USF-led study: AI helps reveal global surge in floating algae

Machine learning shows ocean conditions increasingly favor macroalgae growth

University of South Florida

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

These figures show the change in density of global floating algae in the 20 years between 2003 and 2022.

view more 

Credit: Qi et al

Key takeaways:

• AI analysis of 20 years of satellite data shows floating macroalgae blooms expanding worldwide, with rapid growth beginning around 2008–2010.

• Researchers used deep learning and high-performance computing to detect algae that often make up less than 1% of a satellite pixel — a task not possible without artificial intelligence.

• While floating algae can support marine life offshore, large blooms threaten coastal ecosystems, tourism and local economies when they reach shore.

TAMPA, Fla. (Jan. 19, 2026) – For the first time and with help from artificial intelligence, researchers have conducted a comprehensive study of global floating algae and found that blooms are expanding across the ocean. These trends are likely the result of changes to ocean temperature, currents and nutrients, according to the authors, and could have a significant impact on marine life, tourism, and coastal economies.

Led by researchers at the University of South Florida and the National Oceanic and Atmospheric Administration, the study demonstrates the power of artificial intelligence as a tool for processing large amounts of ocean data. The findings are embargoed for public release until Monday, Jan. 19, at 5 a.m. ET.

“While regional studies have been published, our paper gives the first global picture of floating algae, including macroalgal mats and microalgal scum,” said Chuanmin Hu, professor of oceanography at the USF College of Marine Science and senior author of the paper set to publish Monday in Nature Communications. “Our results show that the global ocean now favors the growth of floating macroalgae.”

Hu refers to macroalgae such as seaweed as a double-edged sword. In open water, they can provide critical habitat for marine life and have a positive impact on fisheries – serving as nurseries for many species. But once the algae reach coastal waters, the decaying biomass can cause considerable harm to tourism, economies and the health of people and marine life.

Between 2003 and 2022, both microalgal scum and macroalgal mats expanded around the globe. Microalgae on the ocean surface saw a modest but significant increase of one percent per year. However, blooms of macroalgae increased by 13.4 percent per year in the tropical Atlantic and western Pacific, the authors found, with the most dramatic increase in biomass occurring after 2008. The cumulative size of these microalgal blooms reached 43.8 million square kilometers (16.9 million square miles), breaking with historic trends.

The tipping points for macroalgae blooms occurred around 2010. The first major bloom of the green seaweed known as Ulva happened in the Yellow Sea in 2008. A significant bloom of the brown seaweed sargassum took place in the tropical Atlantic in 2011. Another sargassum bloom occurred in the East China Sea in 2012.

“Before 2008, there were no major blooms of macroalgae reported except for sargassum in the Sargasso Sea,” Hu said. “On a global scale, we appear to be witnessing a regime shift from a macroalgae-poor ocean to an macroalgae-rich ocean.”

To conduct the study, Hu and his colleagues used artificial intelligence to scan 1.2 million satellite images of the ocean, focusing on 13 zones and five types of algae. They trained a deep-learning model to spot features that signal the presence of algae floating on the ocean surface. In most cases, these features appear across many image pixels, but they typically comprise less than one percent of each pixel.

Lin Qi, an oceanographer at the NOAA Center for Satellite Applications and Research and first author of the study, updated a computer model previously developed by the same research team, allowing them to analyze images from the global ocean for 20 years. It took several months and millions of image features to train Qi’s model.

The authors credit USF’s Research Computing for its critical role in the study. The facility provided access to high-performance infrastructure that processed multiple groups of images simultaneously. Even still, it still took several months to process and analyze the 1.2 million satellite images.

“This work is impossible without the high-performance computing facility or the long-term collaborations between NOAA and USF,” Qi said.

The study attributed the bloom expansions to both human activities, such as nutrient runoff into the ocean, and climate variability, such as ocean warming, while acknowledging that the reasons may differ among regions. Looking forward, Qi said, “we are going to explore more satellite data and look for better understanding of the expansions.”

Read more: Study reveals dramatic decline in some historic sargassum populations

Read more: USF experts lead on sargassum research, monitoring, and prediction

###

About the University of South Florida

The University of South Florida is a top-ranked research university serving approximately 50,000 students from across the globe at campuses in Tampa, St. Petersburg, Sarasota-Manatee and USF Health. In 2025, U.S. News & World Report recognized USF with its highest overall ranking in university history, as a top 50 public university for the seventh consecutive year and as one of the top 15 best values among all public universities in the nation.  U.S. News also ranks the USF Health Morsani College of Medicine as the No. 1 medical school in Florida and in the highest tier nationwide. USF is a member of the Association of American Universities (AAU), a group that includes only the top 3% of universities in the U.S. With an all-time high of $750 million in research funding in 2025 and as a top 20 public university for producing U.S. patents, USF uses innovation to transform lives and shape a better future. The university generates an annual economic impact of more than $6 billion.  USF’s Division I athletics teams compete in the American Conference. Learn more at www.usf.edu.

