- Sabine's Newsletter
- Posts
- Telescope Troubles, Robots’ Data Gap, 1000+ Scam Journals, and Vibe Physics
Telescope Troubles, Robots’ Data Gap, 1000+ Scam Journals, and Vibe Physics
This week’s science bits from SWTG
Marcus Gee-Lim
September 03, 2025
|
Misconduct Claims At 2 Billion Dollar Telescope
Image: Artist’s impression of SKAO antennas in South Africa and Australia.Credit: SKAO, ICRAR, SARAO
The organization of the Square Kilometer Array Observatory (SKAO) has been accused of improper handling of governmental funding, according to a whistleblower report seen by the Guardian. The SKAO is a double array of radio telescopes, one in South Africa and one in Australia, and is still under construction. The allegations, which have been in the rumour mill for about a year, include a lack of information about what the organization spends money on, unexplained overspending, and procurement manipulations. A spokesperson of the SKAO denied any wrongdoing.
This week’s episode of Science News is about “vibe physics” — basically using an LLM to come up with new physics theories. I've tried GPT 5, Gemini, Deepthink, Grok 4, and Claude Opus 4.1 to get some feedback on a physics idea.
Also, you can now create and share your own quizzes on QuizWithIt for free! Each quiz has a unique URL, can be embedded into websites or newsletter, and be shared on social media. Happy quizzing!
Roboticist Warns that Data Gap will Impede Robot Revolution
Roboticist Ken Goldberg from UC Berkeley has published a new paper detailing the problem that we have little data to train AI operated robots on. He calls it the “100,000 year gap,” based on an estimate for the amount of time a single human would have had to spend writing all the data that Large Language Models have been trained on. Training robots will likely require even more data than they currently have, and gathering this data is a slow process. Video is of limited use because it doesn’t detail muscle operations and converting it to 3D is fraught with problems. Training robots in simulations seems (at least so far) to not work very well. Goldberg concludes that the robot revolution is further off than many companies want us to believe: “It’s not going to happen in the next two years, or five years, or even 10 years.” Paper here, press release here.
Fact-based news without bias awaits. Make 1440 your choice today.
Overwhelmed by biased news? Cut through the clutter and get straight facts with your daily 1440 digest. From politics to sports, join millions who start their day informed.
New AI tool Identifies 1000+ Likely Scientific Scam Journals
Computer scientists from the University of Colorado in Boulder have developed an AI-based tool that automatically scans and flags journals with questionable publication practices. Scam journals are part of the pseudoscience market – they offer fake peer reviews for a fee, typically in the range of $1,000. The researchers trained their model on websites of journals that have been identified as scam journals (sometimes called “predatory journals”) in the past. While they do not know exactly what the system picked up on, some of the easier to interpret markers are a lack of reputable scientists in the editorial board, typos on the websites, as well as atypical patterns in citations. It is a much needed help in an arms race against the rise of pseudoscientific practices. Press release here. Paper here.
