What's Happening in AI

First-of-its-kind AI bias lawsuit, GPT-4 for content moderation, AGI in ‘2-3 years’, and more. First, this week’s big story: White House calls for China tech ban

• Why this matters: This Executive Order will guide rule-making on U.S. investment into ‘technologies and products (e.g., semiconductors and microelectronics, quantum information technologies, and AI) in countries of concern’ (China, Hong Kong SAR, Macau SAR).
• What’s next: Treasury lays out proposed rules
• Consider: PE, VC investors on Biden's China tech ban: We're ready
• Comments: “It’s the right thing to do”, “everything could be AI”, “increasingly complex”

Policy

Americas

• EEOC settles first-of-its-kind AI bias lawsuit and U.S. state rules begin to take shape
• Senator Warner outlines what he wants to know about Google’s Med-PaLM2 deployment in healthcare
• DOD announces establishment of generative AI task force
• Waymo, Cruise get approval to offer 24/7 paid robotaxi rides in SF
• ICYMI: The state of state AI laws (more here)

Europe and Africa

• Update on U.K. PM Rishi Sunak’s plans for an AI summit
• EU tech envoy: ‘The winds have changed’ on regulating Silicon Valley
• Nigerian tech leaders cheer insider’s rise to president’s cabinet as ‘watershed’ moment
• ICYMI: TikTok on fulfilling commitments under the EU’s Digital Services Act

Asia and Australia

• How does generative AI learn? Japan prepares disclosure rules
• Google appeals using online content for AI training unless firms opt-out
• Guide to China’s 2023 major research program on AI methods
• ICYMI: ​​China’s Interim Measures for the Management of Generative AI Services took effect Aug. 15

Responsible AI

Data, information, and trust

• Using GPT-4 for content moderation (more here)
• Why watermarking AI-generated content won’t guarantee trust online
• Why it’s impossible to build an unbiased AI language model
• Poll: Americans distrust AI giants

Health and science

• Ex-Google CEO Eric Schmidt to launch AI-science moonshot
• Recursion CTO: AI supercharges drug discoveries, potentially changing medicine
• ML is useful for many things, but not for predicting scientific replicability
• ICYMI: Demo of IBM and NASA’s new climate model (more here)

Security and privacy

• 'Hypnotized' ChatGPT and Bard will try to convince users to pay ransoms
• Zoom rewrites its policies to make clear that your videos aren't used to train AI tools
• ICYMI: Why companies are turning their chatbots over to hackers at DEF CON

Resources

• How to match LLM patterns to problems (more here)
• Livestream with HuggingFace to review the most recent AI models
• AI 101 for Teachers

Where AI is going

• Interview: Anthropic’s Dario Amodei says AGI in ‘2-3 years’
• Why 2024 will be the deepfake election

Launches, partnerships, and updates

• Google’s search updates
• Anthropic raises $100M and Weights & Biases raises $50M
• Nvidia teams up with HuggingFace

Workforce

• Jobs are for humans
• AI can’t build a high-rise, but it can speed up the job
• Robots to the rescue? (more here)

Compute and chips

• Alibaba chief warns of constraints as China AI training ramps up
• Nvidia set to hop AI forward with next-gen Grace Hopper Superchip
• U.S. Commerce Department is reaping hundreds of Chips Act inquiries
• ICYMI: CoreWeave and Tenstorrent close rounds; Nvidia H100 GPUs: supply and demand‍
• Financialization of AI (ex. raising money on the basis of GPU allocations), commentary by Anthropic’s Jack Clark