top of page

Preparing for AI

Your weekly CogX newsletter on AI Safety and Ethics

The week's developments on AI and regulation, explained | 12.01.24

If you’re enjoying this briefing, sign up here — and share it with a friend. Reading time: ~4 minutes


Last year, secret Sino-US diplomacy talks took place between OpenAI, Anthropic and leading Chinese AI experts. The discussions, centered around AI risk and global cooperation,  were a success.  . Despite this, US regulations restricting high-performance chip exports have been introduced for China, raising crucial questions about AI’s role in international diplomacy and geopolitical tensions. 


Shifting focus across the pond, the UK’s 2024 general elections are at high risk from deepfakes, according to a former cabinet member. Will the current detection tech be good enough to mitigate? 


Plus, checkout our AI Dilemma’s section to find out how the controversial “death algorithm” could actually be a good thing and whether an AI tax is a fair solution to job-displacement fears.


We cover these stories, and AI experts’ 48-year prediction revisions, the latest research on an LLM theory of the mind and an AI-enabled political campaign from behind bars.

- Charlie and the Research and Intelligence Team


Ethics and Governance



🤫OpenAI, Anthropic, and Cohere secretly met with Chinese AI experts in Geneva, a rare sign of Sino-US collaboration in advanced tech. The discussions focused on AI risk and cooperation, and future talks are set to cover AI alignment and regulation. 


🤖Bad news for Nvidia: will China buy their downgraded chips? New US regulation has limited exports on high-performance chips. Now, Nvidia is unable to sell its most capable products to a huge percentage of their market, and the ones it can sell, China isn't interested in. 


🗳️Deepfake videos could disrupt the UK's 2024 General Election, warns Former Cabinet Secretary Lord Gus O’Donnell. His warning follows recent targeted deepfake attacks against British politicians, and current shortfallings in deepfake detection technology. 


📈What's next for AI regulation? 2023 was the year of debate, and 2024 will be the year of action. The US will focus on transparency, the EU will enforce its AI Act and China will lay out comprehensive AI laws. Globally? Expect diverse approaches, and emerging regulation.


 

AI Dilemmas


💰Calls for an AI tax grow, amidst rising concerns of job losses. Mass job displacement may be coming, and taxing AI firms will be a way to address societal impacts and balance the scales. International cooperation and debate on tax rate are crucial to tackle this issue promptly.


⏳Is an AI death calculator a good thing? The 'life2vec' algorithm can predict a person's life course and death with high accuracy. While the model is controversial, it has the potential to uncover how social factors influence health and could inform policies to improve life expectancy..


 


Insights & Research



📊AI experts revise predictions by 48 years for a fully-automated labor market, and 13 years for high-level machine intelligence, according to a survey of 2,778 experts. Results also show that tech zoomers outweigh tech doomers yet most respondents fall somewhere in between.


🧠LLMs may have a theory of the mind. This study assesses whether LLMs can understand and predict how humans perceive robot behavior. Initial results suggest LLMs do possess a potential theory of mind, but final conclusions have not yet been made.


🤖Rodney Brooks bursts the ‘AI and self-driving cars’ hype bubble. He highlights that despite predictions, there are still no fully autonomous vehicles on the road. Brooks warns that technological progress often takes longer than expected and suggests an AI winter may be looming. 


✍️Researchers will be the ones to fill the gaps in the EU AI Act, particularly in defining high-risk AI. Experts state that collaboration with the scientific community is essential to address the regulatory issues within the limited timeframe before the Act's full implementation.


In case you missed it


Pakistan’s former Prime minister Imran Khan uses AI to campaign from behind bars. Check out his AI generated speech here:



✍️ Enjoying this newsletter?  Subscribe to our free weekly briefings on Preparing for AI, Cinema & AI, The Race to Net Zero, AI & DeepTech and The Future of Work.

bottom of page