The week in AI: Google DeepMind's robot assistant that learns from humans

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each Thursday, we aggregate the major developments in artificial intelligence; we pass along the news, useful resources, tools or services, and exciting projects in open source. Even if you aren’t an engineer, we’ll keep you in touch with what’s going on in AI.

By combining data from both systems, the robot became much better at doing tasks that involve moving around and using both hands. With 50 examples of each task, the robot learned to do things like cook and serve shrimp, open a big cabinet to put away heavy pots, call and get into an elevator, and wash a pan under a faucet. By learning from these examples, the robot's success rate in the given tasks went up by as much as 90%.

Check out the tutorial to build your own ALOHA for just under $32k (allen keys and hot glue gun sold separately). The hardware code, learning code and datasets are openly available on the project site.

Google’s autonomous driving technology company Waymo has published a study comparing the safety of its driverless vehicle service to human driving benchmarks. The study analyzed over 7 million miles of fully autonomous driving in Phoenix, San Francisco, and Los Angeles, comparing the crash rates from this real-world data to human crash benchmarks derived from police reports and driving studies. The results? Waymo's autonomous system had significantly lower rates of police-reported crashes (57% lower) and any injury crashes (85% lower) compared to human drivers.

Waymo’s driverless vehicle competitor Cruise (General Motors), conversely, has been trying to overcome an ‘all time low’ over a series of recent safety failures.

The maps showed that a whopping 3 out of 4 industrial fishing ships worldwide dodge public monitoring systems, with large clusters operating secretly along the coasts of Africa and Asia. The satellite imagery also spotlighted a surge in offshore wind farms and oil rigs, underscoring the need to align environmental conservation efforts, energy demands, and maritime resource management in increasingly crowded oceans. According to the researchers, this groundbreaking use of AI and satellites will open new possibilities for smarter ocean management and conservation.

Microsoft is hoping to kick off the ‘AI era’ for PC’s in 2024 with the push of a button - or at least the new Copilot key, the first major change to Windows keyboards in 30 years. A new Windows logo-styled key will activate Copilot to engage Microsoft's comprehensive AI assistant. The new key aims to simplify using AI across everyday tasks as the tech becomes more personalized, intelligent and integrated across Windows applications. The update will start appearing on new Windows 11 devices beginning in late February, including on the upcoming Microsoft Surface devices.

Nikon, Sony Group, and Canon are developing new camera technology to combat the rise of AI-generated deepfakes. Photos from these cameras will be embedded with tamper-resistant digital details in photos like date, time, location, photographer identity and other signatures that will help verify the authenticity of images. To facilitate this, an alliance of global news organizations, technology companies and camera makers has launched a free web-based tool called ‘Verify’ for checking images for these signatures, flagging AI-created or altered images.

Sony plans to introduce this technology in professional-grade cameras through a firmware update in 2024, with potential video compatibility. Canon is also developing a similar camera and a technology for adding digital signatures to videos.

More in AI this week:

• OpenAI became the nexus of the technology world in 2023

• Here are all the artists Midjourney allegedly uses to train its AI

• Japan's 'omotenashi' culture can offer an edge in the AI age

• Intel spins out AI software firm with backing from DigitalBridge

• A new kind of AI copy can fully replicate famous people - and the law is powerless.

• Can Midjourney’s CEO stop a storm of fake election images?

• Copyright law is AI's 2024 battlefield

• Chief Justice Roberts: courts will be 'significantly affected' by artificial intelligence

• Dr GPT: 84% Say ChatGPT got their diagnosis right