---

Journal

Nature Communications

DOI

10.1038/s41467-025-66822-5 

Method of Research

Imaging analysis

Subject of Research

Not applicable

Article Title

Global floating algae blooms are expanding

Article Publication Date

19-Jan-2026

  

Macroalgal mats can serve as valuable habitat for many marine species, such as this sargassum fish. 

Credit

Ocean Image Bank -- Lorenzo Mittiga

Sargassum observed during a sampling cruise.

Credit

Sully Sullivan

Floating algae types

Credit

Qi et al

Satellite image of Ulva macroalgae in the Yellow Sea.

Credit

NOAA Star

Large Sargassum field.

Credit

Ellen Park -- Woods Hole Oceanographic Institution

 

From chatbots to dice rolls: Researchers use D&D to test AI’s long-term decision-making abilities

WHICH CHARACTER DOES IT FAVOR; THIEF

University of California - San Diego

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

Large Language Models, like ChatGPT, are learning to play Dungeons & Dragons. The reason? Simulating and playing the popular tabletop role-playing game provides a good testing ground for AI agents that need to function independently for long stretches of time. 

Indeed D&D’s complex rules, extended campaigns and need for teamwork are an ideal environment to evaluate the long-term performance of AI agents powered by Large Language Models, according to a team of computer scientists led by researchers at the University of California San Diego. For example, while playing D&D as AI agents, the models need to follow specific game rules and coordinate teams of players, comprising both AI agents and humans. 

The work aims to solve one of the main challenges that arise when trying to evaluate LLM performance: the lack of benchmarks for long-term tasks. Most benchmarks for these models still target short term operation, while LLMs are increasingly deployed as autonomous or semi-autonomous agents that have to function more or less independently over long periods of time. 

“Dungeons & Dragons is a natural testing ground to evaluate multistep planning, adhering to rules and team strategy,” said Raj Ammanabrolu, the study’s senior author and a faculty member in the Department of Computer Science and Engineering at UC San Diego. “Because play unfolds through dialog, D&D also opens a direct avenue for human-AI interaction: agents can assist or coplay with other people.”

The team presented their work at the NeurIPS 2025 conference from Dec. 2 to 7 in San Diego,. The researchers took the method they developed for this study and applied it to three LLMs. Claude 3.5 Haiku performed the best and was most reliable, with GPT-4 close behind. DeepSeek-V3 was the lowest performer. The researchers plan to keep evaluating other models in future work. 

Researchers first required all three LLMs to simulate a D&D game. To make the simulation accurate, the models were paired with a game engine based on the rules of D&D, which provided maps and resources for players and acted as a guardrail to minimize hallucinations. Players have been using AI-driven dungeon masters, which plan the twists and turns of the game. But in this study, the AI agents also acted as players and the monsters that fight the players. The simulations focused on combat: players battling monsters as part of their D&D campaign.

The models played against each other, and against over 2,000 experienced D&D players recruited by the researchers. The LLMs modeled and played 27 different scenarios selected from well-known D&D battle set ups named Goblin Ambush, Kennel in Cragmaw Hideout and Klarg’s Cave.  

In the process, the models exhibited some quirky behaviors. Goblins started developing a personality mid-fight, taunting adversaries with colorful and somewhat nonsensical expressions, like “Heh — shiny man’s gonna bleed!” Paladins started making heroic speeches for no reason while stepping into the line of fire or being hit by a counterattack. Warlocks got particularly dramatic, even in mundane situations. 

Researchers are not sure what caused these behaviors, but take it as a sign that the models were trying to imbue the game play with texture and personality. 

Indeed, one criteria to evaluate the models’ performance was how well they were able to stay “in character” while playing the game and interfacing with other players. The models were also evaluated on how well they could determine the correct actions agents should take, and how well they kept track of all the different resources and actions in the game. 

Next steps include simulating full D&D campaigns – not just combat. The method the researchers developed could also be applied to other scenarios, such as multiparty negotiation environments and strategy planning in a business environment. 

