Close Menu
Technology Mag

    Subscribe to Updates

    Get the latest creative news from FooBar about art, design and business.

    What's Hot

    The oldest Fire TV devices are losing Netflix support soon

    May 24, 2025

    The Breville Oracle Jet Is Like the iPad of Home Espresso Machines

    May 24, 2025

    The Best Coffee Pod Machines for Hot and Cold Brew

    May 24, 2025
    Facebook X (Twitter) Instagram
    Subscribe
    Technology Mag
    Facebook X (Twitter) Instagram YouTube
    • Home
    • News
    • Business
    • Games
    • Gear
    • Reviews
    • Science
    • Security
    • Trending
    • Press Release
    Technology Mag
    Home » Databricks Has a Trick That Lets AI Models Improve Themselves
    Business

    Databricks Has a Trick That Lets AI Models Improve Themselves

    News RoomBy News RoomMarch 26, 20253 Mins Read
    Facebook Twitter Pinterest LinkedIn Reddit WhatsApp Email

    Databricks, a company that helps big businesses build custom artificial intelligence models, has developed a machine-learning trick that can boost the performance of an AI model without the need for clean labeled data.

    Jonathan Frankle, chief AI scientist at Databricks, spent the past year talking to customers about the key challenges they face in getting AI to work reliably.

    The problem, Frankle says, is dirty data.

    ”Everybody has some data, and has an idea of what they want to do,” Frankle says. But the lack of clean data makes it challenging to fine-tune a model to perform a specific task. “Nobody shows up with nice, clean fine-tuning data that you can stick into a prompt or an [application programming interface]” for a model.

    Databricks’ model could allow companies to eventually deploy their own agents to perform tasks, without data quality standing in the way.

    The technique offers a rare look at some of the key tricks that engineers are now using to improve the abilities of advanced AI models, especially when good data is hard to come by. The method leverages ideas that have helped produce advanced reasoning models by combining reinforcement learning, a way for AI models to improve through practice, with “synthetic,” or AI-generated, training data.

    The latest models from OpenAI, Google, and DeepSeek all rely heavily on reinforcement learning as well as synthetic training data. WIRED revealed that Nvidia plans to acquire Gretel, a company that specializes in synthetic data. “We’re all navigating this space,” Frankle says.

    The Databricks method exploits the fact that, given enough tries, even a weak model can score well on a given task or benchmark. Researchers call this method of boosting a model’s performance “best-of-N.” Databricks trained a model to predict which best-of-N result human testers would prefer, based on examples. The Databricks reward model, or DBRM, can then be used to improve the performance of other models without the need for further labeled data.

    DBRM is then used to select the best outputs from a given model. This creates synthetic training data for further fine-tuning the model so that it produces a better output the first time. Databricks calls its new approach Test-time Adaptive Optimization or TAO. “This method we’re talking about uses some relatively lightweight reinforcement learning to basically bake the benefits of best-of-N into the model itself,” Frankle says.

    He adds that the research done by Databricks shows that the TAO method improves as it is scaled up to larger, more capable models. Reinforcement learning and synthetic data are already widely used, but combining them in order to improve language models is a relatively new and technically challenging technique.

    Databricks is unusually open about how it develops AI, because it wants to show customers that it has the skills needed to create powerful custom models for them. The company previously revealed to WIRED how it developed DBX, a cutting-edge open source large language model (LLM) from scratch.

    Share. Facebook Twitter Pinterest LinkedIn WhatsApp Reddit Email
    Previous ArticleLenovo’s ThinkPad X9 14 Kickstarts a New Era, but at What Cost?
    Next Article OpenAI rolls out image generation powered by GPT-4o to ChatGPT

    Related Posts

    Let’s Talk About ChatGPT and Cheating in the Classroom

    May 23, 2025

    Kentucky’s Bitcoin Boom Has Gone Bust

    May 23, 2025

    Fire Breaks Out at a Data Center Leased by Elon Musk’s X

    May 23, 2025

    Anthropic’s New Model Excels at Reasoning and Planning—and Has the Pokémon Skills to Prove It

    May 23, 2025

    The Time Sam Altman Asked for a Countersurveillance Audit of OpenAI

    May 22, 2025

    Politico’s Newsroom Is Starting a Legal Battle With Management Over AI

    May 22, 2025
    Our Picks

    The Breville Oracle Jet Is Like the iPad of Home Espresso Machines

    May 24, 2025

    The Best Coffee Pod Machines for Hot and Cold Brew

    May 24, 2025

    Whoop is reportedly replacing defective MG trackers

    May 24, 2025

    Our Favorite Computer Monitors for PC Gaming

    May 24, 2025
    • Facebook
    • Twitter
    • Pinterest
    • Instagram
    • YouTube
    • Vimeo
    Don't Miss
    News

    Twelve South’s slick 3-in-1 charging stand has dropped to a new low price

    By News RoomMay 24, 2025

    Memorial Day marks the unofficial start of summer, and if you somehow managed to skip…

    Fujifilm’s X Half, a New OnePlus Tablet, and Fender’s GarageBand Rival—Your Gear News of the Week

    May 24, 2025

    X is down

    May 24, 2025

    15 Best Memorial Day Tech Deals

    May 24, 2025
    Facebook X (Twitter) Instagram Pinterest
    • Privacy Policy
    • Terms of use
    • Advertise
    • Contact
    © 2025 Technology Mag. All Rights Reserved.

    Type above and press Enter to search. Press Esc to cancel.