Developing policies to prevent global catastrophe as Cofounder and Director of Policy @globalshieldhq
Canberra
Joined September 2010
- Tweets2.6K
- Following1.3K
- Followers762
- Likes1.7K
Rumtin retweeted
There's a decent chance the internet ceases to exist as such, at least for a time.
AI-enabled bioweapons are also coming, and there's no "shut down the websites" option for that.
We're approaching the point where frontier AI R&D becomes politically untenable.
Vulnerabilities in a programming language that supports much of the internet are so severe that European and British authorities recommend websites temporarily go offline until they can be fixed. nbcnews.com/tech/security/cr…
Rumtin retweeted
It was great to be back at @dfat this week to present on AI supply chains (from mines to models). I tested ideas from my work on how trade policy should adapt in light of AI, including to help reduce AI risk. This includes an early attempt to... 1/3
Every AI tool depends on a long supply chain: critical minerals, chips, data centres, models. Every link is a potential chokepoint.
This week @devonwhittle, our Australia Director, presented on those chokepoints to the Trade Law and Economic Security Division at @dfat.
AI presents risks beyond those usually associated with trade and supply chains. Expanded access to advanced cyber and biological capabilities greater concentration of power, and loss of control are all on the horizon.
Our work is looking at how trade policy and economic instruments can contribute to dealing with these risks, and help Australia shape its own AI future.
#AI #AusPol
Robert Pape: “I don’t think world war 3 is coming in the next six months. What would I put as a probability? Maybe 2 out of 10.”
Peoples sensitivity to global catastrophe is so low these days…
Rumtin retweeted
The greatest geopolitical risk is no longer a single war, disrupted shipping route or economic shock. It is the simultaneous interaction of the multiple crises that amplify one another faster than governments, markets and international institutions can absorb their consequences.
Like, subscribe, all that stuff!
If you want to discover what is really causing the greatest risks to our world and what measures governments can take with regard to it, then subscribe to the Global Shield monthly briefing 📬
In our October briefing we looked at AI agents turning rogue, war ranking at the top of the UN's list of risks, and new developments at NATO and in the US Congress.
Links below 👇
"human history has given us no experience of facing an actor which is smarter than we are. Is capable of out-thinking us. This is genuinely novel. And deserves the serious attention it is getting, including in MI5."
mi5.gov.uk/mi5-director-gene…
Rumtin retweeted
I think it's plausible that within a year we'll see AI-based cyberattacks on the power grid, financial system, internet infra, or water. Things like the French supermarket cyberattack but worse. As basic prep I would recommend:
Cash: Get a few weeks of spending in cash
Internet: Get a Starlink Mini so that you have Internet if local fiber or cell is down (~$150 plus $10/month in standby mode)
Power, minimal: Enough to run Starlink Mini and phones (~$1,250)
Daily need: about 1 kWh (Starlink Mini 0.7, phones and laptops 0.2; can limit runtime)
Battery: 1 kWh LiFePO4, e.g. EcoFlow DELTA 3 Plus
Solar: 1x400W folding panel, which refills about 1 kWh on a sunny day
Power, better: Enough to run a fridge, Starlink Mini and phones (~$5,000)
Daily need: about 2.5 kWh (fridge 1.5, Starlink Mini 0.7, phones and laptops 0.2)
Battery: 4 kWh LiFePO4, e.g. EcoFlow DELTA Pro 3
Solar: 3x400W folding panels, which refill about 3 kWh/day in sun
Generator: ~2,500W dual-fuel inverter generator; refills battery in a few hours even if weather is bad; get a few 20 lb propane tanks
Water: 14+ gal of water per person for two weeks
Food: 2+ weeks of shelf-stable food
Comms: Hand-crank NOAA emergency radio
Devon gives evidence to the Committee one week, and the next week they take his recommendations forward. Policy impact is rarely this quick!
New accelerationist argument just dropped. Not that doomers are too pessimistic, but too optimistic!
“What’s your p(doom)?”
“Idk, maybe 5%”
“1 in 20 chance of extinction. So you support pausing AI?”
“Ofc not. That would raise it to 30%”
“What?”
“A global pause doesn’t stop people. It just pushes it underground, into secrecy and paranoia, with worse talent and fewer resources. Way riskier.”
“Fine, but what if we actually stopped AI forever?”
“100%”
If there's one thing humans love doing, its finding new ways to off ourselves.
The policy challenge that mirror life presents is both urgent and unusual.
RAND researchers have developed a strategy to prevent mirror life from ever being created. bit.ly/4h8nuRw
"require confidential disclosure of dangerous capabilities and safety incidents; clarify liability standards for models that cause external harm; invest more in interpretability and alignment research; properly fund government capacity for testing high-risk models; and, crucially, ensure that critical infrastructure is hardened against AI-enabled cyberattacks."
So do what AI safety people are asking for?
bloomberg.com/opinion/articl…
The world had forgotten about nuclear risk long before AI came along.
theguardian.com/commentisfre…
Rumtin retweeted
Japan declares cyber space emergency as attacks soar ft.com/content/2d4333d1-8119… via @ft
Rumtin retweeted
Global Shield Australia has made a submission on the Australian Standards for AI.
When something goes wrong with AI, Australians deserve to know.
Our submission sets out a detailed model for mandatory AI incident monitoring and reporting (AIIMR).
How it would work 👇
Rumtin retweeted
"Washington should welcome the division of labor among a wider 'compute coalition.'" Australia should lean into its role in such a coalition, including through effective regulatory frameworks, international agreements, and investment in technologies it requires.
Ever since Trump's landmark decision to sell AI chips to the Middle East, the United States has struggled with a fundamental question:
How should we balance the rewards of exporting compute against the risks of its diversion or transshipment to China?
aei.org/foreign-and-defense-…
Rumtin retweeted
Today, Global Shield Australia's Director @devonwhittle and Associate Maximilian Evans gave evidence to the Joint Select Committee on Artificial Intelligence.
Our message: AI's most serious risks are no longer hypothetical 🧵
Rumtin retweeted
It was great to appear before the Joint Select Committee on AI today, along with the Gradient Institute, to discuss what Australia could be doing to reduce the potential for catastrophic harms from advanced AI. We had great questions and engagement from JSC members. 1/4
