China is only one to two years behind the U.S. in developing AI technologies, according to former Google chairman Eric Schmidt, who chairs the U.S. government's National Security Commission on Artificial Intelligence. Speaking before the Senate Armed Services Committee yesterday, Schmidt said the U.S. should ideally keep a five to 10-year lead over China in AI and other emerging tech.

More:

• Schmidt said China is “relentlessly focused” on becoming the world leader in so-called "high" technologies, such as quantum computing. In fact, it's already "well ahead" in areas such as facial recognition, he noted.
• Due to the diffusion of AI technologies, "anything that’s invented in the open source AI world will immediately be adopted by China,” he said, adding: “The threat is very, very real.”
• The National Security Commission on AI plans to release a report on March 1, which will call on the U.S. and its close allies to invest in 10 high-tech priorities.
• When discussing the report, Schmidt said he was worried the U.S. fails to understand the true threat from China in areas such as AI, machine-learning, hypersonics, semiconductor manufacturing, and 5G technologies. The U.S. should focus on AI and high tech as a matter of national security, not just commercially, he added.

Related:

• The U.S. maintains a "substantial lead" in artificial intelligence worldwide, though China is closing the gap in some areas, according to a January report by the Information Technology and Innovation Foundation.
• The U.S. leads in startup investment and R&D, they found. However, China had the highest number of supercomputers at 214, compared to 113 in the U.S. and 91 in the EU. China also published the most AI research papers in 2018, the last year data was available.
• The researchers predicted that China will eventually catch up to the U.S. in its development and use of AI technologies. Lead author Daniel Castro concluded that the U.S. and EU should pay attention to China, as countries that lead "will shape its future and significantly improve their economic competitiveness, while those that fall behind risk losing competitiveness in key industries."

Robotics firm Berkshire Grey plans to go public through a merger with a blank-check firm, known as a special-purpose acquisition company (SPAC). The deal, expected to provide the company with up to $413M in cash, would value the combined entity at $2.7B.

More:

• Bedford, Mass.-based Berkshire Grey develops technologies that use mobile robots, automation, and AI to process orders inside warehouses. Its system can sense, scan, and grip items during the picking process. It counts Walmart, Target Corp., and FedEx Corp. among its customers.
• TechCrunch's Brian Heater, who visited its HQ last year, said Berkshire sells a "ground-up solution for close to full automation." It doesn't offer so-called "plug and play automation solutions," which are focused on automating businesses faster and less expensively, he noted.
• SPACs are shell companies that raise money in an IPO in order to pursue an acquisition later. In this case, Berkshire would merge with SPAC Revolution Acceleration Acquisition Corp. Closing is expected in Q2 2021.
• Berkshire shareholders SoftBank Group Corp., Khosla Ventures, New Enterprise Associates, and Canaan Partners will roll their equity into the combined company, which projects revenue of $59M in 2021. It expects to become profitable by 2024.

The AI Infrastructure Alliance — a global organization of 20+ AI startups — announced its official launch, with plans to develop common standards for how businesses develop machine learning models at scale. Partnerships in the alliance will create a canonical stack for AI, which will help data scientists and engineers "move up the stack to solve more complex, higher-order problems," rather than duplicating previous methods for each project.

More:

• AI partner organizations include Algorithmia, Determined AI, WhyLabs, Pachyderm, and others. Pachyderm's chief tech evangelist Dan Jeffries will serve as the director.
• The alliance for small to medium-size AI businesses is focused on tools that data scientists and engineers "need to build robust, scalable, end-to-end enterprise AI/ML stacks on-premise or in the cloud."
• In an interview with VentureBeat, Jeffries called the initiative a “rebel alliance against the empire” that offers alternatives to tools delivered by Big Tech cloud providers.
• Because no single tool exists for teams to leverage all of AI's potential, the alliance aims to provide clarity by "building a cohesive framework and bringing together leaders and innovators to help set the standard for how data science teams build models," he noted.

Google Creative Lab launched the open-source AI experiment "Alto," or A Little Teachable Object, that adds on-device machine learning to hardware projects. The project allows users to explore machine learning basics by building a so-called "teachable object," which can recognize things in its environment.

