7 OCTOBER | LONDON 2024
SEPTEMBER 12TH - 14TH
The O2, LONDON
Preparing for AI
Your weekly CogX newsletter on AI Safety and Ethics
The week's developments on AI, safety and ethics, explained | 07.06.24
If you’re enjoying this briefing, sign up here — and share it with a friend. Reading time: ~5 minutes
This week, UK financial watchdog delays on detailed AI rules stating current frameworks suffice. Meanwhile, US antitrust enforcer calls for urgent regulation to curb BigTech's control over AI. Fearmongering, or founded in reality?
Meanwhile OpenAI, Anthropic and DeepMind employees warn AI could lead to "human extinction" without stronger safeguards, Meta’s controversial new policy trains AI on user data (its opt-out form offering limited protection) and Google’s AI search is copying people’s work.
We cover this, plus, a ‘future self’ chatbot for sage life advice, AI that decodes dog barks and what GPT-4 gets like when it's angry.
- Charlie and the Research and Intelligence Team
P.S You are now able to sign up for the super early bird offer (75% off) to the CogX AI Summit in London on October 7th here.
Ethics and Governance
🛡Detailed AI rules not on the cards for now, says UK financial watchdog. The FCA is hesitant to implement rules immediately, due to the rapid pace of technological change, believing current frameworks suffice to hold senior managers accountable.
⚖️US antitrust enforcer calls for urgent scrutiny over BigTech’s control of AI stressing the need for immediate regulation, warning competition in the AI sector could be stifled without prompt action. He's investigating compute, data, cloud services, and hardware.
🆘OpenAI workers raise alarm about AI risks: Current and former employees from top AI firms warn that AI could pose a threat of "human extinction" without stronger safeguards. They call to protect researchers who voice concerns and public involvement in development.
📷Instagram and Facebook train AI on your data - here’s how to opt out: Meta’s new policy will take users' data, including photos and posts, to train AI systems. Users can object by filling out a form on Instagram’s website, but Meta may still use their data in some cases.
AI Dilemmas
⚠️’Google’s AI Search copied my work’: Google's new AI bumped this author’s article down but still referenced it without proper attribution. This raises concerns, as such features can reduce traffic to original sources and diminish incentive to click through to full articles.
🤖AI solutions to loneliness: As loneliness becomes a growing concern, tech companies are proposing AI solutions. Apps like Replika offer digital companionship through chatbots, offering immediate emotional support, despite scepticism about replacing human interaction.
🚀Enjoying our content? We’re also on LinkedIn — Follow us to stay in the loop with the week's most impactful AI news and research!
Insights & Research
🤖AI researchers build ‘future self’ chatbot to inspire wise life choices: MIT scientists can now create a digital image of a user as an older adult, with plausible future memories, to promote long-term thinking about health, finances, and personal goals.
🔍Can AI models trained on human speech help us understand dogs? Yes, according to Researchers at the University of Michigan who’ve developed an AI that identifies whether a bark conveys playfulness or aggression and can indicate the dog's age, breed, and sex.
🌐 Will AI become death, destroyer of worlds? MIT economists warn AI could displace workers and lower wages, but also has the potential to enhance productivity and create jobs. The future hinges on whether it prioritises automation or human augmentation.
🧠Researchers found emotions enhance LLM performance. While GPT-3.5 aligned well with human responses, GPT-4 showed consistent rational behaviour even under emotional influence. Notably, inducing anger in GPT-4 disrupted its performance.
In case you missed it
‘I asked AI where it thought it was from. It’s answer? Nowhere.’ In this short film, visual artist Nouf Aljowaysir examines a question she’s struggled with since childhood.
✍️ 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
🚀 You are now able to sign up for the super early bird offer (75% off) to the CogX AI Summit in London on October 7th here.