NASA UFO report advocates for use of AI

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

NASA releases UFO report

The Story: After a yearlong plus study into UAP/UFO’s (despite the associated reputational risk), NASA just published an independent 33-page study on the issue. While they note that extraterrestrial origins of UFO’s cannot currently be confirmed (nor denied), the goal should be to shift discussion from speculation to scientific investigation through improved data collection and analysis.

More Details:

• The report found no evidence for alien UAPs (unidentified anomalous phenoma, the Pentagon’s updated terminology to encompass ‘submerged and trans-medium objects’), but cannot draw definitive conclusions given current data quality. Cutting-edge drones with erratic/experimental movement capabilities in particular represent a new problem.

• At present, analysis of UAP data is hampered by poor sensor calibration, the lack of multiple measurements, the lack of sensor metadata, and the lack of baseline data.

• Specific suggestions include using NASA satellites and radar networks to detect corroborating evidence of UAPs and rule out natural phenomena. NASA also seeks to crowdsource civilian UAP reports through smartphone apps and investigate any risks to aviation safety.

• The study aligns with a recent Pentagon report on 144 unexplained military sightings and new Congressional interest in UAPs as a national security issue.

Takeaways: Here was the independent study’s ultimate recommendation:

“NASA should contribute to a comprehensive, government-wide approach to collecting future data. The importance of detecting UAP with multiple, well-calibrated sensors is paramount, and NASA could potentially leverage its considerable expertise in this domain to utilize multispectral or hyperspectral data as part of a rigorous data acquisition campaign.

The panel finds that artificial intelligence (AI) and machine learning (ML) are essential tools for identifying rare occurrences, potentially including UAP, within vast datasets. However, these powerful techniques will only work on well-characterized data gathered with respect to strong standards.

NASA’s extensive experience in the application of state-of-the-art computational and data-analysis techniques should therefore be leveraged to provide critical assistance. Once again, appropriate data collection, curation, and distribution are paramount; NASA, with its world-leading experience in these aspects is well-positioned to play a leading role.”

Metaphysic itself built a platform on deepfakes - an avatar of Elvis Presley for a ‘live performance’ on “America’s Got Talent” in September 2022, as well as a social media parody of actor Tom Cruise called “deeptomcruise”. Their website claims they are world leaders in making faking content look real. Looks like a conflict of interest to us.

Panoptic is seeking regulatory clearance in the U.S. to market the software as a medical device. The company is also conducting clinical studies, forging partnerships with healthcare providers, and refining the technology based on user feedback. The company is a finalist in China’s 2023 Spirit of Hong Kong Awards, which recognize innovation and achievement.

• The EU is in China pitching a global AI governance body

• Meanwhile, UK Prime Minister Rishi Sunak is considering bans on China at the first major global AI safety summit amid spying concerns

• What if AI treats humans the way we treat animals?

• AI now generates music with CD-quality audio from text, and it’s only getting better

• UK Lord Justice lauds ChatGPT’s role in helping to write judgment in first known use by British judge

• Why Japan is building its own version of ChatGPT

• AI software/data analytics company Databricks raises $500M more, IPO likely not coming soon

EvoDiff's generative capabilities could help design enzymes for therapeutics, industrial uses, and more. In initial tests, it produced diverse, high-fidelity proteins. While not yet peer reviewed, EvoDiff shows the promise of diffusion models for protein engineering. Read the preprint here to find out more about the project.

• (Open source) Agents is an open-source library/framework for building autonomous language agents

• 3 ways to stop LLM hallucinations

• Replit/Stanford introduce Spellburst: an LLM-powered creative-coding environment