Michael J. D. Vermeer

Happy to be quoted in this Vox article discussing my recent paper from RAND on Evaluating Select Global Technical Options for Countering a Rogue AI.

AI Is About to Change Our Lives Forever. Luckily We Have a User Manual for That. The question isn't whether to slow artificial intelligence, but how to ensure it strengthens rather than undermines the democratic equality the Constitution promises.

"As technologies grow more capable, the question becomes fundamentally constitutional: Will AI be used to fulfill the promise of American democracy, or will it be allowed to create the kind of concentrated power that the founders designed the Constitution to prevent?" https://bit.ly/4hiYYLO

Bottom line: all is not lost. Even if perfect containment is impossible, layered safeguards grounded in computation, information theory, and thermodynamics could likely give humanity time and tools to respond to even threats from a superintelligent AI adversary. (3/3)

Many assume a superintelligent AI could bypass any human-imposed limits. We challenge that assumption. Using standard security engineering approaches – threat modeling, protocols, primitives, and layered defenses – humans can impose real costs and barriers. (2/3)

Can Humans Devise Practical Safeguards That Are Reliable Against an Artificial Superintelligent Agent? Can humans design practical safeguards to protect our critical and digital infrastructure against attacks from a future artificial superintelligence? In this paper, the authors argue that such sa...

Can humans build defenses that could constrain even a future superintelligent AI? Our new @rand.org paper argues yes—by leveraging fundamental scientific limits and proven security engineering principles. 👉 www.rand.org/pubs/perspec... (1/3)

Thanks! Depends on the type of risk you're talking about! For catastrophic risks, I think very long-term. There are plenty of other lesser risks that are short-term or already here though.

How Might Artificial Intelligence Enhance Cybersecurity? When it comes to cybersecurity, attackers have the advantage. But with the right investments in people, process, and technology, advanced artificial intelligence could be used to tilt the balance of c...

RAND's Chad Heitzenrater outlines why advanced AI might fundamentally reshape the economics of cybersecurity. ⬇️ https://bit.ly/4ni8ksU

Could AI Really Kill Off Humans? Many people believe AI will one day cause human extinction. A little math tells us it wouldn’t be that easy

I also wrote a more accessible discussion of the report that was just published in @sciam.bsky.social called "Could AI Really Kill Off Humans?" Take a look!
www.scientificamerican.com/article/coul...

On the Extinction Risk from Artificial Intelligence RAND researchers examined the threat of human extinction posed by artificial intelligence and identified risk indicators. They analyzed three scenarios and assessed that it would be immensely challeng...

Happy to share a new report just published from @rand.org called "On the Extinction Risk from Artificial Intelligence." We tried to take a serious, rigorous look at whether human extinction from AI was possible.
www.rand.org/pubs/researc...

AI and the Future of the U.S. Electric Grid AI data centers suck up energy like small cities. But AI could be a big part of the solution, too. There are risks here and grid operators need to move with caution. But AI could usher in an energy fu...

AI needs power, and lots of it. The new RAND article argues that keeping advanced industries in the U.S. requires more than innovation, it needs strategic infrastructure.

More on: www.rand.org/pubs/article...
Full study done by RAND and RAND Europe: www.rand.org/pubs/externa...

#AI #U.S. #Energy

Artificial General Intelligence's Five Hard National Security Problems Whether artificial general intelligence is ultimately realized or not, its development may pose serious challenges to national security, including the sudden emergence of wonder weapons, systemic shif...

Leading AI labs are in hot pursuit of AGI, which would produce human-level—or even superhuman-level—intelligence across many cognitive tasks.

RAND experts highlight 5 hard national security problems such a development could present: www.rand.org/pubs/perspec...

Rogue AI Moves Three Steps Closer OpenAI’s new o3 model suggests that it will not be long before AI systems are as smart as their human minders—or smarter.

In light of OpenAI’s new o3 model, @petersalib.bsky.social writes that "rogue AI is a concern worth taking seriously—and taking seriously now. This is a problem for which, by its very nature, solutions cannot wait until there is conclusive proof of their need."

