Why can't Google easily compete in AI?
Articles and comments are saying Google is facing great threats from AI, which seems strange considering they have a lot more resources than OpenAI, and a long history of implementing AI into their products, and they have legendary AI researchers on their payroll, with massive compute facilities. It would seem to a naive observer that they are well positioned to develop novel iterations on GPT and augment existing search products with it relatively easily. They already have the reach with the world's most popular domain name, and with all their other resources, seems hardly like the "existential" threat to their business as you see publicized.
I think it a problem of size. Having worked in a big tech company before, I saw first hand how sometimes inefficacy can creep in. It's very hard to execute quickly in new areas once a company gets too big. Size, bureaucracy and politics can reduce the speed of decision making.
Another issue is sometimes big companies don't feel the need to push too hard in one area. Open AIs survival depended entirely on their execution and delivery of GPT. It's very possible that Google would have killed off a GPT like project in it's early stages because it has many other streams of revenue. They can afford to kill of a 1M revenue business cause its merely a rounding error. See https://killedbygoogle.com/
GPTChat is a product, not a significant innovation in machine learning. OpenAI put together existing technologies.
Google obviously is far more innovative in AI. The underlying technologies such as transformers came out of google (although there have been some publications in academia before that too). Google and other big companies most likely have similarish large language models, they haven’t made them public yet. Companies such as Google are also under spotlight if it goes wrong. Wait until articles are published that ChatGPT has created bias, produced false false information, etc.
Google has decided not to offer "raw" language model API access, but other than that I'm not sure they're losing. They should be able to easily clone any AI-based apps that get big.
They are too big to be fast and as a public company they need to implement AI in a way that increases revenue, which makes it harder I think.