Alexandre V. Borovik on AI as "intellectual pet food for the masses"
"A total assault, in all spheres of human activity"
Homo sapiens is not necessarily the only possible subject of capitalist proletarianization. If AI approaches or attains the horizon of singularity, the vistas that open up are not therefore those of inevitable capitalist collapse, but rather of the elevation of machine capital as a literally automatic subject autonomous from human beings.
(Nick Dyer-Witheford, Atle Mikkola Kjøsen and James Steinhoff, Inhuman Power)
Alexandre V. [Sasha] Borovik, the author of the book pictured above, is one of the rare mathematicians who not only devotes considerable energy and imagination to reflect on the dynamics and presuppositions of our discipline, both in its own terms and its relation to the broader society as a material as well as intellectual and cultural force, but is also willing to speak out about his thoughts, even at the risk of destabilizing some of the discipline’s most cherished orthodoxies.
I met Sasha twice at conferences in England on the philosophy of mathematical practice, and again, virtually, through the work of the Azat Miftakhov Committee. He has strong and original opinions on a great variety of subjects and he states them forcefully.1 The letter reproduced below on ChatGPR and automation of , which originally appeared on Sasha’s blog in 2023, is a good introduction to his thought. Note his refusal to be distracted by fashionable questions like “can machine learning produce sentient beings?”
I’m grateful to Sasha for allowing me to republish his letter here. In his last message to me he made a point that deserves no less attention, and that I will certainly find a way to quote in one of my presentations at the Joint Mathematics Meetings in Seattle next month:
There is a dangerous mismatch between the declining value of mathematics in everyday life of people and the dramatic boost to the role of mathematics in warfare. The advent of AI is likely to accelerate both tendencies. Unfortunately, I do not see any discussion of that issue in our community.
It was Marx who said “supply takes demand, if necessary, by force”.
A letter to mathematics and computer science colleagues
Dear Colleagues,
Very recently I wrote to a few friends saying that I expected ChatGPT in its next version becoming able to solve every algebra and calculus problem in A Level (the end of school exams in England) and similar school exams in other countries. For that, ChatGPT simply should be shown how to identify what looks as an algebraic, logarithmic, differential etc. equation or a system of equations or inequalities and plug this thing into one of already existing maths problems solvers, for example, the Universal Math Solver,
https://universalmathsolver.com/
— it does more than finding an answer, it produces a complete step-by-step write-up of a solution.
But this important symbolic threshold was passed much earlier than I expected. Conrad Wolfram posted on his blog on 23 March an announcement “Game Over for Maths A-level”, https://www.conradwolfram.com/writings/game-over-for-maths-a-level. A quote:
“The combination of ChatGPT with its Wolfram plug-in just scored 96% in a UK Maths A-level paper, the exam taken at the end of school, as a crucial metric for university entrance. (That compares to 43% for ChatGPT alone).”
This means that undergraduate pre-Calculus and Calculus undergraduate exams will follow quickly.
I think it is dangerous to sit and wait while we are overrun by events. I suggest that we have to address the issues on the global scale: changes in the technological and socio-economic environments of education will soon affect hundreds of millions of children in dozens of countries and later become truly global. It is the scale of the problem which is the issue.
There is nothing special in the ChatGPT, it is only one of a dozen AI systems of enhanced functionality which have suddenly appeared on the market. They are pushed by some of the mightiest transnational corporations to the market where, unlike many other markets, the rules of the supply-side economics apply in their full strength (remember the story of iPod? Or selfie sticks?). It does not matter, what we think and feel about the AI: very soon, it will be everywhere around us. It was Marx who said “supply takes demand, if necessary, by force”. A classical example, which is likely to be reproduced in the case of AI, is the multibillion pet food industry: the concept of pet food was invented and forced on people (now called, in TV commercials, “pet parents”) in the late 1950s by the American meat packing industry which by that time completely saturated the American market (for human consumption) and looked for new directions to expand. For billions of people around the globe, AI will become an intellectual pet food for the masses. And we have to take into account that the supply-side push of the AI on people, is likely to be a total assault, in all spheres of human activity, much wider than education.
