IQ Check: Still deeply shocked by Deepseek?
Mere common sense, suggests that if one has more intelligence is bound to be better in artificial intelligence too. Risking being called “racist”, last time I checked the Asians had higher average IQ that us Westerners, in the same manner that us Westerners have higher average IQ than others.
The DeepSeek Shock Should Not Be a Surprise
The recent shock caused by DeepSeek should not have come as a surprise to those of us who have collaborated with Chinese and Asian scientists.
In mathematics and computer science, many of the greatest scientists are of Chinese or Indian origin. The undisputed mathematical prodigy of our time, Terence Tao, is of Chinese descent, while DeepMind’s Sir Demis Hassabis is half Greek, half Asian. Meanwhile, Indians are leading some of the Western world’s finest tech companies.
In our field, sharing ideas is inevitable. Personally, I consider one of my achievements to be my ability to read, and attempt to comprehend and apply (a small part of ) the work of Prof. Tao.
As there are no borders in intelligence, the same ought to apply to artificial intelligence.
So please, spare yourselves the argument that DeepSeek merely “copied” OpenAI’s efforts — unless, of course, we are to assume that OpenAI was the result of parthenogenesis.
Rest assured that soon enough Artificial Intelligence will be treated as a Global Common Good.
Economic Competition and the Myth of Inevitability
The growing economic competition between China and the West has fueled tensions to the extent that some have begun contemplating the inevitability of military conflict between the “two worlds.” To me, proponents of such ideas appear to have little understanding of the issues at hand.
DeepSeek is not the first Chinese innovation to outsmart us. Let me offer an example from one of the core technologies that power AI computing: Graph Databases.
The GraphDB Parallelism
Consider China’s progress in the GraphDB landscape, exemplified by Ultipa:
Their hybrid model delivers performance at N:24 depth that surpasses Neo4j’s N:3 depth.
I mention Neo4j because it was the original pioneer in the field of GraphDBs. While Neo4j did not invent the mathematics governing GraphDBs, it created the business and market for them. I still remember meeting the Neo4j team in Oxford in the early 2010s as they attempted to “sell” the — obvious to experts — advantages of graph databases.
For the past couple of years, I have been a happy Ultipa client. It offers the performance I need at a fraction of the cost, enabling me to continue my research despite limited funding.
Clarification: I still use Neo4j and OrientDB — as it will take years to convert my code to Ultipa, while it is always good to keep playing on each other’s strengths while ensuring maximum security via functional decomposition-spread among different platforms
DeepSeek, Intelligence, and Innovation
Some experts suggested that the DeepSeek scientists ended up inventing something smarter because we opted to restrict or ban the sales of advanced chips.
Based on my own experience, I can confirm that necessity is indeed the mother of invention across the board.
A piece of policy advice here: If, following a perceived threat, one does not want its adversaries to acquire new — perhaps even previously unimaginable — skills, restrictions won’t cut it if the said adversary is not lacking intelligence. Let’s stick to our principles of open and free markets for all.
It is sometimes argued that restrictions serve to buy time. Fair enough — but one should never assume that the true beneficiary of “buying time” can be accurately predicted. The delayed party might struggle, or it might innovate its way into an even stronger position.
Generally, it is wise not to disturb one’s adversaries while they are busy making “mistakes” — if, of course, one can assume that becoming an adversary was a mistake on their part in the first place.
Ironically, it seems that the Chinese leadership and people have applied this exact strategy with respect to us.
For years, they allowed us to believe that they were… well, incapable of original ideas. After all, it had been ages since they invented the abacus, the compass, gunpowder, paper, porcelain, etc. Meanwhile, they were content to work under inhumane conditions, producing our supposedly genius, innovative, and creative ideas.
Some in the West bought into this propaganda — the belief that creativity has a pro-Western political and/or racial orientation. Obviously, they were in for a rude awakening.
Race and Intelligence
Since time immemorial, warmongers have invoked national pride by demonizing their adversaries based on race. The rhetoric of “us” — the best, the great, the special — versus “them” — the barbarians — echoes throughout history, even in the writings of Herodotus.
Here is a guide, therefore, for all these warmongers to help them realize where exactly we all stand in the grand scheme of things.
Mere common sense suggests that if one possesses higher intelligence, they are likely to excel in artificial intelligence as well.
So, let’s examine where different nations stand in terms of “plain vanilla” IQ.
Is There an Objective Measurement of IQ by Nation?
It is impossible to measure IQ with complete accuracy or objectivity.
You don’t have to go far: Just consider yourself taking an IQ test while the temperature is 18 degrees Celsius, a similar test in 45 degrees and thirsty and again in -15 degrees and hungry. Or, taking a similar test during the Summer holidays and again two months in the Academic year. You will be surprised with yourself displaying an up to 20% deviation in your performance!
