Assessing AI Risks and Governance

Our 148th episode with a summary and discussion of last week's big AI news!

Read out our text newsletter and comment on the podcast at https://lastweekin.ai/

Email us your questions and feedback at contact@lastweekin.ai

Timestamps + links:

• (00:00:00) Intro / Banter

• Tools & Apps
  • (00:02:43) Google Deepmind unveils its most advanced AI image generator, Imagen 2
  • (00:08:21) Anthropic will help users if they get sued for copyright infringement
  • (00:13:50) Midjourney Alpha is here with AI image generations on the web
  • (00:16:34) Instagram introduces gen-AI powered background editing tool
  • (00:17:09) Microsoft drastically expands Azure AI Studio to include Llama 2 Model-as-a-Service, GPT-4 Turbo with Vision
  • (00:18:54) ChatGPT Is Apparently Becoming Lazy as It Has Started Asking Users to Solve Their Own Problems
  • (00:22:17) You can create your own AI songs with this new Copilot extension
  • (00:23:57) Stability AI announces paid membership for commercial use of its models
• Applications & Business
  • (00:25:42) ByteDance is secretly using OpenAI’s tech to build a competitor
  • (00:31:55) Intel unveils new AI chip to compete with Nvidia and AMD
  • (00:36:36) Chinese chip-related companies shutting down with record speed — 10,900, or around 30 per day, shut down in 2023
  • (00:40:11) TSMC mentions 1.4nm process tech for the first time, says 2nm remains on track
  • (00:42:57) Meta has done something that will get Nvidia and AMD very, very worried — it gave up on GPU and CPU to take a RISC-y route for AI training and inference acceleration
  • (00:46:17) Nvidia rushes to deliver modified AI GPU chips to China customers, allegedly places 'Super Hot Run' priority order with TSMC
  • (00:49:17) Sam Altman’s OpenAI agrees to pay German media giant Axel Springer for using its content to train AI models
• Projects & Open Source
  • (00:52:20) Introducing DeciLM-7B: The Fastest and Most Accurate 7 Billion-Parameter LLM to Date
  • (00:57:27) Introducing Stable Zero123: Quality 3D Object Generation from Single Images
• Research & Advancements
  • (01:00:42) FunSearch: Making new discoveries in mathematical sciences using Large Language Models
  • (01:09:12) OpenAI Demos a Control Method for Superintelligent AI
  • (01:16:41) Cheating Fears Over Chatbots Were Overblown, New Research Suggests
  • (01:18:13) SwitchHead: Accelerating Transformers with Mixture-of-Experts Attention
  • (01:20:04) CogAgent: A Visual Language Model for GUI Agents
  • (01:21:10) Limits to the Energy Efficiency of CMOS Microprocessors
• Policy & Safety
  • (01:24:24) OpenAI announces ‘Preparedness Framework’ to track and mitigate AI risks
  • (01:32:08) Pro-China YouTube Network Used A.I. to Malign U.S., Report Finds
  • (01:37:07) AI is a danger to the financial system, regulators warn for the first time
  • (01:38:42) Anonymous Sudan hacking group sets sights on ChatGPT
  • (01:40:33) Scenario planning for an AGI future
  • (01:42:51) The widening web of effective altruism in AI security
• Synthetic Media & Art
  • (01:49:22) Facebook Is Being Overrun With Stolen, AI-Generated Images That People Think Are Real
  • (01:52:34) Pakistan’s former prime minister is using an AI voice clone to campaign from prison