5q vs AI

Last week I completed the MITx course :    MITx course : Machine Learning with Python: from Linear Models to Deep Learning .   I would highly recommend it to anyone who is starting their AI journey. The course provides a great foundation and shows how to get started quickly. There will be another release in a few months, so I suggest that you prepare for it. The prerequisites listed in the about section are not to be underestimated, especially if you won't be able to spend a lot of time during the run of the course. You need to know those basics upfront, or you will just have to spend a lot more time to be able to complete exercises, exams, and to gain knowledge after all. The audit version gives access to the same lectures and exercises as the upgraded track, except the midterm and final exams. The two tests give very interesting and unusual problems, and it is interesting to solve them or later to see the correct answers. How much you will learn depends on three connected things

 I was nicely surprised the other day about an invitation to meet Bard. Google's new conversational AI had some interesting responses to some of my questions. We chatted about current world political scene, how to save the planet,  how to become a better human, solving wave equations and about this blog. Bard liked my blog and expressed interesting overview about a recent presidents meeting. I was a just a little disappointed that Bard's advises about saving Earth and becoming a better human were more like a generic list with the first 10 things that come in mind, but actually all were valid advices for my questions. Bard appears to remember some short history of our conversation. Was able to catch a question I had just half an hour ago, but returned some random stuff, very slightly related, when I asked" Do you remember my questions from yesterday?".  But the thing that mostly impressed me was the name Bard. It is from Azimov's Robot's stories! The set of boo

About the MITx ML Course As I mentioned in the last post I am enrolled in the  MITx course : Machine Learning with Python: from Linear Models to Deep Learning . I am not going to lie, it is hard. You need a really good math foundation to grasp what is going on in the videos. I am struggling to find time and was considering to just make notes from the lectures and later use them as a reference when I eventually get to those topics. I missed the first two units completely, started in the middle of the second and the learning curve is really steep. Until yesterday I was seriously considering just to get out of the speed lane. What changed my mind was a post  I found in the wild parts of the internet and one of the lectures I had last night. Both, the post and the lecture, were talking about the same thing. The professor explained it by using mathematics. The post tried to explain it by a silly analogy. Hearing the scientific, and the real explanation made my head to explode- Eureka, I fin

 The first month of this project Although I had this idea in my mind since December, I actually decided to start about a month ago. I had a plan for my progress, but as usual the reality is different. After a month I have enough data to get an idea what  needs to be optimized and to think of ways to do it. I can divide the what needs to be optimized in two general groups -  optimizing the learning itself and making more time for it. The process of learning Recently I  started getting emails from a course I did enroll in last year and completely forgot  MIT: Machine Learning with Python: from Linear Models to Deep Learning . On top of all I had planned, this course shifted all the priorities because the free access expire May and I am still considering to purchase the certificate option that might extend the access. At the time I logged into edx, the first unit was already out, together with a prep unit,  and I missed most of the deadlines. The second unit is active right now, I have po

NVIDIA, Cuda and Ubuntu It is so good to have a system administrator at home!  I wanted to install PyTorch and found something very interesting in the process- Nvidia is not playing fairly with Ubuntu.  Following the official installation steps for Cuda ran into the following error.    If you get yourself into the same situation, and you have no clue what to do, my advice is to find someone to help. I asked for help and the overall prices took half an hour. I totally see myself wandering on stack overflow for hours (if not days) about this. The tricky part was to get the proper Nvidia drivers installed and working, and once they were ready Cuda installed smoothly.  Once I got Cuda, switched to the installation of PyThorch. It was trivial. Used the matrix on their official website  to pick my desired combination and found the the necessary command at the bottom of the matrix. My Cuda is v.12.0, so my choice in the matrix was the highest possible 11.7. Finally, a quick check if it is wo

Because t he hardest part is to start,  I am starting with something easy - a roadmap and a story. Once I heard a very good motivational story for a man that tried to prove that any common person can finish a marathon. He hasn't run very much until this point and was the perfect person to prove this hypothesis. For a year and something he kept training. Then he completed a marathon and was pround that he proved his idea. But he was wrong. He was no longer the common person that started, he had become a marathon runner. I will try that very same strategy. Will practice and learn every day and let's see what a common person can achieve in the AI domain for a year. The roadmap will be updated as I start discovering the leaves and will be the starting point if you want to find something on the blog. It has six main branches and I will probably work in parallel on several of them. I usually learn by starting from the fundamentals, even if it is something already familiar to me. I m