Google Researcher Says AI Agents Rediscovered a Maya Site

Research on AI in health care spans pregnancy, childhood, skin conditions, and cancer treatment. A review of six studies examined tools including a smartwatch that checks for low blood pressure during cesarean delivery: at a key threshold, it achieved an area under the curve of 0.87, sensitivity of 79.2% and specificity of 82.0%, but was less sensitive to smaller drops. Other studies explored drug dosing for critically ill children, predicting low phosphate in very preterm infants, and forecasting whether labor pain relief will work. Researchers say these tools need population-specific evidence and careful validation before they inform clinical decisions.

A review of allergic contact dermatitis care notes that at least 20% of people are allergic to common environmental substances, while 5% to 10% experience active symptoms each year. About 2,800 contact allergens have been identified. Alongside patch testing and treatments such as emollients and corticosteroids, future approaches could draw on genetics, skin-barrier biology and digital tools to predict reactions. But product information can be incomplete: one study found 31 allergen omissions across 93 labels. In cancer care, Massive Bio uses AI to compare patient records with clinical trials, sometimes in 17 seconds. The company says it has helped more than 200,000 patients in 17 countries and published a study involving 3,800 patients; its co-founder says patient engagement and education still limit enrollment.

At MIT Sloan, researchers organized more than 2,200 proposed AI safeguards into a framework covering the risks addressed, responsible parties, timing in the AI lifecycle and how each measure works. Fifty-five percent of proposals focus directly on AI systems; developers account for about 41% and organizations using AI for 33%. Fewer than 2% address human oversight of autonomous agents, recovery from AI disruptions or care of computing resources. The team also created PRIMER to help organizations report and compare safeguards. Separately, Cedar Rapids residents joined an AI Reality Checks forum, part of a series across 30 states urging rules for powerful AI systems and attention to AI in the 2028 presidential campaign. Humans in Control has launched a tracker of candidates’ positions.

AI is also reshaping legal work, where tools can search case law, flag contract risks and review large collections of records. The article describes potential time and cost savings but stresses that lawyers must protect privacy and attorney-client privilege and verify outputs; AI remains an aid, not a substitute for legal judgment. Kumar Srivastava makes a related point about education and hiring: because AI can generate polished essays, projects and code, finished work alone no longer proves understanding. He argues for assessments that let people demonstrate, explain and revise their thinking, rather than relying on AI-detection tools.

Other coverage highlights the risks of overstating what AI has achieved and who it serves. Google researcher Pietro Miotti said AI agents found a Maya site in Mexico in NASA lidar data, but archaeologists noted it had already been mapped, although it was absent from the Mexican government’s database. Miotti acknowledged the agents had rediscovered it after missing earlier research. The episode renewed concerns about credit and claims that overlook prior scholarship; remote-sensing experts also note that machine learning and computer vision have long been used in the field. An article on LGBTQIA+ rights warns that AI systems built around past patterns can miss changing language, expression and identities, while market priorities may leave marginalized communities underserved. In San Jose, meanwhile, humanoid robot Sir Kit promotes the city at trade shows for Team San Jose; a “brother” works at San Jose Mineta airport.

Key Takeaways

  • A smartwatch study of low blood pressure during cesarean delivery reported 79.2% sensitivity and 82.0% specificity at a key threshold, with researchers calling for further validation.
  • Contact dermatitis affects a substantial share of people: at least 20% have allergies to common environmental substances, and 5% to 10% have active symptoms annually; one label study found 31 allergen omissions across 93 products.
  • Massive Bio says its AI matches cancer patients to clinical trials in as little as 17 seconds and has helped more than 200,000 patients in 17 countries, but enrollment still depends on patient engagement and education.
  • MIT researchers mapped more than 2,200 proposed AI safeguards; 55% focus on AI systems, while fewer than 2% address several areas including oversight of autonomous agents.
  • PRIMER, developed by the MIT team, is intended to help organizations report safeguards consistently and compare approaches.
  • Legal AI tools can assist with research, contract review and discovery, but lawyers must check outputs and protect privacy and attorney-client privilege.
  • Educators and employers face pressure to assess how people think and work, not just polished outputs; Kumar Srivastava warns that AI-detection efforts are an unwinnable arms race.
  • Google researcher Pietro Miotti acknowledged that AI agents rediscovered a Maya site already mapped by researchers, prompting concerns about credit and overlooked prior work.
  • AI Reality Checks forums are being held across 30 states; a Cedar Rapids event called for AI policy attention ahead of the 2028 election, and Humans in Control launched a candidate-position tracker.
  • Articles also flag risks that AI may underserve LGBTQIA+ people and note San Jose’s use of humanoid robot Sir Kit to promote the city at trade shows.

Six studies explore AI support for care during pregnancy and childhood

Researchers reviewed six studies on how artificial intelligence, drug models, and wearable devices could support care for pregnant people and children. The studies covered linezolid dosing for critically ill children, predicting low phosphate in very preterm infants, smartwatch checks for low blood pressure during cesarean delivery, and models that predict whether labor pain relief will work. The smartwatch showed an area under the curve of 0.87, sensitivity of 79.2 percent, and specificity of 82.0 percent at a key blood pressure threshold, but it was less sensitive to smaller drops. The researchers stressed that promising results need population-specific evidence and careful validation before these tools guide clinical decisions.

