Former Google CEO Eric Schmidt says the U.S. public and private sectors need to develop new AI innovation and talent. During a Brookings Institution event this week, Schmidt said the U.S. risks falling behind China in innovation unless it bolsters its AI efforts surrounding national security.

More:

• AI is key to national security in the U.S., which is trailing China in areas like facial recognition, Schmidt said during the event.
• Schmidt, chairman of the National Security Commission on Artificial Intelligence, made a similar argument about federal AI efforts in a New York Times column published in February.
• Schmidt is also leading a federal initiative to launch the U.S. Digital Service Academy, a university that would train students in areas like AI and new-gen tech. The federal AI commission recently recommended the university to Congress.
• Schmidt was Google CEO from 2001 to 2011 and left as a technical adviser for the company back in February.

Nvidia's Ampere A100 GPU broke 16 world performance records in AI benchmarks. The new A100 chip has more than 54 billion transistors and is specifically designed for scientific computing, data analytics, and cloud graphics.

More:

• Nvidia dominated the third round of MLPerf training benchmark scores for AI models from eight companies: Alibaba, Dell EMC, Fujitsu, Google, Inspur, Intel, Nvidia, and the Shenzhen Institute of Advanced Technology.
• The company's A100-based supercomputer, known as Selene, led among commercially available systems and set records for single node performance and scaled out system performance. Selene, with 2,048 Ampere A100 chips, is the fastest commercial system for AI in the U.S.
• The Ampere A100 has a faster performance speed than its predecessor, the Volta V100, by a factor of up to 2.5x.
• Google’s TPU v3 system and smaller TPU v4 system also performed well. Its TPU v3 based supercomputer had the fastest training time in several benchmarks.
• NLP model BERT (bidirectional encoder representation from transformers) and DLRM (deep learning recommendation model) were new additions in this round.

Ford leased two Boston Dynamics’ robot dogs to 3D scan one of its factories in Michigan. The robots, named Fluffy and Spot, will use onboard cameras to scan the Van Dyke Transmission Plant, collecting data that will help Ford update the plant's design.

More:

• The four-legged robots will start scanning the plant from hard-to-reach areas in August. Ford’s digital engineering manager, Mark Goderis, said the scans show what the plants look like now, which will help them build a new digital engineering model that Ford can use to update the plant for new products.
• Ford says using robots instead of humans to scan is cheaper and cuts the time in half.
• Spot and Fluffy can handle stairs and uneven terrain, moving up to 3 mph. Their battery life is two hours.
• Last month, Boston Dynamics started selling Spot in the U.S. for $74,500. Uses include inspecting dangerous or remote environments and transporting payloads across difficult terrain. It launched Spot 2.0 in May.

Argo AI's valuation has reached a record $7.5b, two months after VW Group finalized its $2.6b investment in the self-driving startup. The VW Group deal builds on Ford Motor's $1B investment in Argo over five years, announced back in 2017.

More:

• On Thursday, Ford reported a $1.1b Q2 profit and ended the quarter with more than $39b in cash. This was helped by a $3.5b increase in the value of Ford’s ownership in Argo.
• Argo was founded four years ago by Peter Rander, formerly of Uber's Advanced Technology Center, and Bryan Salesky, previously of Google's self-driving car project.
• The startup has developed a 360-degree sensor and mapping technology, lidar, and cameras for autonomous driving vehicles, which it has tested in six U.S. cities.
• Ford and VW have equal ownership stakes in Argo. They will use Argo's self-driving system to independently develop  autonomous vehicles at scale.

Three AI companies have joined the business development program created by the London Stock Exchange and Global Accelerated Ventures. The companies will gain access to institutional investors and business coaching via the ELITE Group. The invite program supports members via 200 global partners such as brokers, attorneys, and marketing experts.

The new companies are:

• ModuleQ: An AI assistant to analyze data from Microsoft Teams to make business recommendations.
• Covex 2020: Analyzes data sets to support decision making and does a real-time analysis of areas impacted by pandemics.
• vElement: Specializes in AI, robotics process automation, data science, and the Blockchain.

Chinese AI startup Rokid is planning mass shipments of its next-generation smart glasses. The Rokid Glass 2 is designed to handle more applications than its body temperature-detecting predecessor, the T1 thermal glasses.

More:

• Rokid mainly develops AI and AR tech for applications from gaming to manufacturing. Company founder Zhu Mingming previously led Alibaba's AI research lab.
• The second-gen glasses, announced in January, are lighter and more futuristic, and can be controlled by head movements and speech. They support group live streaming and have an infrared sensor that can check the temperatures of up to 10 people at a time.
• Rokid’s T1 glasses have an infrared sensor that can detect the temperatures of up to 200 people in two minutes. The glasses have been widely used to combat the spread of coronavirus.
• The company, which also sells IoT and software solutions for facial recognition in the glasses, counts Credit Suisse Group and Temasek Holdings as among its investors.

QUICK HITS

• Nvidia is seeking to purchase chip-maker Arm from SoftBank for over $32b in a cash-and-stock deal.
• A large-scale machine learning project helped MIT design a novel COVID-19 vaccine that could vaccinate many more people.
• SoftBank Robotics will start selling its AI vacuum-cleaning robot Whiz in South Korea, competing against LG Electronics.
• Google added biometric support for credit card data in Chrome's autofill for Android.
• A new attack adds 3D adversarial logos to images to trick object-detecting deep neural networks.
• Carnegie Mellon University and Facebook's AI research division developed a machine learning system that helps robots differentiate among objects in the environment and navigate around them.
• Alexa and Google Assistant voice apps developers’ privacy policies are often problematic and raise privacy concerns, according to a non-peer-reviewed study coauthored by Clemson University School of Computing researchers.
• BBC heads of corporate digital Simon Pitt shared these GPT-3 conversations that he calls "witty, wise, and dangerously dark."
• CEOs and CFOs talk the winners and losers of the office of the future.*

*This is a sponsored post

Tweet of the Day: AI researcher and Google engineer François Chollet shared some wisdom about unlocking AI's potential.