Head of AI Policy @a16z. Fmr director of @UNC_TechPolicy and Facebook public policy.
Durham, NC
Joined September 2012
- Tweets981
- Following642
- Followers2.8K
- Likes2.7K
Pinned Tweet
1/
Are the right AI rules... zero rules? Not for Little Tech.
To ensure AI thrives in the long run—and that startups can compete at the frontier—we need smart regulation. Here’s what we believe at @a16z 🧵
One of the most compelling opportunities for AI is making it possible to build things that previously took too much time or cost too much to produce.
@DiodeInc’s @DavideAsnaghi explains how AI can compress machine-programming work from a year to days. For American hardware companies, connecting design with manufacturing opens up new possibilities to scale production here at home.
How can America turn its strength in AI into a manufacturing advantage? A small team of engineers in Brooklyn is building the blueprint, starting with the printed circuit boards that power nearly every electronic device.
I spoke with @DiodeInc CEO @DavideAsnaghi about building hardware faster, giving factory workers better tools, and what it takes to grow US manufacturing.
Watch our conversation here, or listen wherever you get your podcasts: a16zpolicy.substack.com/p/tu…
Why are courts running college sports?
College sports has spent years moving from one lawsuit to the next. That’s partly because the industry has chosen a path that makes it difficult to develop and enforce stable rules.
In my first piece for a new Substack, the Sports Policy Institute, I look at why litigation has become the dominant tool for governing college sports, and how collective bargaining could create a more durable framework that keeps college sports out of courts.
sportspolicyinstitute.substa…
.@RayanKrishnan and @Glenn__Parham of @ValsAI explain the “domain transfer fallacy” in AI.
They argue that for a benchmark to be credible, it must test models on the real-world work they’re expected to perform. Better benchmarks give businesses and governments better evidence to choose the right model for the job.
Check out our conversation on the growing market for AI evals: a16zpolicy.substack.com/p/me…
If a student took the SAT at home, proctored it themselves, and reported their own score, how much would you trust the result?
That’s @RayanKrishnan’s analogy for allowing model developers to run their own benchmark test sets.
My full conversation with the @ValsAI team covering the role of independent evaluators in AI: a16zpolicy.substack.com/p/me…
AI evaluation should be based on data, not vibes.
If policymakers want to understand the capabilities of AI models, they need reliable evaluation tools that keep pace with the technology.
That’s what @RayanKrishnan and @Glenn__Parham are building at @ValsAI. They joined me to discuss how the maturing AI evaluation market can give industry and government with better evidence for decision making.
Our full conversation on the latest a16z AI Policy Brief: a16zpolicy.substack.com/p/me…
Exciting to see the good work that Vals is doing in evaluating AI models, providing valuable information to consumers and governments.
We're thrilled to invest in Vals.
A frontier model can look brilliant on a leaderboard and still struggle with the messy work that actually matters in the real world.
@ValsAI takes a fundamentally different approach to evaluation: test models on the work people actually want them to do, not on contrived exams.
The team works with domain experts to turn real workflows into rigorous benchmarks, then builds automated grading systems that can evaluate the final work product to an expert standard.
We're thrilled to partner with @RayanKrishnan, @langstonnashold, and the entire Vals team as they build the trust layer to underpin the AI economy.
By @JenniferHli, @stuffyokodraws, @RaghuRaghuram, and @shangdaxu
Matt Perault retweeted
America’s software leadership was built on openness. We have the same opportunity in AI.
If we want a competitive, secure AI ecosystem whose benefits extend far beyond a handful of companies, we need to preserve access to open-weight models.
For AI startups, access to data can operate as a form of startup capital. But if lawful learning requires expensive licenses or access to private data, only the biggest companies will be able to compete.
@DerekSlater breaks down why data access is a fundamental issue for Little Tech.
Fair use and functioning markets aren’t at odds. @DerekSlater explains how, throughout the history of technology, copyright’s limits have helped new products, business models, and sources of revenue emerge.
My full conversation with Derek is on the a16z AI Policy Brief: a16zpolicy.substack.com/p/ai…
When information is lawfully accessible, people are free to read it, analyze it, and build on it. @DerekSlater joins me to discuss how AI is putting the freedom to learn principle to a new test: what changes when a person uses a machine to learn?
More from our conversation on AI’s data access question, new on the a16z AI Policy Brief.