AI Leadership Weekly

Top Stories

Hallucinating GPTs in Health Care
It's well known that LLMs hallucinate answers, even when trained on custom data. But, it's being reported that hospitals are still using OpenAI's Whisper transcription tool to deal with patient records, which is setting off alarms; the worry being that the wrong medical information could end up in patient records.

AI Companies Investing Heavily In Robotics
AI models are fast expanding from being just "online" and accessible through a browser, and moving into the physical world of robotics. The San Francisco-based robotics company, Physical Intelligence, has increased its March $70 million valuation to $2.4 billion thanks to a $400 million funding round.

LLMs Being Used in the Defense Industry
Meta has altered its policies to now allow the use of their Llama models within the defense industry. Right now, for example, Oracle is training a Llama model to produce aircraft technical manuals, which they hope will speed up repairs.

In Brief

• Meta’s head of AR glasses joining OpenAI to work on robotics projects

• Chinese researchers create Infinity-MM dataset to train advanced models

• Apple’s new App Intent APIs can read and react to things on your screen

Market Trends

AI Power Goes Nuclear
Energy demands for AI datacentres is only increasing, and many tech companies are turning to nuclear power to meet their needs. Meta was a struck a recent blow, though, when the proposed site for a new datacentre was deemed to impact a rare species of bee.

Google Expanding AI Hub Into Saudi Arabia
AI tech is continuing to expand into non-English-speaking markets, with Google opening up a new AI hub in Saudi Arabia. This particular hub will focus on Arab language AI models, as well as "Saudi-specific AI applications".

AI Regulations Incoming
Anthropic is urging lawmakers to enact regulations around the use of AI, and warn that they don't have long to act to avoid disaster.

Tools and Resources

v0: Interfaces from Screenshots
Need to rapidly prototype or iterate on an MVP? v0 will turn a simple screenshot into a working app. You can further customise this app with custom code, and by using prompts to request special features.

Melies: Short-form AI Video
This nifty webapp walks you through a series of prompts to first generate a logline for a film, then the characters, then, finally, a short film. At each step you're able to directly edit the text it produces.

neural.love
Here you can create high-quality art from your simple prompts, and without any design skills.

Recommended Reading

Human-Compatible AI: Alignment Problems and Solutions
Stuart Russell, a leading AI researcher, explores how to ensure AI systems remain aligned with human values and interests as they move into critical domains. He provides practical frameworks for thinking about AI safety and deployment in high-stakes environments.

Normal Accidents: Living with High-Risk Technologies
While written before the AI era, this book's insights about how complex systems inevitably produce failures is remarkably relevant for today's AI leaders, especially as AI moves into healthcare, robotics, and defense. Perrow's analysis of how tightly coupled systems can produce cascading failures offers valuable lessons for AI deployment in critical sectors.

Hit reply to let us know which of these stories you found the most important or surprising! And, if you’ve stumbled across an interesting link/tweet/news story of your own, send it our way at [email protected] It might just end up in the next issue!

Thanks for reading. Stay tuned for the next AI Leadership Weekly!

Brought to you by Data Wave your AI and Data Team as a Subscription.
Work with seasoned technology leaders who have taken Startups to IPO and led large transformation programmes.