AI Leadership Weekly

Issue #13

Brad
January 09, 2025

Welcome to the latest AI Leadership Weekly, a curated digest of AI news and developments for business leaders.

Top Stories

Source: Nvidia

Nvidia Announces “Personal Super Computer”
Amongst its many CES 2025 reveals, Nvidia has announced "Digits", which it calls a personal AI supercomputer.

The device itself is a small mini-computer that can sit on your desk, and has a starting price of $3,000. It is also said to run LLM models up to 200 billion parameters thanks to having 128GB of unified RAM. A common complaint of Nvidia's own 40-series and the just-announced 50-series of RTX GPUs is their lack of VRAM, limiting their use in AI applications.

Users will also be able to link multiple Digits machines together to run even larger models.

Source: Jason Redmond/AFP/Getty Images

‘Virtual Employees’ Coming This Year?
OpenAI's CEO, Sam Altman, has written on his blog that virtual employees could start appearing in companies as soon as this year.

This is on the back of Microsoft (a large investor in OpenAI) releasing 'AI agents' last year, which are tools that can carry out various autonomous tasks, as well as the release of OpenAI’s ‘operator’, which is rumoured to be coming by the end of January.

Source: Christophe Gateau, picture alliance/Getty Images

Power a Concern for Musk Supercomputer
xAI recently announced their intention to expand their Memphis AI-training facility by 900%. Currently, the facility houses 100,000 GPUs which it uses to train the Grok AI model. But, they plan to expand this to 1 million GPUs in the near future.

The problem, however, is that Memphis isn't sure it can supply the power needed. Other AI and big-data companies are experiencing the same issues, with other companies (such as Meta) looking to build or acquire nuclear power stations to meet their own demands.

In the words of the CEO of Memphis' utility company, “It is a physics problem, not a political problem, about how much energy can be provided here.”

In Brief

Market Trends

Nvidia GPUs Leaning Harder on AI
One of the biggest announcements at CES 2025 was Nvidia's reveal of their 50-series of GPUs. The flagship 5090 with its 32GB of VRAM is what most AI enthusiasts will flock to, although there has been grumbling for months over the rest of their line-up and its lack of enough VRAM.

That said, their claimed gaming performance is very impressive, with CEO Jensen Huang claiming that the new 5070 will match the performance of the 4090. That trick is achieved with all of the AI upscaling built into the cards, and which they have been developing since 2018.

Nvidia COSMOS Focuses on Robotics
Nvidia is the latest AI-related company to venture into the robotics and hardware realm, but from a different angle.

Their COSMOS platform aims to generate large amounts of training data which can then be used by third-parties to train robot AIs. The goal is to be able to generate, amongst other things, edge-case scenarios that are hard to predict or replicate in the real world, then allow AIs to interact with and learn from them.

Microsoft Investing Billions Into India
Microsoft has announced plans to invest upwards of $3 billion in India to expand its offerings of cloud and AI services. Their goal, of course, is to court more South Asian companies onto their platforms and expand their reach in the region.

Amazon has also indicated investing nearly $13 billion in India by 2030. This is similar to the story we reported recently where Chinese AI companies are currently slugging it out for market share in developing regions.

Tools and Resources

GitPodcast
Enter the URI for a GitHub repo and then received either a 5- or 10-minute "podcast" to help you quickly understand the project.

Journals AI
An AI-powered, guided journal to help get your thoughts down.

WikiTimeline
Turn any Wikipedia article into an interactive timeline.