This is a really terrible example for your argument considering that the digital camera was invented at Kodak but they refused to market it
https://archive.nytimes.com/lens.blogs.nytimes.com/2015/08/1...
It's not like Google tinkered with language models and forgot about it, like Kodak and digital photography.
This is a code red.
Not because of tech but because a direct challenge to how a search engine monetizes ads.
They need a new business model.
Kodak would sell you a 1.3 megapixel SLR in 1991! That was a decade before Nikon got their products in order. It is a 100% made-up myth that Kodak did not see the potential of digital. Around 1990 their position in the digital camera market was comparable to the position of Tesla around 2015 in the electric vehicle market: the only company that understood there was demand for a technologically cutting-edge product at a pretty high price.
The reason Kodak went out of business is because they lost a phenomenal amount of money trying to become a pharmaceuticals manufacturer.
Edited to add: This is a really excellent example of the ultimate futility of ChatGPT. All it can be is a novel compression algorithm for folk wisdom, which is often wrong. I asked it why Kodak failed to make digital cameras—a misleading question because Kodak was the only player in the digital camera market prior to 1999. Here is the pack of lies it regurgitates.
"""Kodak was slow to embrace digital technology. While other companies were investing in digital camera research and development, Kodak continued to focus on its film-based products. By the time Kodak entered the digital photography market, it was already facing strong competition from companies that had been established in the digital space for some time."""
This is, again, totally false. The real history is the opposite. Kodak was the established player.
At least with these models you can easily train them on the information you want them to use.
I think it's indisputable at this point that the compressed information is being decompressed and manipulated/recombined by prompting in ways that do useful work. It's clearly not just regurgitation.
You could take one of these models such as text-davinci-003 and using the OpenAI API (or go the open source route) start fine-tuning it on more accurate information.
Also we will soon within the next year or two see multimodal models that have clear abilities to manipulate and ad-hoc query visual/spatial scenarios. Which you can already do to a limited extent with existing models in cases where language semantics happen to capture some relevant spatial concepts.