Hello, I wanted to confess that I failed the first try at the Ph.D. program because I changed my job. In the next few months, I will try to get back on track but again, from the beginning. You may ask why you changed such a fantastic job you had? I will not answer this here. All I can say is that I do not want to answer. Now I am part of something I love, in a lovely team, with more thoughtful people than me… Btw, did you know that if you are the most competent person in the room… it is the wrong room :-). The Ph.D. program is essential to me, especially in the AI/ML field I love. So I picked up one LinkedIn contact and wrote to ask for help… similar to that strategy, that did not happen precisely this way, but the overall strategic move is confirmed. So I asked one business owner can I join the project and was accepted. Like Steve Jobs advised…
Now in September, I will try again, my application is preparing, and then we gonna see what happens. But, you may also ask, what are the fruits of failure? I mention I leave and have a new opening in my career. Also noteworthy is that I started my own company and focused on lessons learned in my first Ph.D. year of study. Thanks to technical university contacts, I found customers interested in image recognition in the field I have learned. I also built the OpenCL version of the Darknet to run on an industrial PC with Intel Core CPU and GPU. The first task I got was to anonymization the cars on the JPG frames, stored for the record, by blurring the front windows of the vehicles and car plates, It was a lovely experience, and my own model was able on i5 CPU to run 10 FPS thanks to OpenCL version.
Another fruit of my failed Ph.D. program was to prepare the first pre-print publication, which you can find on the Ph.D. tab on this blog. For example, I was really impressed that the words “Darknet OpenCL macOS” put my Ph.D. mentor on top and me in the Google web search engine. With quite many few popular links to the work we did. So far darknet on the github got 130 stars and few forks. Also, I am delighted I got 2 pull requests and 2 more contributions to this project.
In the end, I would like to write that I believe one day AMD GPUs beats on OpenCL the… competitions in the field :D. Right now, their top GPUs do not run “thanks to” wrong drivers on GNU/Linux and on the best cloud providers… I tried Google Cloud and Microsoft Azure on VMs with GPUs. So, in the queue, AMD GPUs, Instinct, Radeon Pro, Radeon, and Navi 20, work but only in theory on Ubuntu. The only hope is on macOS, which supports AMD cards beautifully ;-). I wish one day someone figured out that macOS as a Unix via SSH server is the best AI/ML solution :D.