The Startup Is Dead, Its Data Lives On: AI Labs Are Acquiring Digital Remains of Defunct Startups

When Keelan Miskell closed down his crypto startup, he thought none of the data was valuable. Later he sold the company's digital archives and got tens of thousands of dollars in return. What the buyer wanted was the codebase and workspace data.

Matthew King from MIT Technology Review reported on this new market: institutions that help with startup winding down resell the failed firms' digital remains to AI labs. Training labs have hit the upper limit of public data and synthetic data, so they start to look for real-world records to train AI agents that can automatically execute tasks. As the business gets established, legal and ethical issues also emerge. Original link: AI labs buying data from dead startups.

Animal Language and Lunar Cities: Both Are Still Far From Reality

Cassandra Willyard of MIT Technology Review wrote about animal communication in the "Are we there yet?" series. With more devices available and advancing AI, it has become easier to study animals' calls, squeaks and clicks. But fully understanding them remains very complicated. The value of this work lies in that decoding animal communication may let humans see how other species perceive the world. The progress is not as fast as many expect.

Lunar cities are also evaluated in the same series. Both the United States and China are racing to build lunar bases, and NASA hopes to build a small lunar city by the end of this century. The lunar environment is harsh: asteroids, cosmic radiation, sharp lunar dust, and drastic temperature differences. The bases need autonomous robots, nuclear energy, and the ability to use local lunar resources to sustain operation. The report from MIT Technology Review points out that the main obstacle may be political will. Link: Imagining building cities on the moon.

A Few Side Updates

Delaware has proposed a scheme that allows AI agents to run entire enterprises. A report from Ars Technica reminds that this may let business owners evade liability for damages. Cloudflare CEO Matthew Prince told The New York Times that in the next two years, a critical, internet-wide vulnerability will emerge on average every week, since AI can find vulnerabilities faster than people can patch them. Gizmodo reports that Trump wants to ban the term "artificial intelligence" and require everyone to call it "super intelligence" instead. Put together, these few pieces of news look a bit like the tech industry is rewriting rules for content creation, liability and naming at the same time.

Apple's Carbon Neutrality Goal: Controversy Centered on Brazilian Eucalyptus Forests

In 2020, Apple set a goal to achieve net-zero emissions by the end of this century. To meet the target, the company planted millions of eucalyptus trees in Brazil to offset emissions. Gregory Barber of MIT Technology Review reported on this program. Apple bets this will bring a greener future, while critics warn that industrial tree farms do more harm than good. Original link: Inside the controversial tree farms powering Apple’s carbon-neutral goal.

When you look at all these stories together, you will find that AI's huge appetite for data has extended to the hard drives of closed companies, animal calls, and lunar dust. The next point to watch is who has the right to decide how these data are used.