The future of A.I. will be like cloud operators today.
This past decade has been a boon to progress in the A.I. space. Image generation, creation of art and music, automatic note-taking, even highly accurate predictive models — they’ve all enabled advances in business productivity.
Research in math and information theory have been a big reason for this forward progress, but most folks would probably say that the main reason was improvements in large scale data collection and processing. Today, our best A.I. models require massive amounts of data in order to find valuable patterns. This massive amount of data requires massive compute resources in order to find these patterns. On top of that, we need highly specialized expertise to leverage this data and compute power to create the software that actually learns these patterns.
All of these elements are hard to find, and they’ll likely only be available to large, well-capitalized companies.
Cloud A.I. Infrastructure
Today, it’s more or less impossible for smaller companies and startups to create their own generative models. Not only do they not have the resources, but they likely don’t even have the desire, particularly with so many new generative A.I. offerings on the table. It wouldn’t make sense for a company to roll out its own offering unless it’s a core product.
This means that we’re heading into a world where companies that have a high level of expertise in A.I. manage the models and infrastructure necessary to keep generative A.I. models operational while end users interact with those models via APIs. I see a world where these infrastructure companies are almost like the new “cloud providers” of technology. They’ll provide the tools, while app developers leverage those tools to create new products and services.
Just as the current cloud operators today have a great business, but are less visible than the apps that run on top of them (think Netflix running on AWS), we’re going to be seeing a world where a large number of developers start building on top of the new A.I. “cloud operators”, where some of them become huge and take off, generating large amounts of value for their creators, while still requiring the infrastructure operators to exist.
While we’re starting to see this now, I think that the prevalnce of small businesses and startups that leverage these generative A.I. infrastructure offerings is going to dramatically increase in the near future.