Lifeboat Foundation

Tesla has finally decided to release its Autopilot safety data report after taking a break of more than a year.

For years, Tesla used to release a “Vehicle safety report” that tracked miles between accidents in its vehicles based on the level of Autopilot used or not used and compared it to the industry average.

The automaker used the report to claim that its Autopilot technology resulted in a much safer driving experience and that its vehicles would crash much less often than the average car in the US even without Autopilot.

Nuclear physicists have long been working to reveal how the proton gets its spin. Now, a new method that combines experimental data with state-of-the-art calculations has revealed a more detailed picture of spin contributions from the very glue that holds protons together. It also paves the way toward imaging the proton’s 3D structure.

The rhetoric over “superhuman” AI implicitly erases what’s most important about being human.

Still, ChatGPT operates in a mostly siloed fashion. It can’t yet venture out “into the wild” to execute online tasks. For example, if you wanted to buy a milk frother on Amazon for under $100, ChatGPT might be able to recommend a product or two, and even provide links, but it can’t actually navigate Amazon and make the purchase.

Why? Besides obvious concerns, like letting a flawed AI model go on a shopping spree with your credit card, one challenge lies in training AI to successfully navigate graphical user interfaces (GUIs), like your laptop or smartphone screen.

But even the current version of GPT-4 seems to grasp the basic steps of online shopping. That’s the takeaway of a recent preprint paper in which AI researchers described how they successfully trained a GPT-4-based agent to “buy” products on Amazon. The agent, dubbed the MM-Navigator, did not actually purchase products, but it was able to analyze screenshots of an iOS smartphone screen and specify the appropriate action and where it should click, with impressive accuracy.

“The fear is that they can’t afford to let someone else get there first,” said Scott Jenson, a UX designer who left Google last month.

From MIT

Not all language model features are linear.

Recent work has proposed the linear representation hypothesis: that language models perform computation by manipulating one-dimensional representations of concepts (“features”) in activation space.

Continue reading “Not All Language Model Features Are Linear” »

⚔️ Closed-source vs. Open-weight LLMs The gap between closed-source and open-weight models is closing in terms of MMLU.

Post-training, model editing, quantization.

Researchers have developed a new technique to view living mammalian cells. The team used a powerful laser, called a soft X-ray free electron laser, to emit ultrafast pulses of illumination at the speed of femtoseconds, or quadrillionths of a second.

Apptronik, a NASA-backed robotics company, has unveiled Apollo, a humanoid robot that could revolutionize the workforce — because there’s virtually no limit to the number of jobs it can do.

“The focus for Apptronik is to build one robot that can do thousands of different things,” Jeff Cardenas, the company’s co-founder and CEO, told Freethink. “The best way to think of it is kind of like the iPhone of robots.”

Continue reading “NASA partner unveils the ‘iPhone’ of robots” »