💰 OpenAI needs Microsoft's money

OpenAI is banking on Microsoft’s wallet

OpenAI **is focused on its development of artificial general intelligence (AGI) - and its confident **Microsoft will continue to foot the bill.

OpenAI CEO Sam Altman has given an interview to the Financial Times, noting his company’s relationship with Microsoft is “working really well,” and that he hopes to raise “a lot more over time” from the tech giant.

Altman also commented on the challenges that OpenAI faces in achieving AGI, stating there is “a long way to go,” and that “training expenses are just huge.”

Altman acknowledged that AI training costs were getting in the way of OpenAI reaching profitability, but suggested that the partnership with Microsoft is good for both as they “both make money on each other’s success.”

**Why it matters: **Microsoft has already sunk more than $10 billion into OpenAI, and based on what Altman is saying, that figure will only get bigger and bigger. But like any publicly traded company, it will be shareholder reactions that ultimately shape Microsoft’s decisions - including how much cash to put into an enterprise that is yet to turn a profit.

Financial Times | Link to story (paywall).

Autodesk rolls out AI tools for its maker apps

Design and manufacturing software company **Autodesk **launches Autodesk AI, a set of tools that add generative capabilities to its suite of products.

Perhaps most notable from the announcement is that Autodesk’s AutoCAD is getting an AI feature for interpreting handwritten notes and digital markups. As Autodesk explains, the tool can take that information and incorporate design changes, as well as make context-aware recommendations for further adjustments.

Autodesk is rolling out AI features across its suite of apps now.

Why it matters: Autodesk has long been a leader in the maker software space, so the embrace of AI could see it firm up that position for a while yet.

Yahoo Finance | Link to story.

Nvidia announces next-gen AI chip

**Nvidia **has announced H200, a GPU aimed at the AI industry.

A successor to the H200 features 141GB of HBM3 memory, promising better performance when working with LLMs. Nvidia claims the chip can perform tasks twice as fast as its predecessor, the H100. 

The H200 is expected to ship in the second quarter of 2024, and it will be compatible with models trained on the H100. Nvidia is yet to disclose a price.

**Why it matters: **The H100 has been one of the most sought after GPUs for AI firms, with the chip costing anywhere from $25,000 to $40,000. It’s also of note that **OpenAI **made use of the H100 when training its GPT-4 model.

CNBC | Link to story.

Researchers work to stop deepfakes

NPR profiles AntiFake, a tool that aims to disrupt the production of deepfake voices.

The tool, which is still in the building phase, promises to pass uploaded audio through a custom process that makes AI analysis difficult while maintaining a natural sound to human ears.

Ning Zhang, an assistant professor of computer science and engineering at Washington University, and part of the team working on AntiFake, says the aim of the tool is to prevent an “AI-based synthesize engine from generating an effective copycat.”

The research team have not said if or when AntiFake will be made publicly available, though they are planning to present their work at the ACM CCS security conference later this month.

**Why it matters: **Celebrity deepfakes are not a new phenomenon, but with cultural and political divides seemingly wider than ever, the potential weaponization of them has become a concern for many around the world.

NPR | Link to story.