Introducing def/acc at EF

tl;dr: I believe def/acc (defensive acceleration) – building better defensive technology – is one of the most important ideas in the world today. I think it’s the best way to accelerate technological progress while taking risk seriously – and its a particularly good way to think about AI. I want to see more great companies of this kind, so I’m going to personally run a small EF cohort on this theme in London in September. If you’re a future founder or a very early stage company in this space, sign up for our launch event here or apply here. If you have questions, you can also email me on matt at joinef dot com.

Def/acc is about building technology to protect us from the biggest threats we face – everything from pandemics and cybercrime to powerful AI and nuclear war. It’s the idea that the most powerful solution to technological risk is often more technology. It’s a way to reconcile technological optimism with taking dangerous capabilities seriously.

I recommend this essay by Vitalik Buterin and this one by Michael Nielsen. In Vitalik’s words:

There are certain types of technology that much more reliably make the world better than other types of technology. There are certain types of technology that could, if developed, mitigate the negative impacts of other types of technology. The world over-indexes on some directions of tech development, and under-indexes on others. We need active human intention to choose the directions that we want

I believe that almost everything we value is downstream of scientific and technological progress. I’ve spent my entire career at Entrepreneur First (and more recently at ARIA) trying to accelerate both. I’m proud that we’ve funded things like the world’s best cancer screening AI, satellite-based laser communication, new rocket engines and regenerative liver drugs.

Of course, sometimes technological progress creates dangerous capabilities. It’s impossible to read Annie Jacobsen’s recent book Nuclear War: A Scenario (highly recommended!) and not think as you read it: “what can we do to make sure this never happens?”

Broadly, we have two options: “blunt the spear” or “strengthen the shield”. The first means banning or regulating the technology; the second means building better defensive technology. This is def/acc (Jacobsen’s book would end after just a few pages, and rather happily, if we had better missile defence).

But def/acc isn’t (just) about technology for the military: we need better pandemic detection, better vaccine development, better defensive cyber capabilities, and so on… That’s not to say that technology alone will save us. But whatever you believe about the impact of popular movements or regulation, better defensive technology is close to a free lunch.

Def/acc seems particularly valuable in the area I know best: AI. Powerful AI is coming. It has the potential to transform the world like few other technologies. The upside is incalculable, but the transition will create real risks and huge adjustment challenges. We can blunt the spear or strengthen the shield. 

Last year, I spent time advising the UK government on the creation of the AI Safety Institute, then took a sabbatical to co-lead the negotiations for the AI Safety Summit the UK hosted in Bletchley Park in November. When I returned to EF, I asked myself what I could do in the private sector to help ensure AI is an amazing positive for the world. My answer is: strengthen the shield – help people build defensive AI companies.

There are a number of future AI risks where better technology might be an important part of the solution (and produce some great companies in the process). I’m interested in questions like:

• What technologies do we need to make a world with millions or billions of capable AI agents go well?
• What tools will governments need to fight terrorists and other bad actors who might have much more powerful offensive cyber capabilities than they do today?
• What tools do we need to preserve a shared sense of reality in a hostile information environment? (“Community notes for everything”?)
• What technologies can preserve or amplify human agency in a world of powerful AI?
• What technologies do we need to prevent, detect or mitigate against another global pandemic, natural or engineered?

If the idea of building defensive technology resonates with you, I’d love to help you build a company in this space.

EF helps exceptional people build companies from scratch. Unlike most investors, we start with individuals, not companies. We carefully curate a group of extremely talented people and pay a stipend to cover 12 weeks living expenses while you explore cofounders and ideas. In exchange for the stipend, we get an option to invest at the end of that period. 

If we invest, you get $125,000 for 8%, plus a $125,000 MFN SAFE. We then work with you for a further 12 weeks to help you get ready to raise your seed round. Optionally, you can choose to do the second half of EF in our San Francisco office.

The previous paragraph is true of all EF programs. What’s different about the def/acc stream?

• First, I’ll be running this program personally and be an advisor to each team that we fund. 
• Second, we’ll provide access to some extraordinary speakers and advisors. To kick this off, Reid Hoffman will be holding a special event with me on Thursday 30 May to launch the program in London. Apply to join us here.
• Some of the other people who’ve agreed to get involved include: Eric Schmidt, former CEO and Chair of Google; Kate Bingham, former Chair of the UK Vaccine Taskforce; Jack Clark, co-founder of Anthropic; Jason Matheny, CEO of RAND and former national security official; and Patrick Vallance, former Chief Scientific Adviser to UK Government. More details on other speakers and advisors soon…
• Third, although EF usually works exclusively with individuals “pre-company”, for this program I’ll consider applications from teams that are already building. I’ll take a very small number of these. The program will be exactly the same, but you’ll get the investment upfront, rather than after 12 weeks.

The final point of admission to the program will be September 2nd. We can start helping people who are ready sooner as soon as we make them an offer. We have an “early deadline” for applications on 12 June and a “final deadline” on 17 July. We’ll fill up places as we go, so I encourage you to apply early. I’m taking applications here.