Researchers discuss AI expectations for 2023: continued problematic art-generating apps, incoming regulations, open-source and decentralized efforts, and more
Context & Ripple Effects
This year-end survey lands after two months in which lawyers, analysts, and startup employees could only flag unresolved copyright and fair-use questions around generative AI rather than answer them unresolved copyright and fair-use questions. The researchers' expectation of incoming regulation sits against a specific warning already on record: experts argued the EU's proposed AI Act would create legal liability for general-purpose AI systems while undermining their development EU AI Act liability warning.
The prediction of rising open-source and decentralized AI work also has a checkable sequel in the coverage: the open-source boom turned out to be precariously built on Big Tech's giant models like LLaMA and GPT-3, vulnerable to a decision by Meta or OpenAI to close access open-source boom built on Big Tech models.
First-order effects
- Art-generating apps enter 2023 still operating without settled answers on whether their training data constitutes fair use, leaving artists and rightsholders with no clear legal recourse.
- EU policymakers face pressure to finalize the AI Act despite expert warnings that its liability provisions would penalize general-purpose systems at the development stage.
Second-order effects
- Startups building on general-purpose models inherit whatever compliance burden the AI Act imposes, giving them a reason to back decentralized or open-source alternatives over closed APIs.
- If regulators move, measurement becomes the bottleneck — which is what NIST's later launch of a generative AI assessment program, including benchmarks and content-authenticity detection tech, addresses directly NIST generative AI assessment program.
Third-order effects
- The pattern points toward generative AI being governed through state-mediated frameworks — liability rules plus government-run evaluation infrastructure — rather than left to platform self-policing.
- Open-source AI's dependence on a few Big Tech model releases means decentralization may not reduce concentration so much as relocate it to whoever controls the base weights.
The trend: Generative AI is shifting from an unregulated free-for-all toward state-mediated governance, even as its open-source layer remains structurally dependent on Big Tech's model releases.