FrontierMath is a new benchmark for testing how well current AI systems do at solving math problems. According to Epoch AI,
FrontierMath presents hundreds of unpublished, expert-level mathematics problems that specialists spend days solving. It offers an ongoing measure of AI complex mathematical reasoning progress.
Epoch AI itself is a research organization that aims to understand the development of AI better through research.
In my opinion as a mathematician, the creation of FrontierMath as a rigorous testing suite for AI is extremely dangerous, because it provides an attractive benchmark for those who develop AI. It will only further motivate AI researchers and companies to make AI even more devious and capable. AI is dangerous and needs to be destroyed.
There is no social benefit to companies like Epoch AI. Interestingly, there is a quotation on their website by Professor Yoshua Bengio of Mila – Quebec AI Institute, that says
Epoch AI’s research brings both clarity and depth to the rapidly evolving landscape of artificial intelligence. Their work is a relevant resource for anyone who wants to understand the trajectory of AI and its societal implications.
Such statements are misleading in every way. The societal implications of AI are already clear: extremely dangerous with no net benefit. Sadly, organizations like Epoch AI and people like Professor Bengio constantly spread falsehoods that AI should be studied in depth to understand it. Of course, they are right in a way: the forces of modern technology and capitalism needs to study AI to make it just as safe as it needs to be so that people accept it long enough to be completely subjugated by it.
The creation of FrontierMath is a perfect example of the naïvety of mathematicians and computer scientists. In particular it displays a complete lack of wisdom on the part of the mathematicians who have collaborated in the creation of these problems. These problems will only accelerate the development of AI, and aid in exactly the sort of regulation of AI that will make it easier for large corporations to integrate AI further into our daily lives. Sadly, we have created a society where it is possible for so many to sit in the academic ivory tower and amuse themselves with abstract knowledge, without having to take any responsibility for its practical use.
The truth is, we don’t need to analyze AI. We need to destroy it, and ban all development of AI and machine learning. In particular, that means banning all machine-learning academic programs, banning all commercial AI development, and permanently deleting all code related to machine learning. I condemn Epoch AI and FrontierMath for creating a tool that will only aid the development of AI.