Welcome to Eye on AI, where AI reporter Sharon Goldman reports from traveling Jeremy Kahn. In this issue… A new open-source AI platform helps nonprofits and public agencies track a changing planet… Getty Images narrowly wins, but mostly loses, landmark UK lawsuit against Stability AI image generator… Anthropic projects $70 billion in revenue… China provides tech giants with cheap electricity to fuel the development of domestic AI chips...Amazon Employees oppose the company’s AI expansion.
I am very happy to share a story about “Artificial Intelligence Benefiting Humanity” in today’s program Pay attention to artificial intelligence: Imagine if conservation groups, scientists, and local governments could easily use AI to address challenges such as deforestation, crop failure, or wildfire risk, without requiring any AI expertise.
Until now, this has remained out of reach—requiring massive, inaccessible data sets, large budgets, and specialized AI knowledge that most nonprofits and public agencies lack. Platforms such as Google Earth AI and other proprietary systems released earlier this year have demonstrated what is possible when satellite data is combined with artificial intelligence, but these are closed systems that require access to cloud infrastructure and developer expertise.
Now that’s changing with OlmoEarth, a new open source, no-code platform that runs powerful artificial intelligence models trained on millions of Earth observation data (from satellites, radars and environmental sensors, including open data from NASA, NOAA and the European Space Agency) to analyze and predict planetary change in real time. It was developed by Ai2 (The Allen Institute for Artificial Intelligence), a Seattle-based nonprofit research lab founded by the late Allen in 2014. Microsoft Co-founder Paul Allen.
Early partners are already putting OlmoEarth to use: In Kenya, researchers are mapping crops to help farmers and officials increase food security. In the Amazon, conservationists are seeing deforestation happening almost in real time. In mangrove areas, early tests show accuracy as high as 97%, cutting processing time in half, helping governments take faster action to protect fragile coastlines.
I spoke with Patrick Beukema, who leads the Ai2 team that builds OlmoEarth, a project that launched earlier this year. The goal is not just to release a powerful model, Beukema said. Many organizations are working to connect raw satellite and sensor data into usable artificial intelligence systems, so Ai2 built OlmoEarth as a complete end-to-end platform.
“Organizations find it extremely challenging to build pipelines from all these satellites and sensors, and even basic things are difficult to do—one model may need to connect to 40 different channels from three different satellites,” he explains. “We’re just trying to democratize these organizations that are working on these really important problems and super important missions — we think technology should basically be open and easy to use.”
One specific example Beukema gave me was assessing wildfire risk. A key variable in wildfire risk assessment is how wet the forest is, as this determines how flammable the forest is. “Currently, what people do is go into the forest, collect branches or logs, and weigh them before and after dehydration to measure the humidity of the site in one go,” he said. “Park rangers do the work, but it’s extremely expensive and difficult.”
With OlmoEarth, artificial intelligence can now estimate forest moisture from space: The team used years of expert field data from forest and wildfire managers to train the model, pairing these ground measurements with satellite observations from dozens of channels, including radar, infrared and optical imagery. Over time, the model learned to predict how wet or dry an area would be by analyzing combinations of signals.
Once trained, it can continuously map humidity levels across an entire region, updating as new satellite data arrives, and is millions of times cheaper than traditional methods. The result: near-real-time wildfire risk maps that help planners and rangers take action faster.
“Hopefully this will be helpful to the people on the front lines doing this important work,” Beukema said. “That’s our goal.”
More AI news follows.
Sharon Goldman
sharon.goldman@fortune.com
@sharongoldman
If you want to learn more about how AI can help your company succeed and hear from industry leaders about where this technology is heading, I hope you’ll consider joining Jeremy and me at the Fortune Brainstorm AI Conference in San Francisco on December 8-9. Speakers confirmed so far include Google Cloud CEO Thomas Kurian; Intuit CEO Sasan Goodzi, Databricks CEO ALI GHODSI, GLEAN CEO Arvind Jain, Amazon’s Panos Panos Panay and many more. Register now.
The wealth of artificial intelligence
Palantir hits $1.2B in quarterly revenue, but stock slides after big jump——Jessica Matthews
Sam Altman sometimes wants OpenAI to go public so haters can short the stock – ‘I’d like to see them get burned for it’ ——Marco Queiroz-Guitierrez
Artificial intelligence enables criminals to launch “massively customized attacks” but can also help businesses strengthen their defenses, tech industry leaders say ——Author: Angelica Ang
Artificial Intelligence News
Getty Images has mostly lost a landmark UK lawsuit against Stability AI image generator. Reuters reported today that a London court ruled that Getty only narrowly won the case against Stability AI, but mostly lost, finding that Stable Diffusion infringed Getty’s trademark rights by copying Getty’s watermark in images generated by artificial intelligence. But the judge rejected Getty’s broader copyright claim, saying Stable Diffusion “does not store or copy any copyright work” – a technical distinction that lawyers said exposed a loophole in UK copyright protection. The mixed verdict did not resolve a core issue: whether training an AI model on copyrighted data constituted infringement, and both companies claimed the outcome was a partial victory. Getty said it plans to use the ruling to bolster its parallel lawsuits in the United States, while calling on governments to strengthen transparency and intellectual property rules for artificial intelligence.
