Training data sourcing and licensing, data quality, and synthetic data.

Current topic Data & Training · 33 Switch topic

Latest stories

  1. Australia's central bank finds AI hitting young workers first, especially in software and customer service

    Australia's central bank, the Reserve Bank of Australia (RBA), finds that AI has barely displaced jobs in Australia so far, but that young workers in the most AI-exposed occupations are doing worse. In programming, clerical work and customer service, workers aged 15 to 24 and young graduates show weaker employment and higher unemployment. Since late 2022, employment for young people in these jobs has fallen while it has risen for older workers. The RBA says it has not established causation and that it measures theoretical exposure rather than actual AI use. A Stanford study covered in the same article found employment for 22- to 25-year-olds in exposed occupations fell 13% in relative terms since late 2022.

    View in daily briefing Read original

  2. Google’s AlphaGenome 2 zooms in on the cellular roots of disease

    Google DeepMind unveiled AlphaGenome 2, an AI model that predicts how DNA changes affect different kinds of cells. Almost every cell in the body carries the same DNA, but heart muscle cells, skin cells and infection-fighting immune cells switch different genes on, so the same genetic variants can act differently in each. DeepMind says the model, trained on more than 100 million individual cell profiles across 68 tissues, finds nearly twice as many variants that change gene expression as its previous best model, and it can handle new cell types and states without retraining. It is available for non-commercial research through the AlphaGenome API, with a limited private preview for commercial research.

    View in daily briefing Read original

  3. Google, Anthropic and others to support new AI crawling opt-out standard

    Google, Microsoft, Anthropic and OpenAI said they will support a new standard that lets a website block every crawler collecting training data for generative AI with a single robots.txt entry. Today a site must add a separate entry for each AI provider, and for each crawler a provider runs. The four companies have already submitted a draft to the IETF, the nonprofit that maintains the Robots Exclusion Protocol, and models that perform exclusively non-generative tasks would not be covered.

    View in daily briefing Read original

  4. OpenAI plans to train models on Daybreak Blue cybersecurity session data that was previously off-limits

    Starting November 29, OpenAI will use Daybreak Blue cybersecurity session data from members of its Trusted Access program to train its models. That data had previously been excluded from training. After backlash to the first email, OpenAI sent a second one promising a separate opt-out. Interactions flagged in safety reviews can still be used for safety purposes.

    View in daily briefing Read original

  5. Google Deepmind creates a global map of forests with satellite data and AI

    Google DeepMind and the World Resources Institute, a nonprofit research organization, built ForTy, a map that assigns the world's 2020 land to six classes: primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops such as oil palm, rubber or cocoa, and other land. Experts checking 8,190 sites found it correct 90.2 percent of the time, and its forest-type score of 90.9 percent beat the European Commission's 87.8 percent. But it found only about half of tree crops and planted forests and 60 percent of plantations.

    View in daily briefing Read original

  6. OpenAI agents analyze mouse brain data in 32 minutes, a job that would take humans months

    OpenAI, working with PerturbAI and Scripps Research, used thousands of AI agents to analyze biological data from millions of mouse cells in 32 minutes. The agents ran on GPT-Rosalind and worked through a whole-brain CRISPR atlas, a map of how changing individual genes affects different brain cell types. Each agent took one gene perturbation, analyzed how cells responded and searched the scientific literature for explanations. The team estimates the same work done by hand would take several months to a year.

    View in daily briefing Read original

  7. Beyond tree cover: mapping the world’s forest types

    Google DeepMind and Google Research, with the World Resources Institute (WRI) and the International Institute for Applied Systems Analysis (IIASA), released Forest Typology (ForTy) 2020, a global map that classifies every 10-by-10-meter patch of land into one of six types. Unlike maps that only show whether trees are present, it separates primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops such as oil palm, rubber and cocoa, and other land. An early yearly time series covering 2017 to 2025 is included. Validated against more than 9,000 randomly chosen sites, it identified the correct forest type 90% of the time and forest versus non-forest 95% of the time.

    View in daily briefing Read original

  8. Meta drops the word "train" from its commercial TOS after AI backlash

    Meta has dropped the word “train” from its commercial terms of service, which take effect October 30, but legal experts, including Cornell’s James Grimmelmann, say the wording is still broad enough to allow scraping websites. Last week’s terms drew backlash by letting Meta use “any content” from sites linked on business pages to provide, develop, improve, research, train, and maintain its products. The revised terms delete “any content” and say Meta may “periodically pull updates” from sites to develop advertising and business products. Meta did not add language promising to respect robots.txt.

    View in daily briefing Read original

  9. Meta tries to scapegoat staff who downloaded torrents to spare execs

    Meta has argued that nobody “decided” to pirate books for AI training, and Magistrate Judge Thomas Hixon has now ordered it to name the executives who authorized the downloads, Ars Technica reported, citing TorrentFreak. The order applies to a class action by authors including Richard Kadrey and Sarah Silverman and to a suit by publisher Entrepreneur Media. In a filing, Meta described its early data gathering as an “informal approach” in which a decision was made only in the sense that staff “technically” downloaded torrents. The plaintiffs argue that downloading 80.6 terabytes from LibGen and 35.7 terabytes from Anna’s Archive means Meta must also have been seeding, that is, uploading the files to others.

    View in daily briefing Read original

  10. Meta removes language that raised scraping and AI training concerns

    Meta removed a disputed passage from its commercial terms after saying nothing had changed. The passage said Meta could use any content from a business account holder's domains, web pages or URLs to develop and train its products. Legal experts told Nieman Lab it could be read as permission to scrape publishers' sites to train AI models. Meta said the crawling terms apply only to its ads and business products, that it has made this explicit in the terms, and that how it uses web crawlers has not changed.

    View in daily briefing Read original

October 20264

September 202610

August 20269