Почему развитие в ИИ стоит начинать с изучения математики и алгоритмов
Руководитель Школы анализа данных Яндекса в подкасте Machine Learning Podcast рассказывает, почему фундамент (матан, линал, теорвер, алгоритмы) — это не скучная теория, а база для работы с ИИ в 2026.
Вы узнаете, как глубокое понимание математики помогает писать эффективный код, отлаживать модели и ориентироваться в разных областях ML. А ещё — почему даже опытным разработчикам полезно возвращаться к фундаментальным дисциплинам.
Перейти к прослушиванию
#подкаст#ML
#подкаст
Энакин Скайуокер. Андрей Зайцев. Но не Дарт Вейдер.
Именно герою Андрея Энакину Оби-Ван Кеноби сказал: «Ты был Избранником! Предрекали, что ты уничтожишь ситхов, а не примкнёшь к ним! Восстановишь равновесие Силы, а не ввергнешь её во мрак!»
Андрей восстановил в своей жизни равновесие Силы и его ролики на канале PRO.GOLOS прекрасны.
О том, как стать джедаем озвучки – эксклюзивный подкаст Андрея Зайцева с Сердитым пряником и Михаилом Хрусталёвым.
Доступно наVK Видео
Доступно на RUTUBE
Доступно на YouTube
Доступно на Яндекс Музыке
Как создать подкаст под ключ? Расскажет @AndPolina
Does AI dream of electric patents?
Google faces a legal problem: lawyers are unsure if they could patent plans created by AI algorithms. The company had filed patents describing a ML technique used to design and map out components in the custom AI accelerator TPU chips.
However, US laws recognize and protect intellectual property created only by "natural persons". Although Google engineers built AI models, after training algorithms generated their products automatically with minimal human effort. Therefore, a legal catch arises: is it permissible to patent the outputs created by these systems?
During the meeting held by US Patent and Trademark Office, Laura Sheridan, senior patent counsel at Google, said that company pursued only patterns on ML models, not the floorplans it had created.
This case remains an important issue for the entire IT industry nevertheless. AI technologies already produce a lot of outcomes that could become valuable intellectual property for a business and entrepreneurs. ML systems can, for example, write a code, hunt for new drugs, and create digital art. That is why application of the patent laws to the AI algorithms’ outcomes should be clarified in the shortest time.
#AI#ML
Researchers at MIT, Stanford University, Intelligence Lab, and the Autodesk AI Lab developed AI that can figure out Lego Instructions
Scientists collaborated to develop a learning-based framework that can travel 2D instructions to build 3D objects. This system called the Manual-to-Executable-Plan Network (MEPNet) was successfully tested on Lego sets and Minecraft-style building plans.
So it will definitely help people who were driven mad with confusing Lego manuals. But the key idea is to integrate neural 2D keypoint detection modules and 2D-3D projection algorithms for high-precision prediction of unseen components.
Interpreting 2D instructions could be tricky for AI. The key problems are identifying correspondence between 2D and 3D objects, and dealing with a lot of basic objects, which could be assembled into complex forms. «It requires inferring 3D poses of unseen components composed of seen primitives," the researchers said.
At first, MEPNet analyses the current state of Lego set and creates 3D model of all components. Then the algorithm predicts a set of 2D keypoints and masks for each component.
Once that's done, the 2D keypoints "are back-projected to 3D by finding possible connections between the base shape and the new components." The combination "maintains the efficiency of learning-based models, and generalizes better to unseen 3D components," the team wrote.
The full paper of MEPNet is available via the link. And the algorithm’s code is also posted on GitHub.
#AI#ML