"We've been really pushing on'thinking,'" says Jack Rae, a principal research scientist at DeepMind. Such models, which are built to work through problems logically and spend more time arriving at an answer, rose to prominence earlier this year with the launch of the DeepSeek R1 model. They're attractive to AI companies because they can make an existing model better by training it to approach a problem pragmatically. That way, the companies can avoid having to build a new model from scratch. When the AI model dedicates more time (and energy) to a query, it costs more to run.

In May, I was confronted with a robot version of my writer self. It was made, at my request, by a Silicon Valley startup called Writer, which specializes in building artificial-intelligence tools that produce content in the voice of a particular brand or institution. In my case, it was meant to replicate my personal writing voice. Whereas a model like OpenAI's ChatGPT is "trained" on millions of words from across the Internet, Robot Kyle runs on Writer's bespoke model with an extra layer of training, based on some hundred and fifty thousand words of my writing alone. Writer's pitch is that I, Human Kyle, can use Robot Kyle to generate text in a style that sounds like mine, at a speed that I could only dream of.

Within four days of ChatGPT's launch, Habib used the chatbot to build QuickVid AI, which automates much of the creative process involved in generating ideas for YouTube videos. Creators input details about the topic of their video and what kind of category they'd like it to sit in, then QuickVid interrogates ChatGPT to create a script. Other generative AI tools then voice the script and create visuals. Tens of thousands of users used it daily--but Habib had been using unofficial access points to ChatGPT, which limited how much he could promote the service and meant he couldn't officially charge for it. That changed on March 1, when OpenAI announced the release of API access to ChatGPT and Whisper, a speech recognition AI the company has developed.

When OpenAI, the San Francisco company developing artificial intelligence tools, announced the release of ChatGPT in November 2022, former Facebook and Oculus employee Daniel Habib moved quickly. Within four days of ChatGPT's launch, Habib used the chatbot to build QuickVid AI, which automates much of the creative process involved in generating ideas for YouTube videos. Creators input details about the topic of their video and what kind of category they'd like it to sit in, then QuickVid interrogates ChatGPT to create a script. Other generative AI tools then voice the script and create visuals. Tens of thousands of users used it daily--but Habib had been using unofficial access points to ChatGPT, which limited how much he could promote the service and meant he couldn't officially charge for it.

Writer, the only full-stack generative AI platform built for business, today launches three new proprietary large language models (LLMs) designed for enterprise-ready generative AI. Palmyra Small (128M), Palmyra Base (5B), and Palmyra Large (20B) are the only in-production LLMs that were trained on a set of data specifically curated to power AI use cases for the enterprise. Palmyra Small and Base LLMs are accessible via free download on Hugging Face. Writer's enterprise customers have their generations all powered by Palmyra Large through the Writer platform, and Writer enterprise customers are also now able to integrate generative AI capabilities directly into their products and to scale and improve their experience with Writer via Writer's new API to Palmyra Large. "Writer was built from the ground up to take AI into the enterprise. It all starts with our proprietary model, where customers own their inputs, training data, and outputs," said May Habib, CEO of Writer.

Generative AI is coming for videos. A new website, QuickVid, combines several generative AI systems into a single tool for automatically creating short-form YouTube, Instagram, TikTok and Snapchat videos. Given as little as a single word, QuickVid chooses a background video from a library, writes a script and keywords, overlays images generated by DALL-E 2 and adds a synthetic voiceover and background music from YouTube's royalty-free music library. QuickVid's creator, Daniel Habib, says that he's building the service to help creators meet the "ever-growing" demand from their fans. "By providing creators with tools to quickly and easily produce quality content, QuickVid helps creators increase their content output, reducing the risk of burnout," Habib told TechCrunch in an email interview.

Writer is an AI-powered tool for checking and guiding content creators in organizations where voice and branding are essential. Its new feature CoWrite does that writing itself -- but don't worry, this isn't quite the content apocalypse we've been worried about. CoWrite is the latest in a new wave of tools that use large language models like GPT-3, but modify them using "fine tuning," a common phrase but with a special meaning in the machine learning world. Basically it means giving the big, general model a specific set of content to imitate more closely than the rest of the language it understands -- a bit like telling an image creation model to make a picture in a certain style by feeding it examples. Writer's tools already do this to a certain extent, ingesting style guides and other data to provide a live style-check service: "use this preferred word instead of that," or "use active voice in headlines," depending on what your organization likes.

Argonne looks to exascale and beyond, sorting out the relationship between computing and experimental facilities, the need for speed and AI's role in making it all work. In 1949, physicists at the U.S. Department of Energy's (DOE) newly minted Argonne National Laboratory ordered the construction of the Argonne Version of the Institute's Digital Automatic Computer, or AVIDAC. A modified version of the first electronic computer built at the Institute for Advanced Study in Princeton, New Jersey, it was intended to help solve complex problems in the design of nuclear reactors. With a floor area of 500 square feet and power consumption of 20 kilowatts, AVIDAC boasted remarkable computing power for the time. It possessed a memory of 1,024 words (about 5.1 kilobytes in total), could perform 1,000 multiplications per second, and had a programming capability that allowed it to solve problems consistently and accurately. Today, your smart phone can store around 100 million times more data, and can do in a single second what would have taken AVIDAC two months.

The following article is part of a series on Argonne National Laboratory's efforts to use the predictive power of artificial intelligence, specifically machine learning, to advance discoveries in a broad range of scientific disciplines. High-energy physics and cosmology seem worlds apart in terms of sheer scale, but the invisible components that comprise the field of one inform the composition and dynamics of the other -- collapsing stars, star-birthing nebulae and, perhaps, dark matter. For decades, the techniques by which researchers in both fields studied their domains seemed almost incompatible, as well. High-energy physics relied on accelerators and detectors to glean some insight from the energetic interactions of particles, while cosmologists gazed through all manner of telescopes to unveil the secrets of the universe. " … it would be interesting to know if image classification techniques from machine learning that have been used successfully by Google and Facebook can simplify or shorten the development of algorithms that identify particle signatures in our 3D detectors."

EcoFactor is one of several startups with a cloud computing platform to manage and analyze data from smart thermostats and other home energy devices. But it also specializes in using that data to monitor and predict performance problems and impending failures of the air conditioners keeping houses cool. That kind of technology could have a lot of value to the companies that make heating, air conditioning and ventilation equipment -- enough to make it worth owning. On Tuesday, HVAC giant Trane announced it has acquired EcoFactor's energy analytics software for an undisclosed sum. Trane, a brand of Ingersoll Rand, plans to integrate EcoFactor's "unique artificial intelligence (AI) capabilities for energy efficiency and HVAC fault detection" into its existing Nexia home automation line.