Back to All Events

AIIF Masterclass: AI Capabilities and Evaluations

This Masterclass event will be held on April 9, 2024, at 16:00 (Japan time). Hybrid (online and in-person) access is available. For access: contact@aiindustryfoundation.org.

----

Last week, we dove into Gladstone AI’s report, and whether the recommendations align with other clear actions on AI from the U.S. government, such as the NIST framework, U.S. AISIC, and U.S. EO on AI.

This week, we’ll continue examining the Gladstone AI report by vetting the "Survey of AI R&D Trajectories" - the companion explainer for the Gladstone Action Plan.

The text discusses the paradoxical nature of artificial intelligence (AI) as both a significant driver of economic growth and innovation, particularly in critical fields like medicine, energy, and climate science, and a source of potential risks equivalent to weapons of mass destruction due to its capability for weaponization, accidents, or loss of control. It highlights the astonishing pace at which AI capabilities have evolved, from struggling to generate coherent text to enabling high-risk activities such as automated phishing, identity theft, autonomous hacking, and potential bioweapon design. This rapid advancement is primarily attributed to increases in computational power and data, which have led to larger and more capable AI models.

Concerns are raised about the development of artificial general intelligences (AGIs) that could match or exceed human performance across all tasks, posing unprecedented risks if weaponized. The potential for such AI systems to execute catastrophic attacks or operate with autonomy in ways harmful to human welfare is noted as plausible within the next few years. The proliferation of AI technology is facilitated by both proprietary innovations by major labs and the AI community's open-source culture, leading to an irreversible spread of weaponizable AI systems. The text emphasizes the complex challenge that advanced AI poses to national security and public policy, calling for immediate attention to manage its dual-use nature and prevent catastrophic outcomes.

— Chat-GPT 4’s summary of Harris et al., “Survey of AI R&D Technologies”, Gladstone AI Inc. 2024

(If you have trouble accessing the link, please go to Gladstone AI’s website, and scroll to the “Survey of AI R&D Trajectories”)

Previous
Previous
3 April

Guest Nathan Henry Presents: “A Hormetic Approach to the Value-Loading Problem: Preventing the Paperclip Apocalypse?”

Next
Next
10 April

Guest Alex Spies Presents: “Structured World Representations in Maze-Solving Transformers”