We’re going to see more focus on the “models” which is just a database of knowledge. Instead of paying OpenAI for access, you’re going to see:
1. Marketplaces. Example: buy access to this “finance” model, it knows everything about financing businesses. Very similar to selling encyclopedias!
2. Ideally, these are models we can own forever, like an encyclopedia you can save on a disk. (Self-hosted enterprise solutions.) But mostly these will be services you subscribe to, and the models will be hosted remotely, with gatekeepers, maintained by domain specialists (more cool jobs for nerds.)
Of course, a dumb encyclopedia can’t do much without capabilities. New businesses will need to create the capable tentacles, the integrations with the APIs, to give the AI models useful capabilities in the real world. This will create even more jobs, and it won’t be easy, because businesses will not want private data and API keys in the possession of 3rd party integrators, leading to a spectrum of solutions with various capabilities in various niche markets. Unfortunately, the general public won’t be as concerned about privacy, whereas businesses have valuable assets, trade secrets and private data to protect.
Basic example: Ask your AI model to generate a “curl” command to send an SMS message through Twilio to your customers, and pipe that to a Linux command line.
Once these AI models (encyclopedias) have capabilities, they can be packaged together and sold as assistants. For example, a “Blog Media Assistant” with WordPress API access could do various CRUD operations in your Media Library. “Convert all these PNG files to JPG, reduce the brightness by 20%, add my logo in the top right corner, center my phone number at the bottom, and upload them to my Media Library.”