This material may not be published, broadcast, rewritten, or redistributed. Quotes displayed in real-time or delayed by at least 15 minutes. Market data provided by Factset . Powered and implemented by FactSet Digital Solutions . Mutual Fund and ETF data provided by Refinitiv Lipper . Concerns remain over AI's impact on young people amid boom Tech expert praises New York's school cellphone ban as social media concerns rise Trump advisor details administration's push to boost AI hiring Kash Patel to close FBI's Hoover building in DC permanently Santa is'PACKING HEAT' during a traffic stop Trump has made AI a'key part' of his agenda, expert says Youth conservative movement has'never been in better hands,' podcast host says Most-used passwords of 2025 include '123456' and'password' Megan Garcia, a mother who lost her son to suicide after he allegedly became emotionally attached to an AI chatbot, discusses the dangers of the technology on'Fox News Sunday.'

'You'll jump into the deep end pretty much immediately,' Altman said while announcing the vacancy. 'You'll jump into the deep end pretty much immediately,' Altman said while announcing the vacancy. 'This will be a stressful job': Sam Altman offers $555k salary to fill most daunting role in AI New head of preparedness at OpenAI will face unnerving in-tray amid fears from some experts that AI could'turn on us' Mon 29 Dec 2025 09.44 ESTLast modified on Mon 29 Dec 2025 10.10 EST The maker of ChatGPT has advertised a $555,000-a-year vacancy with a daunting job description that would cause Superman to take a sharp intake of breath. In what may be close to the impossible job, the "head of preparedness" at OpenAI will be directly responsible for defending against risks from ever more powerful AIs to human mental health, cybersecurity and biological weapons. That is before the successful candidate has to start worrying about the possibility that AIs may soon begin training themselves amid fears from some experts they could "turn against us". "This will be a stressful job, and you'll jump into the deep end pretty much immediately," said Sam Altman, the chief executive of the San Francisco-based organisation, as he launched the hunt to fill "a critical role" to "help the world".

When you purchase through links in our articles, we may earn a small commission. From overheating smart beds and privacy snafus to generative AI foul-ups and bricked smart thermostats, we had a bumper crop of smart home fails this year. Sometimes the smart home doesn't seem so smart, like when a cloud outage renders your devices useless or your smart bed starts roasting you in your sleep. We see these kinds of smart-home snafus every year, but 2025 was a tad different thanks to the arrival of generative AI assistants like Alexa+ and Google Gemini, which arrived with great fanfare but failed to truly blow our minds. So without further ado, we present the biggest smart home flops, fails, and foul-ups of the year, ranked in ascending order.

The battle for AI dominance has left a large footprint--and it's only getting bigger and more expensive. When Sam Altman said one year ago that OpenAI's Roman Empire is the actual Roman Empire, he wasn't kidding. In the same way that the Romans gradually amassed an empire of land spanning three continents and one-ninth of the Earth's circumference, the CEO and his cohort are now dotting the planet with their own latifundia--not agricultural estates, but AI data centers . Tech executives like Altman, Nvidia CEO Jensen Huang, Microsoft CEO Satya Nadella, and Oracle cofounder Larry Ellison are fully bought in to the idea that the future of the American (and possibly global) economy are these new warehouses stocked with IT infrastructure. In the earliest days of computing there were giant power-sucking mainframes in climate-controlled rooms, with co-ax cables moving information from the mainframe to a terminal computer.

Switch 2 games are on sale through Jan. 5 OpenAI is hiring a new Head of Preparedness to try to predict and mitigate AI's harms CEO Sam Altman posted about the role on X, saying the models'are starting to present some real challenges.' OpenAI is looking for a new Head of Preparedness who can help it anticipate the potential harms of its models and how they can be abused, in order to guide the company's safety strategy. It comes at the end of a year that's seen OpenAI hit with numerous accusations about ChatGPT's impacts on users' mental health, including a few wrongful death lawsuits . In a post on X about the position, OpenAI CEO Sam Altman acknowledged that the potential impact of models on mental health was something we saw a preview of in 2025, along with other real challenges that have arisen alongside models' capabilities. The Head of Preparedness is a critical role at an important time, he said.

