You added the ethics, right?

Not my job

Amazon, Google, Microsoft, Tesla and other AI leaders set up shop inside one of the world’s most important computer vision conferences in June in New Orleans.

• They were on the hunt for computer vision talent: interns who could help build tomorrow’s medical and autonomous vehicle software products.
• But while big tech talks up the importance of ethical AI, the young computer scientists at the Computer Vision and Pattern Recognition Conference are immersed in a learning environment that is resistant to ethical considerations.

This year’s conference marked the first time CVPR organizers “strongly encouraged” researchers to consider negative societal impacts of their research.

• Sure, computer vision holds promise for finding disease faster and making autonomous vehicles safer, but it also enables discriminatory surveillance tech, controversial emotion AI and disinformation-spreading deepfakes.
• Some researchers I spoke to at CVPR pushed back. One told me researchers “are not good at telling you what the applications of their research are. It’s not their job.”
• Another helping advance emotion AI said, “It’s not the techniques that are bad; it’s the way you use it."

Resistance among researchers hindered goals of CVPR organizers to move the computer vision community toward a more ethics-friendly approach to their work.

• Although some researchers acknowledged possible downsides of their work in papers presented at CVPR — including research that could be used to improve deepfakes for spread disinformation — many did not.
• CVPR program chairs didn’t push too hard. ““The community as a whole is not at this point yet. If we make a very radical change, then the reviewers will not really know how to basically take that into account in the review process,” Stefan Roth, a CVPR chair and computer science professor at Germany’s Technical University of Darmstadt, told me.

But AI ethics proponents balked.

• “That is exactly what pisses me off,” said Timnit Gebru, founder and executive director of the Distributed Artificial Intelligence Research Institute, and a researcher with a Ph.D. in computer vision. “[Computer vision researchers] have convinced themselves that it’s not their job.”
• Meanwhile, far away in Seoul, the top AI ethics conference — FAccT (Fairness, Accountability and Transparency) — was held the same week.
• But hundreds of computer vision researchers from Asia, many from South Korea, were thousands of miles away advancing their craft with relatively minimal concern for the societal implications of their work.
• Alice Xiang, a general co-chair of this year’s FAccT conference and head of Sony Group’s AI ethics office, told me, “We do sometimes worry about whether practitioners who actually develop AI technologies might feel that this is just a conversation for ethicists.”

Read my full dispatch from the CVPR conference and learn why ethics at the research level matters for AI businesses.

— Kate Kaye (email | twitter)

SPONSORED CONTENT FROM DATASTAX

You're either real-time or out of time: Applications like this power our daily lives, and if they can’t access massive amounts of data stored in a database as well as stream data “in motion” instantaneously, you — and millions of customers — won’t have these in-the-moment experiences.

Read more from DataStax

The fools on the Hill

Fifty-two billion dollars in proposed funding for U.S. chip manufacturing subsidies that in January looked like it would sail through Congress is no longer a sure bet. It has been languishing in congressional reconciliation for weeks and has transformed from a widely supported piece of legislation into a partisan football.

The rub might be that, like it or not, the chip subsidies have never been at the top of the congressional priority list. Its fate has been tied to USICA, a law that Senate Minority Leader Mitch McConnell has vowed will not pass if the “Democrats are pursuing a partisan reconciliation bill.” If some form of the bill fails to get to the president’s desk by the August congressional recess, the chip funding is all but dead until the new Congress begins governing in 2023.

The chip giants who have announced tens of billions of new or expanded chip fabrication facilities in the U.S. are mad. But anger hasn’t translated into a change of plans yet: Intel delayed the ceremonial start to construction at its Ohio site, but the company has begun building the new chip factory anyway. A TSMC board member suggested construction — which is already underway at its Arizona site — may slow if lawmakers don’t get their act together and pass the subsidy package. Samsung, too, has begun construction at its Texas fab.

Financial corner

Coalition was valued at $5 billion after raising $250 million for its cybersecurity software and insurance products.

Flexe was valued at $1 billion after raising $119 million for its logistics software.

Swimlane raised $70 million for its cybersecurity automation platform.

Around the enterprise

GlobalFoundries and STMicroelectronics announced plans to build a new chip factory in France as a joint venture. It’s expected to open by 2026.

SPONSORED CONTENT FROM DATASTAX

You're either real-time or out of time: Many of the challenges facing our world today are increasingly complex and critical, such as climate change, talent shortages and supply chain disruptions. Solving these problems requires analyzing large data sets, quickly. Additionally, organizations must use data to predict future issues and then determine the most effective solution.

Read more from DataStax