I was starting to feel quite uneasy. This paper hit the spot. Right on time.
As a researcher in applied AI in the '00s, I loved purely the challenge of solving a problem. And the WOW effect produced when showcasing the techno-magic, of course. I was into developing rendering engines before and now I enjoyed the complementarity of computer vision and computer animation. There is inherent magic in the complementarity of analysis and synthesis that is highly rewarding if you work in a visual domain. As a young scientist, I proposed a back-propagation algorithm - first implementing it via shaders and then using the first version of CUDA. And I really did not know how to apply this 100x speedup to any practical task!
At the very beginning of the '10s, I enjoyed massive budgets when leading the creation of AI for virtual production via text input. But we never had to use a substantial part of that we never suffered for computational power or data. Rather on developers. The visual analytics startup for the retail industry, a year before that, we built a competitive one with a lower budget on training data markup and computational resources.
That started to change quickly following years, with the mass adoption of deep learning. First, competition for training data. Then, enormous budgets for GPU compute.
Frankly, I spent 2016 to 2019 building AI products that I could not support financially and I nearly burned out. But maybe I was working hard, not smart.
These years we are bombarded with facts on enormous budgets used to create and own LLMs and have to accept the fact that the AGI age is nearly here and it will be owned by Big tech.
Is that so? FOMO is about emotions overtaking reason. If feeling subdued because of missing the hype train or feeling that AI is a battlefield for Big tech, please read this paper and then disconnect from technical media for a while.
Choose Your Weapon:
Survival Strategies for Depressed AI Academics
Author:
Bogdan Sevriukov.