Anthropic expects revenue to reach $70 billion by 2028 and cash flow to reach $17 billion. Anthropic, the maker of the Claude chatbot, is expecting explosive growth, with revenue forecast to reach $70 billion by 2028, up from about $5 billion this year. information. The company expects most of its growth to come from businesses using its artificial intelligence models through APIs, and expects OpenAI’s comparable sales to roughly double next year. Unlike ChatGPT maker OpenAI, which consumes billions of dollars in computing costs, Anthropic expects to be cash flow positive by 2027 and generate up to $17 billion in cash the following year. Those numbers could help it target a valuation of between $300 billion and $400 billion in its next funding round, positioning the four-year-old startup as a financially efficient challenger to OpenAI’s dominance.
China provides tech giants with cheap electricity to fuel the development of domestic artificial intelligence chips. according to financial times, China is ramping up subsidies for its largest data centers, slashing electricity bills by as much as 50% for facilities using domestic artificial intelligence chips to reduce reliance on Nvidia and strengthen its local semiconductor industry, according to the company. financial times. Local governments in provinces including Gansu, Guizhou and Inner Mongolia are rolling out new incentives after tech giants including ByteDance, Alibaba and Tencent complained that Chinese chips from Huawei and Cambrian were less energy efficient and more expensive to run. The move underscores Beijing’s commitment to making its artificial intelligence infrastructure self-sufficient, as domestic chips still require 30% to 50% more power than Nvidia despite a surge in power demand for the country’s data centers.
Amazon employees oppose the company’s artificial intelligence expansion. Last week, a group of Amazon employees Publish an open letter The company’s “overdrive” push for artificial intelligence comes at the expense of climate goals, worker protections and democratic accountability, it warned. The signatories, who claim they helped build and deploy Amazon’s artificial intelligence systems, argue the company’s planned $150 billion data center expansion will increase carbon emissions and water use, especially in arid regions, even as the company continues to provide cloud tools to oil and gas companies. They also criticize Amazon’s growing ties to government surveillance and military contracts, and claim internal artificial intelligence programs are accelerating automation without supporting worker advancement. The group calls for three commitments: no AI powered by dirty energy, no AI without employee input, and no AI used for violence or mass surveillance.
Follow artificial intelligence research
What if large AI models could read each other’s thoughts instead of chatting over text? That’s the idea behind a new paper from researchers at CMU, Meta AI and MBZUAI, titled Exchange of ideas in multi-agent collaboration. The team proposed a system called ThoughtComm that lets AI agents share their underlying “thoughts” — the representations underlying their reasoning — rather than just exchanging words or tokens. To do this, they used sparse regularized autoencoders, a type of neural network that compresses complex information into a smaller set of the most important features, helping to reveal which “ideas” really matter. By understanding which ideas agents share and which ones they retain, the framework enables them to coordinate and reason more efficiently – hinting at a future where AI collaborates not by talking but by “thinking” simultaneously.
you have a calendar
November 10-13: Web Summit, Lisbon.
November 19: Nvidia third quarter earnings report
November 26-27: World Artificial Intelligence Conference, London.
December 2-7: NeurIPS, San Diego
December 8-9: wealth brainstorming AI San Francisco. Apply to participate here.
brain-boosting foods
How AI companies quietly train for paid journalism
I want to emphasize a new one Atlantic investigation Written by staff writer Alex Reisner, it reveals how Common Crawl, a nonprofit that crawls billions of web pages to build a free archive of the internet, became a backdoor for paid-content AI training. Reisner reports that despite Common Crawl’s public claims that it avoids content behind a paywall, its dataset includes full articles from major news outlets, and these articles end up in the training data for thousands of AI models.
Common Crawl insists it did nothing wrong. When asked about publishers’ requests to remove their content, Common Crawl director Rich Skrenta dismissed the complaints, saying, “If you don’t want your content to be on the Internet, you shouldn’t be putting it on the Internet.” Skrenta told Reisner that he sees the archive as a kind of digital time capsule — a “crystal cube on the moon” — and as a record of civilization’s knowledge. Regardless, it certainly highlights the growing tension between artificial intelligence’s thirst for data and journalism’s copyright battles.