• IBM's AI watermarking method protects models from theft and sabotage

    3 monthes ago - By VentureBeat

    What if machine learning models, much like photographs, movies, music, and manuscripts, could be watermarked nearly imperceptibly to denote ownership, stop intellectual property thieves in their tracks, and prevent attackers from compromising their integrity? Thanks to IBM's new patent-pending process, now they can. In a phone conversation wi...
    Read more ...