Sponsored by

Happy Thanksgiving and welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each Thursday, we aggregate the major developments in artificial intelligence; we pass along the news, useful resources, tools or services, and exciting projects in open source. Even if you aren’t an engineer, we’ll keep you in touch with what’s going on in AI.

• On Friday, OpenAI made an official blog post about a leadership transition change, with CTO Mira Murati taking over for Altman as interim CEO. The board claimed Altman was not consistently candid in his communications, and that they no longer had confidence in his ability to continue leading OpenAI. They also demoted Brockman, who immediately resigned. Other major senior researchers also immediately resigned.

• On Monday, Microsoft CEO Satya Nadella announced that Altman, Brockman, and other researchers from OpenAI would be joining Microsoft to lead a new AI division.

• Later on Monday, a staggering 95% of the company’s 770 employees signed a letter threatening to resign or follow Altman to Microsoft unless a) Altman and Brockman were reinstated and b) a new board be established.

• The letter seems to have been a successful play. Altman and Brockman returning appears to be a done deal at this point. There will be a new board, and three of the four members that initiated the ouster of Altman, including head scientist Ilya Sutskever, have been removed.

Little information has been revealed about the original decision to remove Altman, but rumors are swirling. Some are speculating that Project Q*, a mysterious discovery at OpenAI, was a major research breakthrough towards artificial general intelligence that may not have been handled properly internally. We’ll provide more updates as they come to light.

Some updates and coming upgrades happening with popular ChatGPT competitors:

Anthropic has released Claude 2.1, the latest version of its AI assistant, with several major upgrades including a 200,000 token context window, reduced rates of hallucination, and new tool use capabilities. Claude already had the largest context window in the business at 100,000 tokens; 200,000 is equivalent to 500 pages of text. Claude 2.1 is also more accurate and reliable in its outputs - a 30% drop in incorrect answers and 3-4x lower rates of unsupported claims.

Inflection is about to upgrade their popular chatbot Pi with Inflection-2, which they claim is the second most capable LLM in the world behind GPT-4. It exceeds Inflection-1 in factual knowledge, stylistic control, and reasoning abilities. Although not a primary focus, Inflection-2 showed strong capabilities in code and math reasoning benchmarks - outperforming all peers except GPT-4. It can be fine-tuned for enhanced coding abilities.

Google’s Bard can now chat with you and answer questions about YouTube videos.

UnitedHealth, the largest health insurance company in the U.S., is facing a lawsuit for allegedly using a flawed AI algorithm which overrides doctors' judgments to deny critical health coverage to Medicare patients. This situation is part of a broader issue in the healthcare industry where AI use often results in biased and inaccurate outcomes. The lawsuit, seeking class-action status, highlights instances where coverage denials based on UnitedHealth’s ‘nH Predict’ algorithm were overwhelmingly overturned on appeal.

UnitedHealth insists that the tool is only for guidance and that coverage decisions adhere to plan terms and the standards and guidelines set by the Centers for Medicare & Medicaid Services (CMS), a federal agency within the United States Department of Health and Human Services. The lawsuit accuses UnitedHealth and NaviHealth of multiple legal violations, seeking damages and an end to the AI-based denials.

ChatGPT with voice chat is now available to all app users, including the free tier. To get started with voice, go to Settings → New Features on the mobile app, and opt into voice conversations. The new voice capability is powered by a new text-to-speech model, capable of generating human-like audio from just text and a few seconds of sample speech. OpenAI collaborated with professional voice actors to create the voices and uses Whisper, their open-source speech recognition system, to transcribe spoken words into text.

More in AI this week:

• EU AI Act: Germany, France and Italy reach agreement on the future of AI regulation in Europe

• Founder and CEO of GM’s self-driving car unit resigns in wake of safety problems

• FTC eases AI use investigation process

• ChatGPT Plus subscriptions are being listed on eBay after OpenAI blocked signups

• Amazon aims to provide free AI skills training to 2 million people by 2025 with its new ‘AI Ready’ commitment

• OpenAI employees say chief scientist has been making strange ‘spiritual’ claims

• What the OpenAI drama means for AI progress — and safety

Harness The Economist's approach to business writing

How do you write persuasively for business? Study the psychology, craft and purpose of writing with experts from The Economist.

The six-week Professional Communication: Business Writing and Storytelling online short course from The Economist is designed to help business professionals attract and address broad audiences across various written formats. 

On completion of this course, you’ll walk away with:

• Practical writing, editing and data-visualisation skills for creating clear and compelling business communications.

• A writer’s toolkit based on The Economist’s editorial approach, helping you to navigate questions of structure, word choice and style.

• The ability to source, structure and convey information in a way that is accurate, engaging and ethical.

We hope you have an amazing Thanksgiving! See you next week.