Poorly predict the future.

Looking Back on the Big Policy Stories of 2024 RAND president and CEO Jason Matheny reflects on the many ways RAND is helping leaders of all stripes better understand and address humanity's greatest challenges.

Throughout 2024, RAND has been helping leaders better understand and address humanity's greatest challenges.

RAND president and CEO Jason Matheny reflects on some of the major policy stories of the year—and where we intend to build in 2025. 🧵
www.rand.org/pubs/comment...

Understanding and Managing Global Catastrophic Risk In 2022, Congress passed the Global Catastrophic Risk Management Act to provide policymakers, emergency management planners, and others with a strategy to respond to catastrophic risks. A new assessme...

Supervolcanoes. Nuclear war. Asteroids colliding with the earth.

New RAND research provides an overview of global catastrophic risk: www.rand.org/pubs/researc...

Welcome to the New rand.org RAND's newly redesigned website has a fresh look and feel, improved functionality, and enhanced performance.

Free to read. Easy to navigate. Accessible for all.

Welcome to the new rand.org. ⤵️ www.rand.org/pubs/article...

All - Bluesky Directory A curated collection of all things relating to the Blue Sky social media platform.

Starter packs are genius, but I was surprised there wasn't a list of them for people to find.

So I built it:
blueskydirectory.com/starter-pack...

The website monitors the packs being shared and adds the ones it finds to the database.

Missed your stater pack? Message me and I'll get it added.

I didn't get a bunch of new followers, but I'm going to let you all what I do anyway.
👋

I research national security implications of emerging technologies (e.g., quantum computing, 5G, AI) for RAND, a nonprofit, non-partisan research institution.

Looking to follow others with similar interests!

Introduce yourself either some jobs you have done apart from what you do now.

1. Car washer.
2. Lifeguard.
3. Drill press operator.
4. Golf course greensperson.
5. Tutor.
6. Chemist.

AI Governance: Lessons from Earlier Technologies What can be done to maximize AI's benefits while minimizing its potential risks? The approaches to governing four earlier technologies—nuclear technology, the Internet, encryption products, and geneti...

Is AI like nuclear technology? The Internet? Cryptography? Genetic engineering? Should we govern AI like we governed any of these technologies?

In my latest @rand.org report I explore each of these examples and how the lessons learned can be applied to AI governance!
www.rand.org/pubs/researc...

EFF at 25: Remembering the Case that Established Code as Speech One of EFF's first major legal victories was Bernstein v. Department of Justice, a landmark case that resulted in establishing code as speech and changed United States export regulations on

www.eff.org/deeplinks/20...

I had the privilege to review this excellent paper, and I highly recommend it for folks interested in AI and cybersecurity!

have a joke about Sisyphus, but I retell it every day without ever getting to the punchline.

Exciting breakthough in quantum computers, specifically quantum error correction. That would likely be the most powerful quantum computer to date. 4 reliable logical qubits from 30 physical qubits. Low error rate. Works on ion-trap architecture! s7d9.scene7.com/is/content/q...

Periodic Table Regions xkcd.com/2913

Foundation Model Convening These proceedings capture perspectives emerging from a convening focused on artificial intelligence safety and security issues, such as legal liability, international governance, and cybersecurity.

Great summary of a great event that I was privileged to attend and speak to. It's a short, free read, so I recommend it if you're interested in the governance and security of foundation AI models.
www.rand.org/pubs/conf_pr...

Was just writing up a review and Word suggested I follow the phrase "--> Conclusion" with "-iℏt"

First, I'm flattered it thought I was writing about quantum mechanics, but also - no.

It has been a long time since I had to use the reduced Planck's constant.

I'm going to confidently reply YES to this one.

Heart-shaped pizza

Made a heart-shaped pizza for my family for Valentine's Day, and it was delightful!

Fair assessment.

I won the spelling bee and the geography bee and I was on the math team.