More:

• With Alto, users can gain a "basic handle on machine learning in a way that’s adaptable to hardware projects of all sorts," according to SlashGear.
• Users teach Alto to recognize items, like an Apple, by pressing a button on the side. Once Alto "sees" an item it has been taught, it points at it with one of its arms.
• The teachable object uses a Coral USB accelerator, which provides an Edge TPU as a co-processor to computers.
• Google's Coral — a collection of hardware and software components for building devices with edge AI — graduated out of beta in October 2019. Its other components include a Coral Dev Board mini-computer with AI chip and a 5-megapixel camera accessory.
• The Alto code and template are available on GitHub.

The Echo Show 10 — Amazon's sixth smart display, and the most expensive at $250 — will start shipping tomorrow, Feb. 25. The Alexa-powered display has a rotating screen and base, powered by Amazon's own AZ1 Neural Edge processor, that automatically tracks users as they move.

More:

• Its 10.1-inch display sits atop a motorized circular base, which has a built-in camera, microphones, and tracking algorithms that rotate it to match the user's line of sight. Amazon says it's designed to (silently) make sure "video calls, recipes, and shows are always in view."
• The touchscreen display has a 1280 X 800 resolution (not 4K). The device comes with three speakers (woofer and two tweeters) and an upgraded 13-megapixel camera, which can digitally pan and zoom during calls.
• As CNN points out, the video that the device uses for motion tracking only "sees" a cubed outline of the user, which is processed locally on the device. You can also turn off the movement with voice commands, in settings, or by closing the shutter.

A version of this story first appeared in Inside Amazon. You can read the full issue here.

Thirteen members of the Forbes Technology Council shared what they consider common mistakes businesses often make when implementing their AI initiatives. Here are a few that stood out:

• Rolling out too many AI programs at once. Doing so can prevent businesses from seeing positive results. "AI is a delicate tool that can provide tremendous value to your business, but you have to be attentive and improve it." — Thomas Griffin, OptinMonster.
• Not using enough data. The outputs from AI systems don't add sufficient value if there isn't enough data or that data isn't reliable. "Another challenge is continuously training the algorithm in production and providing feedback." — Hed Kovetz, Silverfort.
• Lacking a big enough AI team. "Implementing AI requires a right-sized team to always keep your algorithms in their best shape." As a result, many businesses turn to outsourcing for their AI projects, including using on-demand services to scale existing teams. — Nacho De Marco, BairesDev.
• Believing AI can solve all their problems. Many businesses view AI as a "catch-all" solution. But they're better off taking the opposite view, using AI for a particular purpose or to address specific problems. "Then decide if you can employ an off-the-shelf solution or if a custom one is really required." — Marc Fischer, Dogtown Media LLC.
• Not identifying the right problem. Many business owners don't know whether a problem should be solved via AI, machine learning, or automation. They also run the risk of misidentifying the problem in the first place. "Identifying the right business problem, choosing the best tool or platform, inputting the required data sets, and finding the key partners to deliver are the four pillars of AI success." — Soumen Chatterjee, Wipro

• This company is reinventing home and renters insurance. Get a quote to see how much you could save.*
• FarmWise, a San Francisco-based startup developing weed-killing robots, has raised $24M in a funding round led by Playground Global.
• Arizona-based firm Gnosis IG claims to have developed AI software that can predict student performance based on datasets supplied by teachers. Phoenix's Isaac Elementary School District is now using the system, which provides predictions via a dashboard for each student.
• A new machine learning approach can simulate atomic motions in aluminum and other materials, potentially aiding in computational materials discovery.
• Zorroa Corp., a Google-backed machine-learning integration solutions provider, launched its ML SaaS platform Boon AI for media supply chain workflows.
• Researchers plan to deploy facial recognition technology to recognize and collect data on rare snub-nosed monkeys living in the Qinling mountain habitat of China.
• The New York Police Department used its four-legged robot "Digidog" — a painted version of Boston Dynamics' Spot — to investigate a home invasion and barricade situation in the Bronx.
• Virginia lawmakers unanimously passed “a de facto ban” on local police use of facial recognition; the bill now goes to Gov. Ralph Northam, who hasn't said if he will sign it.
• What are the costs of inefficiencies in your product release cycle? Read how top companies save over $1M each year.*

*This is a sponsored post.

Tweet of the Day: Researchers from Google, the University of Washington, and UC Berkeley used StyleGAN2 (Generative Adversarial Network) to "rephotograph" famous figures from history: