit bigger than getting the name of George Washington’s wife wrong or Jay Leno’s birthday off by 3 years. Some of these answers seem almost cartoonish in their wrongness I almost half suspect some engineer at Google is fucking with it to prove a point.
Is it just that Googles AI sucks? I’ve seen other people say that AI is now getting info from other AIs and it’s leading to responses getting increasingly bizarre and wrong so… idk.
Google in particular is in a bit of a bind because they were late to the AI bandwagon. The number of people who actually understand these generative tools well enough to be project leaders on them is very small, so it’s not a problem that they can just throw a ton of money at to solve - they have to build their institutional knowledge about LLMs and their products from the ground up the hard way.
Conflicting with this is the pressure to release something to please the investors. So Google has an unfinished product made by people who are still learning how it works, and while the smart thing to do would be to keep it under wraps until it can compete with ChatGPT they’re forced to polish that turd and release it anyway.
Model collapse and all the other things people are talking about are factors to keep an eye on, but I think that’s the primary reason why Google’s go at AI was so fantastically worse than the others we’ve seen.
Googles highest end ai is ranked really highly though, second only to the 3 best gpt-4 models. Even outranks a couple other gpt-4 variants. They weren’t really that late (I don’t think they were late at all tbh), they just didn’t invest in a completely dedicated AI company like OpenAi and Microsoft, they actually built their own from the ground up. I think the main problem with googles search model is that it’s really bad at crawling the web for data and turning that into a coherent answer. But their LLM alone and as a chatbot is top notch
You are absolutely right about having something to please investors because Microsoft was way more prepared to integrate Copilot into all their products. Googles models are super powerful, they just were not prepared to package it into a consumer product as soon as they did