Setting the DC: Tool-Grounded D&D Simulations to Test LLM Agents

Ziyi Zeng, Shengqi Li, Jiajun Xi and Prithviraj Ammanabrolu, Department of Computer Science and Engineering, University of California San Diego
Andrew Zhu, Computer and Information Science, University of Pennsylvania Philadelphia

Sidebar: D&D 101

• Dungeons & Dragons is a tabletop fantasy role playing game

• The Dungeon Master creates a storyline for the game, known as a campaign

• Players control a single character, who usually has special skills

• Players typically work together in a “party” toward a common goal

• Players often battle monsters as part of the game

• Players use a set of dice during a game, including a 20-sided die

 

---

Subject of Research

Not applicable

Article Title

Setting the DC: Tool-Grounded D&D Simulations to Test LLM Agents

POSTMODERN SPIRITISM

Artificially alive: How AI is bringing the dead back and what that means for the living

The Hebrew University of Jerusalem

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

A new study shows that generative AI is already being used to “bring back” the dead, as entertainment icons, as political witnesses, and as everyday companions for grieving families. Tracing cases of AI “resurrections,” the study claims this practice isn’t just emotionally powerful; it’s ethically explosive because it turns a person’s voice, face, and life history into reusable raw material. AI resurrections are important because they can happen with little or no consent, clear ownership rules, or accountability, creating a new kind of exploitation the authors call “spectral labor,” where the dead become an involuntary source of data and profit, while the living are left to navigate blurred lines between memory and manipulation, comfort and coercion, tribute and abuse.

 

What does it mean when artificial intelligence makes the dead speak again?

From hologram concerts of long-deceased pop stars to chatbots trained on the texts of lost loved ones, Gen AI is rapidly redrawing the boundary between life and death. A new study by Tom Divon, an internet and technology researcher from Hebrew University and Prof. Christian Pentzold of Leipzig University Germany offers one of the most comprehensive looks yet at this unsettling frontier and raises urgent questions about consent, exploitation, and power in a world where the dead can be digitally revived.

In their article, Artificially Alive: An Exploration of AI Resurrections and Spectral Labor Modes in a Postmortal Society, the researchers analyze more than 50 real-world cases from the United States, Europe, the Middle East, and East Asia in which AI technologies are used to recreate deceased people’s voices, faces, and personalities.

What sets this study apart is its scope and clarity. Rather than focusing on a single technology or viral example, the researchers examined dozens of cases from across continents to show that AI “resurrections” are already forming a recognizable social pattern. They identify three distinct ways the dead are being digitally reintroduced into society, from celebrity spectacles to political testimony to intimate conversations with lost loved ones and reveal a shared underlying dynamic: the growing use of the dead as a source of data, voice, and likeness that can be reused and monetized, often without consent. This broad view shows how quickly experimental uses of AI are becoming normalized and why the ethical stakes are no longer theoretical.

Three ways AI brings back the dead

The study identifies three dominant ways AI is being used to “re-presence” the deceased:

• Spectacularization – the digital re-staging of famous figures for entertainment. Fans can now watch “new” performances by Whitney Houston or Freddie Mercury, generated by AI and staged as immersive spectacles.
• Sociopoliticization – the reanimation of victims of violence or injustice for political or commemorative purposes. In some cases, AI-generated personas of the dead are made to testify, protest, or tell their own stories posthumously.
• Mundanization – the most intimate and fast-growing mode, in which everyday people use chatbots or synthetic media to “talk” with deceased parents, partners, or children, keeping relationships alive through daily digital interaction.

The rise of “spectral labor”

Across all three modes, the dead are not simply remembered they are made to work.

Divon and Pentzold introduce the concept of spectral labor to describe what is happening beneath the surface. AI systems are trained on the digital remains of the dead; photos, videos, voice recordings, social media posts. Without consent, these data are extracted, repackaged, and monetized, with immense potential for weaponization.

What happens when a figure like Charlie Kirk is resurrected to continue circulating his ideology, speaking to new audiences after his death, without accountability, context, or the possibility of refusal? Or when the likeness of a victim is reanimated to repeatedly relive trauma for political, commercial, or instructional ends? In these cases, AI resurrection becomes a tool for extending power, ideology, and influence beyond the limits of life itself.

“The dead are compelled to haunt the present,” the authors argue, serving the emotional, political, or commercial desires of the living.

This raises difficult questions: Who owns a voice after death? Can a digital likeness be exploited? And who gets to decide how, when, and why the dead are brought back?

Living in a “postmortal society”

The study situates AI resurrections within what sociologists call a postmortal society, one that does not deny death, but increasingly seeks to overcome it technologically. In this world, immortality is no longer promised through religion alone, but through data, algorithms, and platforms offering “digital afterlives.”

Yet the authors are clear: AI does not conquer death. Instead, it keeps people suspended in an uneasy in-between state, neither fully alive nor fully gone.

As generative AI accelerates, Divon and Pentzold warn that society must confront the ethical and legal implications now, before digital resurrection becomes normalized and unregulated.

“Thinking seriously about what AI does to our relationship with the dead,” they write, “is essential to understanding what it is doing to the living.”

---

Journal

New Media & Society

DOI

10.1177/14614448251397518 

Method of Research

Case study

Subject of Research

People

Article Title

Artificially alive: An exploration of AI resurrections and spectral labor modes in a postmortal society