Artificial intelligence rarely makes waves in the Czech Republic, but the departure of Tomáš Mikolov, a leading AI scientist, from the Czech Technical University in Prague (ČVUT) was just such an event. The announcement came at the AI Zážeh spolupráce conference two weeks ago, and since then it has been discussed from all angles. I met with Tomáš Mikolov to hear his reasons firsthand.
First and foremost, it was clear that the whole thing had affected him deeply. He recounted stories of rejected grants, strange offers to broker grants, and even a funny story about how information about the evaluators of his application was accidentally leaked. Some time ago, the Ministry of Education, Youth and Sports published an Excel file with the results of the call for proposals. However, it simply hid some rows, so it was possible to read that his proposal for OP JAK project no. CZ.02.01.01/00/22_008/0004582 focused on artificial intelligence was evaluated by Alberto Bonetti, an evaluator in the field of engineering, and Melanie Sophie Dulong de Rosnay, a lawyer.
The server Věda žije (Science Lives) subjected the evaluators to a survey and found that most of them had obtained their PhDs only a few years earlier. The website further states: The entire project application was so complicated that it was practically impossible to fill it out correctly without the assistance of specialized companies (and the associated financial costs), and even then it took several man-months of work that researchers could have otherwise devoted to the true purpose of their profession—science.
Tomáš Mikolov says the same thing. In five years, he has not received a single decent grant. He was rejected, for example, for a “radical change in project management,” when he actually reduced his workload on the project from 0.7 to 0.6 so that he could transfer these funds to students. “The conditions are complicated and constantly changing. A project can be rejected even if you have complied with a condition that was originally required. Without specialized agencies that want a percentage of the project, new scientists don’t stand much of a chance. And I worked with colleagues who have a lot of experience in writing grants,” says Mikolov. He believes the system is inefficient and has used even harsher words. “The system only works for a limited group of people who know how to navigate it,” he adds. Even contacting ministers did not change anything. So he decided that he had done what he could and would not bother with it anymore. He will continue to research artificial intelligence at BottleCap AI, a company he co-founded.
Mikolov is not alone in his criticism of the subsidy system. It has long been said that grants mainly flow to established projects and that evaluators are not always among the most knowledgeable in their field, and there are often quality issues with the results of their work. However, not much has changed. Admittedly, the M17+ methodology was established by Prime Minister Babiš’s cabinet and adopted by Minister Langšádlová, i.e. Prime Minister Fiala’s cabinet. However, it concerns the evaluation of scientific workplaces and is mainly used in EU operational programs, where teams with a better M17+ rating have a better chance of success. And the quality assessment is based on the M17+ methodology.
In reality, however, its contribution to grants is clearly insignificant. The grant environment is turbulent, and without constant monitoring, it is simply unlikely that you will submit a “high-quality” project for a significant amount of money. The Czech system is notorious for its high administrative burden and bureaucracy, where even minor accounting tasks are complicated. This is why professional consulting companies drain the system, working for a commission (around 10%, but sometimes more, depending on the type of project) to develop the project and ensure that everything is in order or “on call,” meaning that in the event of problems, the workplace is asked to remedy the situation rather than being penalized.
Add to this the fact that spending on science and research in the Czech Republic is below the EU average. And it’s no wonder there is such despair.
The solution? None before the elections. After the elections… well, I would say, as always…
It is no wonder then that demotivation ends in burnout, personal demotivation, or, in the best case, departure to the private sector.
PS: In mid-May, Bottlecap AI announced a breakthrough discovery, reducing the cost of training AI models by 50%. Details have not yet been announced, but they are reportedly in talks with those who could benefit from such a discovery.
