Eric Wall

abdel Verified account @AbdelStark · 5d

Some personal news: I joined Lucid Computing (@LucidCompute) as Director of Research Partnerships & Standards.

My core focus will be to help foster more open source collaboration across the AI verification ecosystem, between the many teams working on similar / adjacent topics and pieces of technology.

I am so excited to finally be able to share this next chapter of my career. As I posted recently, I have been completely obsessed with AI safety.

It accelerated in March this year, when @ercwl put me in a group chat with @Kristian_Ronn. At the time, I was actively exploring the use of Zero Knowledge Proofs for critical AI use cases (mostly in insurance, defense, healthcare and robotics).

We immediately had great vibes together. We shared the same vision and the same enthusiasm, and we both saw AI safety as the most important thing humanity could work on.

Since then, I have met Greg, Lucid’s co-founder and CTO, and it was the same kind of connection, the same feeling of a shared vision and mission. And everyone I have met at Lucid since I joined is the same kind of person: talented, mission driven, and with a deep willingness to have a real impact with our work.

It’s a real honor to join them all on this epic journey.

I genuinely believe that cryptography and math are our only hope of avoiding a dystopian, totalitarian world with massive concentration of power, gatekeeping of intelligence, and abuse in many forms.

The beauty of cryptography and math, and Zero Knowledge Proofs in particular, is that they are tools that can break the asymmetry of power. This is precisely what we need for AI right now.

We also need solutions designed for the most adversarial conditions possible, including resisting superpowerful nation-state actors. That’s the only way to aim for credible international treaties to “pace AI”.

The threats and risks are immense, and AI safety is probably the greatest challenge humanity has ever faced. Despite that, I am hopeful, and I remain optimistic that we still have the power to change course. This will inevitably require cooperation and collaboration at many levels.

From what I have seen, there is a strong appetite for more open source collaboration in the AI verification ecosystem. That’s a great starting point, and I have high hopes that we can stimulate and nurture a very healthy, collaborative environment, with no zero-sum mentality.

We need to join forces, more than ever.

The space is way too small today. It will have to grow massively over the coming months, and I am sure it will. That makes it even more important to build solid foundations and a strong culture in the ecosystem, so it can grow and scale the right way.

I will do my very best to help build that collaborative environment. I love building in public, I am an open source maxi, and I am used to working with many entities across distributed ecosystems.

The topics and primitives where I will focus and facilitate open source collaboration:
- ZK for provable inference
- Proof of exhaustion schemes
- Workload classification, via different techniques
- Hardware roots of trust, TEEs, hardware formal verification
- Formal verification

If you are working on any of these in AI verification, I would love to connect and see how we can join forces on our shared mission.

If you are not working on AI verification (or AI safety more broadly): first, you are missing out. Second, we need you. We need a lot more talent working on these problems.

Lastly, I will work on public communication and educational material about AI verification and our work at Lucid. More on this very soon.

Follow @LucidCompute and @Kristian_Ronn, and stay tuned.

Let's build AI using tech to make it more verifiable and safer!

View quoted post on X