Artificial intelligence is a foundational technology driving innovation in every corner of the economy.
Rapid advances and breakthroughs in AI have drawn policymakers’ attention—as evidenced by the flurry of hearings, task forces, executive orders issued, and bills introduced. These breakthroughs have enabled a flourishing ecosystem of startups building AI, building with AI, and using AI to better everyday tasks.
As policymakers continue to dive in to varying facets of AI with proposals, guidance, frameworks, and incentives, it is imperative that they have a solid grasp of the breadth and interconnectedness of the AI ecosystem.
By volume, startups make up the vast majority of the AI ecosystem. But like many technology policy issues, the perspectives of startups can be drowned out by attention on the largest players in an industry.
Startups are building with AI in a few distinct ways. Some startups are building their own machine learning models to perform specific tasks. Most startups are leveraging foundation models—either by licensing from market leaders or accessing open source—to fine-tune and build unique products. Often, startups leverage multiple foundation models, finding some perform better at certain tasks and others better at different tasks. This increasingly describes how startups are building—90 percent of the startups that joined our network in the past year used AI in their main product. And finally, most startups are leveraging others’ AI tools as part of their product or to augment basic business functions.
Startups are implicated at every level of the AI ecosystem. For example, a startup leveraging the latest foundation model to create a product giving feedback on startups’ pitch decks sits in the middle and touches companies up and down the stack. Policies impacting one part of the stack—like rules on deploying AI, or another, like requirements for foundation model development—will reverberate across the ecosystem.
Policymaking in the AI space requires not one bill or one framework, but smart policies across all issues impacting AI developers, deployers, and users. To chart a course forward, policymakers must begin with a strong foundation.
Read the report here.
