Happy Tuesday!

In Today’s Email:

• Ex-Meta researcher’s AI startup 👨‍🔬

• Alibaba’s new AI models 👩‍💻

• AI in the Air Force 🛩️

Ex-Meta researcher’s AI startup 👨‍🔬

Former Meta researchers have launched an AI biotech startup.

The researchers had previously developed an AI language model for biology at Meta. Their newest venture, EvolutionaryScale, has already raised at least $40 million. The team at the startup is led by Alexander Rives, who previously ran the AI protein-folding team at Meta until it was shut down in April.

EvolutionaryScale has created a transformers-based model off of GPT-4 or Google’s Bard. The model is only trained on data about protein molecules, which allows it to predict the structure of other unknown proteins. The database powering the model already contains 700 million possible structures.

Use cases? The possible structures could be used to develop drugs and microorganisms. The startup has already pitched venture capitalists, with the initial round of funding being led by Lux Capitol. According to sources, prominent AI investors Daniel Gross and Nat Friedman also participated in funding rounds.

Alibaba’s new AI models 👩‍💻

Don’t count out Alibaba’s AI developments.

On Friday, the tech company launched two new AI models, that the company says can carry out more complex conversations and understand images. The two new models, Qwen-VL and Qwen-VL-Chat, are open-source. Similar to Meta, Alibaba is making its new models available to researchers, academics, and companies worldwide.

Qwen-VL-Chat can carry out more “complex interactions”. For example, the model can compare multiple image inputs and answer several rounds of questions. According to the company, the model can write stories and create images based on photos the user inputs, as well as solve equations shown in a photo.

Qwen-VL can respond to open-ended queries about different images and generate picture captions. Although we think other companies like OpenAI, Google, and Meta are further ahead with more advanced models, Alibaba shouldn’t be discredited. Their massive e-commerce infrastructure is a huge market for the use of generative AI tools.

AI in the Air Force 🛩️

The US Air Force wants to build using AI.

The Air Force is looking for $5.8 billion to build nearly 2,000 Valkyrie aircraft, powered by AI. The XQ-58A Valkyrie aircraft is meant to serve as an aid to human pilots, by maneuvering in situations that might be difficult and providing cover. The New York Times reported the aircraft would be ideal for suicide missions.

The aircraft is scheduled to be tested in a chase-and-kill simulation later this year. According to Business Insider, this model of the Valkyrie can easily cruise at 550 mph, with an operational altitude of 45,000 feet. This model has already been used as a datalink for F-22s, F-35s, and the Air Force’s Skyborg program.

The budgetary estimate lists $5.8 billion over the next five years to build the vehicles. The NYT reported each Valkyrie will cost far less than a manned pilot jet, only coasting $3-$25 million. This is all a part of the Air Force’s “Next Generation of Air Dominance” program which has the ultimate goal of fielding a “family of systems” to succeed the F-22 Raptor.

That’s it for today, we’ll see you back here tomorrow!