Review explores personalized care and AI for contact dermatitis

A review by Sean E. Mangion, Samantha A. Sandelowsky, Evan D. Hicks, and Ross L. Pearlman examines how precision medicine and digital tools could improve care for allergic contact dermatitis. At least 20 percent of people are allergic to common environmental substances, and 5 to 10 percent have active symptoms each year. About 2,800 contact allergens have been identified, including fragrances, preservatives, dyes, natural rubbers, and industrial metals. Current care includes patch testing, avoiding allergens, emollients, and corticosteroids, while future approaches may use genetics, skin-barrier biology, and digital technologies to help predict and prevent reactions. The review also notes that ingredient information can be incomplete, citing a study that found 31 allergen omissions across 93 product labels.

MIT researchers map more than 2,200 proposed AI safeguards

Researchers at MIT Sloan School of Management created a framework to organize more than 2,200 proposed ways to reduce AI risks. It sorts safeguards by the risks they address, who should act, when action should happen in the AI lifecycle, and how the safeguard works. The study found that 55 percent of proposals focus directly on AI systems, while developers account for about 41 percent of safeguards and organizations using AI account for 33 percent. Less than 2 percent of the proposals addressed human oversight of autonomous AI agents, society’s ability to recover from AI disruptions, or care of computing resources. The team also developed PRIMER to help organizations report safeguards consistently and compare approaches.

AI is changing legal research, contracts, and client services

Artificial intelligence tools are changing how lawyers research cases, review contracts, and support litigation. Research platforms can scan case files, laws, and legal precedents quickly, while contract tools can flag risky clauses, inconsistencies, and missing terms. AI can also help review large groups of emails and records during discovery and analyze past court decisions for patterns. The article says these tools may reduce costs and speed up work for clients, but it warns that lawyers must protect privacy and attorney-client privilege and check AI outputs for errors. It presents AI as an assistant, not a replacement for professional legal judgment.

AI agents rediscovered a Maya site already mapped by researchers

Google AI researcher Pietro Miotti said his AI agents found an ancient Maya site in Mexico by analyzing NASA lidar data. Archaeologist Sarah Parcak and others pointed out that the site had already been mapped, even though it was missing from the Mexican government’s database. Miotti later acknowledged that the agents had rediscovered the site because they missed earlier research. The incident raised concerns about AI researchers receiving credit for existing discoveries and about exaggerated claims that overlook experts’ work. Geoscience consultant Alex Janevski noted that machine learning and computer vision have long been used in remote sensing.

Massive Bio uses AI to match cancer patients with clinical trials

Dr. Arturo Loaiza-Bonilla co-founded Massive Bio after seeing eligible cancer patients go unnoticed by clinical trial systems. Its AI platform reviews patient records and matches them with trials in as little as 17 seconds. The company says it has helped more than 200,000 patients in 17 countries, received an NCI SBIR grant, and published a study involving 3,800 patients. Loaiza-Bonilla says better matching alone has not increased enrollment, and that patient engagement and education remain major challenges.

AI is changing how schools and employers should assess knowledge

Kumar Srivastava argues that AI can now produce essays, projects, and software that once showed a person’s skills and learning. He says polished work is no longer proof that someone understands how it was made, citing job candidates who could not explain their code. Schools and employers have responded in different ways, from banning AI to allowing it, while some use monitoring tools to detect its use. Srivastava warns that detection is an unwinnable arms race and calls for assessments that let people show, explain, and revise their thinking.

Cedar Rapids forum calls for AI guardrails before the 2028 election

Cedar Rapids residents met on Saturday to discuss artificial intelligence and ways to urge 2028 presidential candidates to address it. The event was part of AI Reality Checks, a series held across 30 states to build support for rules on the most powerful AI systems. Gavin Nelson, Iowa state director of Humans in Control, said local involvement is key to finding broad solutions. The organization also launched a tracker to record 2028 presidential candidates’ views on AI.

AI robot Sir Kit promotes San Jose at trade shows

Sir Kit, an AI-powered humanoid robot, is San Jose’s newest cheerleader. The robot has accompanied Team San Jose to trade shows. Sir Kit also has a “brother” working at San Jose Mineta airport.

Experts warn AI systems can harm LGBTQIA+ people

An article argues that AI poses risks to democracy and LGBTQIA+ rights, and says concerns raised in 2021 are already affecting people. Anthropologist and media scholar Mary L. Grey, a 2020 MacArthur Fellow, said then that AI would “always fail” the LGBTQIA+ community and now stands by that view even more strongly. Many AI tools rely on patterns from past examples, which can miss the changing language, expression, and identities within the community. The article also says the AI market tends to prioritize what the largest number of people want, leaving marginalized groups at risk.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI in healthcare clinical decision support maternal health pediatric medicine allergic contact dermatitis cancer clinical trials AI safeguards AI governance PRIMER framework AI policy legal AI education and assessment AI detection archaeology remote sensing LGBTQIA+ inclusion humanoid robots AI validation privacy human oversight

Comments

Loading...