Lifeboat Foundation

MIT looked at the original Roboat as “quarter-scale” option, with the Roboat II being half-scale; they’re slowly working up to the point of a full-scale option that can carry four to six passengers. That bigger version is already under construction in Amsterdam, but there’s no word on when it’ll be ready for testing. In the meantime, Roboat II seems like it can pretty effectively navigate Amsterdam — MIT says that it autonomously navigated the city’s canals for three hours collecting data and returned to where it left with an error margin of less than seven inches.

Going forward, the MIT team expects to keep improving the Roboat’s algorithms to make it better able to deal with the challenges a boat might find, like disturbances from currents and waves. They’re also working to make it more capable of identifying and “understanding” objects it comes across so it can better deal with the environment it’s in. Everything the half-scale Roboat II learns will naturally be applied to the full-scale version that’s being worked on now. There’s no word on when we might see that bigger Roboat out in the waters, though.

Elon Musk has extended his thanks to Tesla owners who received the company’s limited Full Self-Driving beta last week. The information Tesla is gathering from early access FSD beta testers will be invaluable as the company’s AI team continues to enhance and refine the EV automaker’s autonomous driving software.

The founder of Tesla Owners Club Vancouver Islands James Locke asked Elon Musk about his view on the content early access FSD testers were sharing. “Yes, very helpful,” said the Tesla CEO. “Thanks to all beta testers.”

Last week, Musk announced that Tesla plans to roll out the FSD beta to the general public later this year. Tesla will need all the information it can get to make sure that the full release of the Full Self-Driving beta goes smoothly.

AI practitioners can draw lessons from scientists working with animals to train reinforcement learning agents, a recent paper found.

Transformer model, a deep learning framework, has achieved state-of-the-art results across diverse domains, including natural language, conversation, images, and even music. The core block of any Transformer architecture is the attention module, which computes similarity scores for all pairs of positions in an input sequence. Since it requires quadratic computation time and quadratic memory size of the storing matrix, with the increase in the input sequence’s length, its efficiency decreases.

Thus, for long-range attention, one of the most common methods is sparse attention. It reduces the complexity by computing selective similarity scores from the sequence, based on various methods. There are still certain limitations like unavailability of efficient sparse-matrix multiplication operations on all accelerators, lack of theoretical guarantees, insufficiency to address the full range of problems, etc.

Microsoft, in collaboration with MITRE research organization and a dozen other organizations, including IBM, Nvidia, Airbus, and Bosch, has released the Adversarial ML Threat Matrix, a framework that aims to help cybersecurity experts prepare attacks against artificial intelligence models.

With AI models being deployed in several fields, there is a rise in critical online threats jeopardizing their safety and integrity. The Adversarial Machine Learning (ML) Threat Matrix attempts to assemble various techniques employed by malicious adversaries in destabilizing AI systems.

AI models perform several tasks, including identifying objects in images by analyzing the information they ingest for specific common patterns. The researchers have developed malicious patterns that hackers could introduce into the AI systems to trick these models into making mistakes. An Auburn University team had even managed to fool a Google LLC image recognition model into misclassifying objects in photos by slightly adjusting the objects’ position in each input image.

If humans want to travel about the solar system, they’ll need to be able to communicate. As we look forward to crewed missions to the Moon and Mars, communication technology will pose a challenge we haven’t faced since the 1970s.

We communicate with robotic missions through radio signals. It requires a network of large radio antennas to do this. Spacecraft have relatively weak receivers, so you need to beam a strong radio signal to them. They also transmit relatively weak signals back. You need a large sensitive radio dish to capture the reply. For spacecraft beyond the orbit of Earth, this is done through the Deep Space Network (DSN), which is a collection of radio telescopes custom designed for the job.

The only major crewed mission we currently have is the International Space Station (ISS). Since the ISS orbits only about 400 kilometers above the Earth, it’s relatively easy to send radio signals back and forth. But as humans travel deeper into space, we’ll require a Deep Space Network far more powerful than the current one. The DSN is already being pushed to its data limits, given the large number of active missions. Human missions would require orders of magnitude more bandwidth.

Now there’s another feat to add to GPT-3’s list: it wrote a screenplay.

It’s short, and weird, and honestly not that good. But… it’s also not all that bad, especially given that it was written by a machine.

Continue reading “OpenAI’s GPT-3 Wrote This Short Film—Even the Twist at the End” »

At the Cronin Lab at the University of Glasgow chemists developed a robotic chemist called a “chemputer” that turns words into molecules.

U.S. company can keep hosting vast coronavirus-related project but must protect French citizens’ health data from American government, court rules.

A French court has ruled that Microsoft Corp. can continue hosting a government-run project aggregating citizens’ anonymous health data to use for AI-based research, but must guarantee no data will be sent to the U.S. or be shared with American intelligence authorities.

The ruling, handed down last week, contradicts the stance of France’s data protection authority, which told the court this month that any U.S. cloud provider could be forced to comply with U.S. surveillance laws and should therefore not be allowed to host sensitive health data. The regulator’s opinion could provide clues for other companies handling such data, legal experts say.

Continue reading “French Court Asks Microsoft for Safeguards Against U.S. Surveillance of Health Data” »