Machine Learning World

Новый видосик с пошаговыми инструкциями https://youtu.be/rFwAibqvDrE

В продолжение темы про работу https://www.youtube.com/watch?v=G6Sq8cJkTrI

Свежый видос про поиск работы https://www.youtube.com/watch?v=ap4bGxdxBvg

Flutter приложение для приютов с животными. День 19 Небольшой апдейт по тому как же следить за помощью ChatGPT #ANDevHowTo #ANDevHowToFlutter

For the first time: a powerful model from GPT-4 in your living room. Preliminary results confirm the rumors, the +Command R model has officially become the first open model ever to bypass GPT-4 in the Chatbot-arena test. --- Last weekend the +Command R model was released (from Cohere) and gained a lot of interest due to its high performance in several different tests. Link: https://huggingface.co/CohereForAI/c4ai-command-r-plus Hours after the release, rumors began to circulate that this model is so successful that it is even more successful than GPT-4. Today we receive for the first time an official result from one of the most important tests today for the language model ability test: Chatbot Arena. For the first time: An open model bypasses GPT-4. The model is available for download, for lecture and training at your home. Question: How many of you are not sure how to operate the model on cheap hardware? Do you want me to make a video (or streaming) how to use it? Beyond the tremendous contribution to the community, this is also another huge achievement: the model is significantly smaller than GPT-4 and contains only about 6% of the amount of parameters in it. It's a trend, all models are becoming very effective lately. Restrictions: True. The license does not allow commercial use of the model. True. 105 billion parameters are not simple for a lecture "at home". True. The closed models are still leading (by a pretty small gap already). But this is a turning point. That many were skeptical about her. From now on you have a better model than GPT-4-0314 at home. (Up to GPU)

Друзі, дуже прошу підтримати мій тікток канал підписавшись на нього. Це допоможе робити більше цікавих відео. #ANDevHowTo https://www.tiktok.com/@miidosvid

На днях спрашивал как обрабатывать 1ТБ логов в день на 500-600$/мес Вот мое решение для данного вопроса. https://youtu.be/9GJuKWSSXDY?si=WiX4USSwAjtdziPu #ANDevHowTo #ANDevHowToArchitecture #ANDevHowToLogging

Продовжимо тему про роботу мозку, так поговоримо в контексті навчання. Якщо ви задавались питанням чому ви забули майже все чого вас вчили в школі та універі, то це відео дасть відповідь. А також розповість як вчитися щоб не забувати. #ANDevHowTo #ANDevHowToLearning https://youtu.be/0Y4PyiWZ4Q8?si=MKFLmuWIqKi4aimD

Мой учитель с Мех-Мата Михаил Городний, когда то сказал "Хороший математик отличается от плохого тем что знает когда задача решена". Эти слова сегодня как то сами всплыли у меня в голове пока я размышлял об ИИ и интелекте вцелом. И тут у меня возник вопрос - Быть может мы уже создали искусственный интеллект, просто еще не поняли это? Именно эту тему я пытаюсь раскрыть в этом видео. https://www.youtube.com/watch?v=S7SWRrkcVvY&ab_channel=%D0%9C%D1%96%D0%B9%D0%94%D0%BE%D1%81%D0%B2%D1%96%D0%B4%7C%D0%90%D0%BD%D0%B4%D1%80%D1%96%D0%B9%D0%9D%D1%96%D0%BA%D1%96%D1%88%D0%B0%D1%94%D0%B2

Can ChatGPT generate just a white picture? It seems like it should be easy, right? But alas, it's not. As you can see, it only succeeded when I described the specific steps on how to do it, and even then, it took a second attempt. This is why current solutions cannot replace humans and are merely a more intelligent search and another tool in the hands of people to simplify their daily work. Right now, I'm using it to write an application in Flutter, with which I have zero experience. And although there are quite trivial things, even on those, it falters and goes in circles because it's not capable of combining data for new solutions, and it can't even understand from the context that it has already recommended this. #ANDevHowTo #ANDevHowToAI #ANDevHowToML