There is no doubt that 2025 will be remembered as the year of slop. A popular term for incorrect, weird and often downright ugly AI-generated content, slop has rotted nearly every platform on the internet. Enough slop has accumulated over the past few years that scientists can now measure its effects on people over time. Researchers at the Massachusetts Institute of Technology found that people using large language models (LLMs) such as those behind ChatGPT to write essays show far less brain activity than those who don't. And then there are the potential ill-effects on our mental health, with reports that certain chatbots are encouraging people to believe in fantasies or conspiracies, as well as urging them to self-harm, and that they may trigger or worsen psychosis.

GPU prices could follow RAM's big rise OpenAI's house of cards seems primed to collapse In 2025, it fell behind the one company it couldn't lose ground to: Google. OpenAI CEO Sam Altman speaks during the US Federal Reserve Board of Governors' Integrated Review of the Capital Framework for Large Banks Conference at the Federal Reserve in Washington, DC, on July 22, 2025. OpenAI is in a far less commanding position than it was following the public release of ChatGPT a few short years ago. Back in 2022, the sudden popularity of ChatGPT sent Google into a panic . The company was so worried about the possibility of the upstart chatbot disrupting its Search business, executives sounded a code red alert inside of the company and called Sergey Brin and Larry Page out of retirement to help it formulate a response to OpenAI.

This ebook is available only for subscribers. In this exclusive subscriber-only eBook, you'll learn about a new method that scientists have uncovered to look at the ways our bodies are aging. We're learning more about what vitamin D does to our bodies Jessica Hamzelou OpenAI's new LLM exposes the secrets of how AI really works Will Douglas Heaven We're learning more about what vitamin D does to our bodies The sunshine vitamin could affect your immune system and heart health. The idea that machines will be as smart as--or smarter than--humans has hijacked an entire industry. But look closely and you'll see it's a myth that persists for many of the same reasons conspiracies do. The experimental model won't compete with the biggest and best, but it could tell us why they behave in weird ways--and how trustworthy they really are.

Plus: Australia's social media ban for teens has just come into force. Stardust Solutions believes that it can solve climate change--for a price. The Israel-based geoengineering startup has said it expects nations will soon pay it more than a billion dollars a year to launch specially equipped aircraft into the stratosphere. Once they've reached the necessary altitude, those planes will disperse particles engineered to reflect away enough sunlight to cool down the planet, purportedly without causing environmental side effects. But numerous solar geoengineering researchers are skeptical that Stardust will line up the customers it needs to carry out a global deployment in the next decade. MIT Technology Review Narrated: Is this the electric grid of the future?

With a quarter of people in the UK using AI to find products, marketers must not only appeal to shoppers directly but to AI bots and their opaque decision-making process. With a quarter of people in the UK using AI to find products, marketers must not only appeal to shoppers directly but to AI bots and their opaque decision-making process. 'What to buy Dad for Christmas': is retail ready for the AI shopping shift? Consumer test drive: can AI do your Xmas gift shopping for you? While traditional internet search, social media - especially TikTok and Instagram - and simply wandering a local high street will still be the main routes to presents for most this year, about a quarter of people in the UK are already using AI to find the right products, according to PricewaterhouseCoopers.

GPU prices could follow RAM's big rise You can start using the apps for free, with some limitations. A ChatGPT user asks the chatbot to make an image more vibrant through Photoshop. At the time, the company said more software was on the way, and now one of the most popular professional applications is available through the chatbot. Starting today, you can access Photoshop, Acrobat and Adobe Express inside of ChatGPT. All the apps are free to use through OpenAI's website, though before you can begin generating PDFs and illustrations using Acrobat and Adobe Express, you'll need to sign into your Adobe account.

Keio University President Kohei Ito (left) and OpenAI Chief Strategy Officer Jason Kwon sign a memorandum of understanding on Tuesday in Tokyo. Keio University is working with OpenAI to integrate artificial intelligence into its education system. Keio University President Kohei Ito and OpenAI Chief Strategy Officer Jason Kwon signed a memorandum of understanding Tuesday, making Keio the first Japanese university to form a strategic partnership with the producer of ChatGPT. "We will develop an environment where students and researchers can proactively learn and utilize AI," Ito said. In a time of both misinformation and too much information, quality journalism is more crucial than ever.

Sam Altman's appearance on is part of a larger charm offensive currently being waged by the tech establishment. OpenAI CEO Sam Altman was at the center of Silicon Valley's most visible publicity push in recent memory Monday night when he appeared on . In a predictably softball interview with host Jimmy Fallon, Altman explained how ChatGPT has helped him alleviate the anxiety that comes with being a new parent. It was a distinctly clever, if somewhat surprising, choice from Altman who has mostly kept his personal life out of the media spotlight. But Altman is a salesman, and a good salesman understands the optics of good television.

A large amount of work has been done in Multi-Agent Systems (MAS) for modeling and solving problems with multiple interacting agents. However, most LLMs are pretrained independently and not specifically optimized for coordination. For example, existing LLM fine-tuning frameworks rely on individual rewards, which require complex reward designs for each agent to encourage collaboration. To address this challenge, we model LLM collaboration as a cooperative Multi-Agent Reinforcement Learning (MARL) problem. We develop a multi-agent, multi-turn algorithm, Multi-Agent Group Relative Policy Optimization (MAGRPO), to solve it, building on current RL approaches for LLMs as well as MARL techniques. Our experiments on LLM writing and coding collaboration demonstrate that fine-tuning multiple LLMs with MAGRPO enables agents to generate high-quality responses efficiently through effective cooperation. Our approach opens the door to using MARL methods for LLM collaboration and highlights the associated challenges.

The emergence of large language models (LLMs) has sparked much interest in creating LLM-based digital populations that can be applied to many applications such as social simulation, crowdsourcing, marketing, and recommendation systems. A digital population can reduce the cost of recruiting human participants and alleviate many concerns related to human subject study. However, research has found that most of the existing works rely solely on LLMs and could not sufficiently capture the accuracy and diversity of a real human population. To address this limitation, we propose CrowdLLM that integrates pretrained LLMs and generative models to enhance the diversity and fidelity of the digital population. We conduct theoretical analysis of CrowdLLM regarding its great potential in creating cost-effective, sufficiently representative, scalable digital populations that can match the quality of a real crowd. Comprehensive experiments are also conducted across multiple domains (e.g., crowdsourcing, voting, user rating) and simulation studies which demonstrate that CrowdLLM achieves promising performance in both accuracy and distributional fidelity to human data.