Lab explores using artificial intelligence to help police catch criminals – EFF • The Register

Briefly The Pacific Northwest National Laboratory is researching how artificial intelligence techniques can be used to create a “Digital Police Officer” or “D-PO” of the future.

Freedom of information requests from the Electronic Frontier Foundation show that the US Department of Energy-funded lab envisions that police might one day be able to partner with a virtual assistant to fight crime. D-PO It would be able, for example, to eavesdrop on facial recognition systems to alert a police officer on patrol to a suspect nearby, and could also advise on how best to apprehend the suspect.

The EFF warned against scrambling with software like D-PO, citing concerns about inaccurate facial recognition matches and biased predictive police policies.

“The good news is that in the emails we got, one of the authors acknowledges in the internal emails that items like D-PO taking command are ‘out of reach’ and that monitoring the live drone feed is ‘not a near-term capability,'” The nonprofit organization focused on digital privacy He said.

The National Laboratory also described how a separate virtual assistant, BITS, could provide US border and customs agents with visual data to help them crack down on drug smuggling. “Records received by the EFF do not indicate any official interest from CBP or the Department of Homeland Security. However, BITS may not be as far into the future as D-PO,” she said.

Uber is back in the self-driving taxi game

Transportation giant Uber has announced a 10-year deal with self-driving car startup Motional Inc. to develop self-driving taxis.

Uber scrapped efforts to develop its own self-driving vehicles in-house, and sold its Advanced Technology Group unit to self-driving biz Aurora in 2020. Now, it’s trying to revive efforts to get back into the robotics business by hiring a third-party vendor to build the necessary cars.

Motional will provide the devices, while Uber will deliver them to passengers. “This agreement will be instrumental in the widespread adoption of the robot hub,” said Carl Enanema, President and CEO of Motional, He said in the current situation.

“Motional now has unparalleled access to millions of passengers and a roadmap to expand exponentially over the next 10 years. We are proud to partner with Uber to bring both driverless transportation and delivery to life in cities across the United States.”

Both companies have worked together before. Motional launched tests of its IONIQ 5 self-driving electric vehicle to deliver food to Uber Eats customers in Santa Monica, California in May.

Optimize PyTorch Models Across AMD and Nvidia GPUs

Researchers at Meta have released AITemplate, a new framework aimed at accelerating the inference process for AI models running on GPUs from Nvidia and AMD.

Developers have to rewrite code if they want to improve their models on different devices, which makes it a huge daunting task to switch. However, AITemplate (AIT) gives them more flexibility by converting neural networks written in PyTorch to CUDA or HIP C++ code to support Nvidia’s A100 and AMD’s MI200 GPU systems.

“With AIT, it is now possible to run efficient heuristics on hardware from both GPU providers,” Meta He said In a blog post. “We used AIT to achieve performance improvements of up to 12x on NVIDIA GPUs and 4x on AMD GPUs compared to the passionate mode within PyTorch.”

The developers plan to expand AIT to support Apple’s M-series GPUs as well as CPUs. You can access the program source over here.

Amnesty International US Bill of Rights

The White House this week proposed the Artificial Intelligence Rights Act, a non-binding scheme to help agencies, the private sector and citizens guide the development and deployment of algorithms.

People should have the right to be protected from unsafe systems and discrimination and their privacy should be respected. They should also know when and how their personal data is used, and should be able to opt out of automated services if possible.

“Automated technologies are driving great innovations and shaping important decisions that affect people’s rights, opportunities, and access. The AI ​​Rights Act blueprint is for everyone who interacts daily with these powerful technologies—and everyone whose lives have been changed by unaccountable algorithms,” Alondra Nelson, deputy director of the Office of Policy Science and Technology , He said in the current situation.

The bill is designed to help lawmakers craft policies and encourage developers to safely build AI systems. However, what this means in practice is unclear given that the bill falls short of enforcing any new laws aimed at regulating AI. ®

Leave a Comment