This article is more than 1 year old

Clearview AI plans tech to ID faces as they age, seek big government deals

Plus: AI imparts brain surgery skills more effectively than human tutors

In brief Controversial facial recognition startup Clearview AI plans to employ more staff in order to pursue big lucrative US government contracts worth many millions of dollars. 

CEO Hoan Thon-That told Reuters Clearview's current annual contracts with its 3,100 customers are relatively small.

"We know that some of these agencies are seeing great success, but they are only at a small five-figure purchase or a six-figure purchase. And so it's 'Can we get a few of those to the seven-figure, maybe eight-figure purchases?'."

To pursue more substantial projects, Clearview AI will increase its size by a third and build new capabilities such as matching photographs of people in their younger and older years to improve identification.

Clearview is best-known for scraping personal images of people from social media platforms like Facebook or Instagram, as well as picture-sharing sites like Flickr or Getty Images. That practice earned the company legal problems in the US and Canada.

Will RISC-V chips go on to dominate AI?

The number of chips based on the RISC-V architecture is predicted to grow 73.6 per cent per year until 2027, the majority of which will power AI and machine learning software, according to research and consulting group Semico.

Hardware startups building custom AI accelerators are turning to RISC-V's open source blueprints to avoid paying the licensing fees required when using x86 and Arm architectures. RISC-V's economical instruction set also allows chip designers to build processors that are smaller and require fewer transistors. The resulting products are more power efficient than rivals. 

Arm-based chips remain the market leader for AI hardware, with RISC-V designs accounting for just 15 per cent of total revenue for CPU core architectures. 

Semico research principal analyst Rich Wawrzyniak told IEEE Spectrum that RISC-V is expanding rapidly. "It's not 50 per cent, but it's not five per cent either. And if you think about how long RISC-V has been around, that's pretty fast growth." Around 25 billion machine learning chips are forecast to be built by 2027, an industry totaling some $291 billion.

AI algorithms are better at teaching students how to perform brain surgery than remote human instructors

Medical students' learning curves improved when using a neurosurgical simulator and a machine learning coach to study removal of virtual brain tumors, according to a study published in JAMA Network last week.

A group of 70 students from McGill University, Canada, were sorted into three different groups. One received instructions and feedback from remote human tutors who guided them through the model procedures. Others were taught by an AI system known as the Virtual Operative Assistant (VOA), while a third group were given no help at all.

Researchers found that students picked up on surgical skills 2.6 times faster and had 36 per cent better performance when they learned from the VOA, compared to those that were taught remotely by real experts. 

"Artificially intelligent tutors like the VOA may become a valuable tool in the training of the next generation of neurosurgeons," said Rolando Del Maestro, senior author of the study and a researcher at the Neurosurgical Simulation and Artificial Intelligence Learning Centre. ®

More about


Send us news

Other stories you might like