Welcome to the Future of Life Institute newsletter. Every month, we bring 40,000+ subscribers the latest news on how emerging technologies are transforming our world - for better and worse.

If you've found this newsletter helpful, why not tell your friends, family, and colleagues to subscribe here?

Today's newsletter is a 7-minute read. We cover:

• 📻 Imagine A World is out today!

• 📊 New polling shows overwhelming AI concerns and support for regulation

• 🗓️ The UK AI Safety Summit is two months away - will China be invited?

• 👩‍💻 PhD and postdoc fellowship applications are now open

Can you imagine a world in 2045 where we manage to avoid the climate crisis, major wars, and the potential harms of Artificial Intelligence?

Imagine A World is a podcast series exploring a range of plausible and positive futures with advanced AI. We interview the creators of eight diverse and thought provoking imagined futures that we received as part of the worldbuilding contest FLI ran last year.

The first two episodes are available now on all major podcast platforms, and we’ll be releasing a new episode each week. Peace Through Prophecy explores how new governance mechanisms could help us to coordinate, while Crossing Points unpacks the importance of designing and building AI in an inclusive way.

Listen to them now on YouTube, Spotify, or Apple Podcasts, and be sure to like and subscribe!

The results reveal bipartisan support for guardrails on AI development. 64% of voters “support a policy that would require any organization producing advanced AI models to obtain a license, that all advanced models be evaluated for safety, and that all models must be audited by independent experts”.

Full results from both AIPI polls are available here; full results from Pew are available here.

We are delighted to announce that applications are back open for our PhD and postdoctoral fellowships focused on AI existential safety research. The fellowship is global and open to all regardless of nationality or background; we are seeking a diverse applicant pool. All Fellows will receive applicable tuition and fees, as well as a stipend and research/travel fund.

Current or future PhD student intending to research AI existential safety? The deadline for PhD fellowship applications is 16 November 2023 at 11:59 pm ET.

Current or future postdoc working on AI existential safety research? The deadline for postdoctoral fellowship applications is 2 January 2024 at 11:59 pm ET.

AI policy:

• UN Secretary-General António Guterres has announced the creation of a multi-stakeholder High-level Advisory Body on AI, in line with the UN Roadmap for Digital Cooperation.

  • FLI participated in the creation of the Roadmap, and helped put forward the recommendation that such a body be created.

• The UK’s AI Safety Summit will take place 1-2 November, convening international governments and stakeholders to discuss action needed to centre safety in AI development.

  • While it’s not confirmed whom will be invited, as FLI President Max Tegmark recently discussed, the Summit’s success is contingent on the inclusion of China - it could represent a “turning point” on AI safety.

• In advance of the Summit, the House of Commons Science, Innovation and Technology Committee recently released a report urging the UK government to act quickly on AI regulation, citing twelve main risks presented by the technology - including extinction, privacy breaches, national security threats, and more.

Nuclear weapons:

• In a rare move, more than 100 medical journals shared a joint call-to-action, urging the health community to work together to “reduce the risks of nuclear war and to eliminate nuclear weapons”.

• The UN International Day against Nuclear Tests on 29 August was met with renewed calls for the 1996 Comprehensive Nuclear-Test-Ban Treaty to come into force. Unfortunately, eight countries have yet to ratify it, leaving it at a standstill.

• In the Bulletin of the Atomic Scientists, FLI’s Dr. Emilia Javorsky and Hamza Chaudhry wrote an in-depth examination of AI convergence risks - the ways in which AI can amplify and accelerate risks from other technologies.

• FLI co-founder and tech pioneer Jaan Tallinn appeared on Al Jazeera’s The Bottom Line to discuss the risk to humanity presented by AI with host Steve Clemons.

• FLI President Max Tegmark was profiled in The Wall Street Journal, discussing the “dangerous race” among AI labs, and his shift from optimism about AI to alarm.

• On the FLI podcast, host Gus Docker spoke to social scientist and GovAI International Governance Lead Robert Trager on the need for, and challenges of, international AI governance.

Finding new vulnerabilities: IBM released new research on large language models (LLMs) and cybersecurity, demonstrating how remarkably easy it is to circumvent LLM safety guardrails simply through language, by “hypnotizing” them in the context of playing a game.

Why this matters: This is only one example of a plethora of research that’s come out recently exposing the weaknesses of widely-used LLMs. Researchers are particularly concerned that these vulnerabilities will empower bad actors who wish to use LLMs to collect user data, spread disinformation, or obtain knowledge that could be used to cause harm.

• A proposal for AI governance: You may have heard this discussed on the FLI podcast - recent guest Robert Trager co-authored this new paper on applying a jurisdictional certification approach to international AI governance.

• “Open-washing” AI: What does it mean for big tech to be “open”, in the context of AI development? David Gray Widder, Sarah West, and Meredith Whittaker examine this in a new research paper.

• A DIY disinformation machine: An anonymous developer made waves by publicising their recent project on AI disinformation, having essentially created an LLM-powered anti-Russia propaganda machine that produced fake tweets, articles, and even journalists and entire news sites.

• “Invisible” bio labs: TIME reports on privately-owned bio labs in the U.S. falling through the regulatory cracks, potentially exposing the public to great biosecurity risks.

6 and 9 August marked the 78th anniversary of the nuclear bombings of Hiroshima and Nagasaki respectively - the only two occasions thus far when an atomic bomb has been used in war. An estimated 215,000+ people, mostly civilians, were killed, with many more injured and/or left homeless as the two cities were effectively demolished.

Many contemporary nuclear bombs - an estimated 13,000 exist today - are several hundred times more powerful than those used on Hiroshima and Nagasaki 78 years ago. The continued threat of these weapons, an exchange of which would guarantee mass devastation, is unacceptable. We must work to ensure they are never used again.

Read more on the risks of nuclear weapons, and our work to mitigate them, here.