In many countries, politicians, state bureaucrats, theoreticians of mathematics education, and school teachers led by them, made everything possible to turn students into a kind of biorobots trained for passing school exams. And here comes the moment of truth: if real robots pass exams with much better marks — what is the purpose of the current model of mathematics education?
And we should not be distracted by general philosophical questions of the kind “can machine learning produce sentient beings?” The real, and immediate issue, is the disruption which will be caused by still non-sentient AI in the human society (made of sentient beings).
It is interesting to glimpse a politician’s view of these issues. Please see below some examples of uses of mathematics as given by Rishi Sunak, Prime Minister of the UK, in his speech on improving attainment in mathematics, 17 April 2023, https://www.gov.uk/government/speeches/pm-speech-on-improving-attainment-in-mathematics-17-april-2023 . Interestingly, the speech was given at the London Screen Academy – this is why examples start with “visual effects”, etc.
You can’t make visual effects without vectors and matrices.
You can’t design a set without some geometry.
You can’t run a production company without being financially literate.
And that’s not just true of our creative industries. It’s true of so many of our industries.
In healthcare, maths allows you to calculate dosages.
In retail, data skills allow you to analyse sales and calculate discounts.
And the same is true in all our daily lives…
… from managing household budgets to understanding mobile phone contracts or mortgages.
With a possible exception of the first line (about visual effects1), all that in 5 (or at most 10) years from now will be done by a combination of AI and specialist mathematics (or maybe accounting) tools — and done much better than 90% of people can do. For example, an app on a smartphone which has access to all financials accounts of the owner – bank accounts, credit cards, tax account, mortgage, etc. and linked to powerful AI servers on the Internet, will be able to take care of household budgets. This app will ask the user, after each contactless payment in the shop, under which heading this payment should be entered in the ledger of the household budget, offering most likely options (maybe deducing them from the shops’ names, like Mothercare or Bargain Booze).
It is widely accepted now that in most areas of human activity ChatGPT and other AI systems are no more than imposters faking answers to questions they do not understand.
However, routine mathematical by their nature tasks of household budgeting, etc. are likely to be important exceptions — because they are intrinsically well structured and less ambiguous. And AI paired with mathematical problem solving software will pass standard school exams better than students or their teachers can do.
I summarise the situation in three bullet points:
What we see now is a slow motion car crash of the traditional model of mathematics education. Sunak (and practically everyone in the area of education policy) are asleep at the wheel and do not see the road ahead. But in the education policy, we have to look at least 14 years ahead – this is the length of school education (in the UK), from 4 to 18 years of age.
Most politicians are able to think ahead only on the time scale of the election cycle, 4 or 5 years. They cannot comprehend the scale of quantities and magnitudes (the latter include time) involved in economic and social problems (and even less so in all the mess around the climate change).
Most politicians lack basic skills of project management and do not understand that work on a serious project should start with the step-by-step reverse planning from the target to the present position.
This why I appeal to professional mathematicians and computer scientists:
Of all people involved in some way in mathematics /computer science education, you are perhaps the only ones free from mental handicaps listed in the three bullet points above. Let us discuss, at first perhaps only in our circle, this fundamental question:
What kind of mathematics education is needed in the era of AI?
Perhaps we have to split the question:
What kind of mathematics should be taught
(a) To future developers, controllers, masters of AI?
(b) To the general public, the users (and perhaps victims) of AI?
If these questions are not answered in our professional communities, we should not expect an answer coming from elsewhere.
Alexandre Borovik
18 April 2023
By the way, I don’t agree with all of them. For example, in a discussion of the proof of the Classification of Finite Simple Groups in the middle of an article entitled “A mathematician’s view of the unreasonable ineffectiveness of mathematics in biology,” he wrote
We have to admit that mathematics faces an existential crisis.
Without switching to systematic use of computer-based proof assistants, and corresponding changes in the way how mathematics is published and taught, mathematics will not be able to face challenges of biology – moreover, it is likely to enter a spiral of decay.
The recent proliferation of extremely long proofs does indeed represent a challenge for mathematicians, but I see no sign of crisis or decay. That’s a subject for another discussion, however.