I said it here when GPT-4 first came out, it just was too good for development, there was no way it was going to be allowed to stay that way. Same way Iron Man never sold the tech behind the suit. The value GPT-4 brings to a company outweights the value of selling it as a subscription service. I legit built 4 apps in new languages in a few months with Chat GPT 4, it could even handle prompts to produce code using tree traversal to implement comment sections etc. and I didn't have to fix its mistake that often. Then obviously they changed the model from GPT 4 to GPT 4 Turbo which was just not as good and I went back to doing things myself since now it takes more time to fix its errors than to just do it myself. Copilot also went to s** soon after so I dropped it as well (its whole advantage was auto completion, then they added gpt 4 turbo and then I had to wait a long time for the auto complete suggestions, and the quality of the results didn't justify the wait).

Now why do I think all that (that the decision to nerf it wasn't just incompetence but intentional), like sure maybe it costs too much to run the old GPT 4 for chat GPT (they still have it from the API), it just didn't make sense to me how openAI's chatGPT is better than what Google could've produced, Google has more talent, more money, better infrastructure, been at the AI game for a longer time, have access to the OG Google Search data, etc. Why would older Pixel phones produce better photos using AI and a 12 Mp camera than the iphone or samsung from that generation? Yet the response to chatGPT (with Bard) was so weak, it sure as hell sounds like they just did it for their stock price, like here we are as well doing AI stuff so don't sell our stock and invest in openAI or Microsoft.

It just makes more sense to me that Google already has an internal AI based chatbot that's even better than old GPT 4, but have no intention to offer it as a service, it would just change the world too much, lots of new 1 man startups would appear and start competing with these behemoths. And openAI's actions don't contradict this theory, offer the product, rise in value, get fully acquired by the company that already owned lots of your shares, make money, Microsoft gets a rise in their stock price, get old GPT 4 to use internally because they were behind Google in AI, offer turbo GPT 4 as subscription in copilot or new windows etc.

The holes in my theory is obviously that not many employees from Google leaked how good their hypothetical internal AI chatbot is, except the guy who said their AI was conscious and got fired for it. The other problem is also that it might just be cost optimization, GPU's and even Google TPU's aren't cheap after all. etc.

Honestly there are lots of holes, it was just a fun theory to write.