As machine learning-enabled Text-to-Image (TTI) systems are becoming increasingly prevalent and seeing growing adoption as commercial services, characterizing the social biases they exhibit is a necessary first step to lowering their risk of discriminatory outcomes. This evaluation, however, is made more difficult by the synthetic nature of these systems' outputs: common definitions of diversity are grounded in social categories of people living in the world, whereas the artificial depictions of fictive humans created by these systems have no inherent gender or ethnicity. To address this need, we propose a new method for exploring the social biases in TTI systems. Our approach relies on characterizing the variation in generated images triggered by enumerating gender and ethnicity markers in the prompts, and comparing it to the variation engendered by spanning different professions. This allows us to (1) identify specific bias trends, (2) provide targeted scores to directly compare models in terms of diversity and representation, and (3) jointly model interdependent social variables to support a multidimensional analysis. We leverage this method to analyze images generated by 3 popular TTI systems (Dall E 2, Stable Diffusion v 1.4 and 2) and find that while all of their outputs show correlations with US labor demographics, they also consistently under-represent marginalized identities to different extents. We also release the datasets and low-code interactive bias exploration platforms developed forthis work, as well as the necessary tools to similarly evaluate additional TTI systems.

OpenAI GPT-5.2 replaces GPT-5.1 with behind-the-scenes improvements for presentations, spreadsheets and image interpretation, though users may struggle to notice differences.

Text-to-image generation has recently witnessed remarkable achievements. We introduce a text-conditional image diffusion model, termed RAPHAEL, to generate highly artistic images, which accurately portray the text prompts, encompassing multiple nouns, adjectives, and verbs. This is achieved by stacking tens of mixture-of-experts (MoEs) layers, i.e., space-MoE and time-MoE layers, enabling billions of diffusion paths (routes) from the network input to the output. Each path intuitively functions as a painter for depicting a particular textual concept onto a specified image region at a diffusion timestep. Comprehensive experiments reveal that RAPHAEL outperforms recent cutting-edge models, such as Stable Diffusion, ERNIE-ViLG 2.0, DeepFloyd, and DALL-E 2, in terms of both image quality and aesthetic appeal. Firstly, RAPHAEL exhibits superior performance in switching images across diverse styles, such as Japanese comics, realism, cyberpunk, and ink illustration. Secondly, a single model with three billion parameters, trained on 1,000 A100 GPUs for two months, achieves a state-of-the-art zero-shot FID score of 6.61 on the COCO dataset.

Recent progress in generative artificial intelligence (gen-AI) has enabled the generation of photo-realistic and artistically-inspiring photos at a single click, catering to millions of users online. To explore how people use gen-AI models such as DALLE and StableDiffusion, it is critical to understand the themes, contents, and variations present in the AI-generated photos. In this work, we introduce TWIGMA (TWItter Generative-ai images with MetadatA), a comprehensive dataset encompassing over 800,000 gen-AI images collected from Jan 2021 to March 2023 on Twitter, with associated metadata (e.g., tweet text, creation date, number of likes). Through a comparative analysis of TWIGMA with natural images and human artwork, we find that gen-AI images possess distinctive characteristics and exhibit, on average, lower variability when compared to their non-gen-AI counterparts. Additionally, we find that the similarity between a gen-AI image and natural images is inversely correlated with the number of likes. Finally, we observe a longitudinal shift in the themes of AI-generated images on Twitter, with users increasingly sharing artistically sophisticated content such as intricate human portraits, whereas their interest in simple subjects such as natural scenes and animals has decreased. Our analyses and findings underscore the significance of TWIGMA as a unique data resource for studying AI-generated images.

There has been significant recent progress designing deep generative models that generate realistic sequence data such as text or music. Nevertheless, it remains difficult to incorporate high-level structure to guide the generative process, and many such models perform well on local coherence, but less so on global coherence. We propose a novel approach for incorporating global structure in the form of relational constraints between different subcomponents of an example (e.g., lines of a poem or measures of music). Our generative model has two parts: (i) one model to generate a realistic set of relational constraints, and (ii) a second model to generate realistic data satisfying these constraints. For model (i), we propose a constrained optimization algorithm that infers the relational constraints present in the training data, and then learn a generative model based on the resulting constraint data. In our experiments, we show that our approach significantly improves over state-of-the-art in terms of capturing high-level structure in the data, while performing comparably or better in terms of low-level structure. We also show that using constrained optimization for part (ii) as well leads to increased controllability with little decrease in quality compared to pure learning-based models.