Nevertheless, there are attempts to objectively measure it and the OECD’s PISA Intelligence Rankings provide our best available tool for assessing IQ performance at a national level.
As we all know, net intelligence is only a contributing factor to varying measure of “success(*)”, whereas other factors do have impact to the measurement of IQ which is subject to the following:
- Data Quality: In some cases, data came from non-representative samples (e.g., small studies, specific regions rather than the entire country, or older data that may not reflect current trends).
- Cultural Bias: IQ tests are subject to cultural, linguistic, and socioeconomic biases that make cross-country comparisons challenging.
- Ecological Validity: Critics question whether a single number (mean IQ) can capture the complex educational, economic, health, and cultural factors that influence test outcomes.
Below is a broad overview of the publicly reported PISA 2018 results, grouped (informally) by continent or region. These scores are the mean performance of 15-year-old students in three core domains: Reading, Mathematics, and Science.
Dataset:
## Africa
| Country | Reading | Mathematics | Science |
|----------|--------:|------------:|--------:|
| Morocco | 359 | 368 | 374 |
| Tunisia | 361 | 353 | 363 |## Asia
| Country/Region | Reading | Mathematics | Science |
|--------------------------|--------:|------------:|--------:|
| B-S-J-Z (China) | 555 | 591 | 590 |
| Hong Kong (China) | 524 | 551 | 517 |
| Macao (China) | 525 | 558 | 544 |
| Singapore | 549 | 569 | 551 |
| Japan | 504 | 527 | 529 |
| South Korea | 514 | 526 | 519 |
| Chinese Taipei (Taiwan) | 503 | 531 | 516 |
| Malaysia | 415 | 440 | 438 |
| Thailand | 393 | 419 | 426 |
| Indonesia | 371 | 379 | 396 |
| Israel | 470 | 463 | 462 |
| United Arab Emirates | 432 | 435 | 434 |
| Qatar | 407 | 414 | 419 |
| Saudi Arabia | 399 | 373 | 386 |
| Jordan | 419 | 400 | 429 |
| Lebanon | 353 | 393 | 384 |## Europe
| Country | Reading | Mathematics | Science |
|----------------------|--------:|------------:|--------:|
| Estonia | 523 | 523 | 530 |
| Finland | 520 | 507 | 522 |
| Poland | 512 | 516 | 511 |
| Ireland | 518 | 500 | 496 |
| United Kingdom | 505 | 504 | 505 |
| Germany | 498 | 500 | 503 |
| France | 493 | 495 | 498 |
| Sweden | 506 | 502 | 499 |
| Denmark | 501 | 509 | 493 |
| Netherlands | 485 | 519 | 503 |
| Belgium | 499 | 508 | 499 |
| Spain | 477 | 481 | 483 |
| Italy | 476 | 487 | 468 |
| Russia (Partner) | 479 | 478 | 478 |
| Greece | 457 | 451 | 452 |
| Turkey | 466 | 454 | 468 |## North America
| Country | Reading | Mathematics | Science |
|----------------|--------:|------------:|--------:|
| Canada | 520 | 512 | 518 |
| United States | 505 | 478 | 502 |
| Mexico | 420 | 409 | 419 |## Central & South America
| Country | Reading | Mathematics | Science |
|--------------------|--------:|------------:|--------:|
| Chile | 452 | 417 | 444 |
| Uruguay | 427 | 418 | 426 |
| Costa Rica | 426 | 402 | 416 |
| Brazil | 413 | 384 | 404 |
| Colombia | 412 | 391 | 413 |
| Peru | 401 | 400 | 404 |
| Argentina | 402 | 379 | 404 |
| Panama | 377 | 353 | 365 |
| Dominican Republic | 342 | 325 | 336 |## Oceania
| Country | Reading | Mathematics | Science |
|-------------|--------:|------------:|--------:|
| Australia | 503 | 491 | 503 |
| New Zealand | 506 | 494 | 499 |Interpreting PISA Scores
- Mean OECD score in each domain is typically set around 480–500 points.
- A 30-point difference in PISA roughly equates to one year of schooling, according to OECD benchmarks.
- Performance can be influenced by a variety of factors, including socioeconomic conditions, language background, education policies, cultural context, and more.
These results give a broad snapshot of how 15-year-olds in various countries perform on standardized tasks in reading, math, and science — but they are not direct measures of “national IQ.” They are often used as proxies for the quality of a country’s education
Important Notes
- Not all countries participate in PISA, and some participate only in certain cycles.
- The People’s Republic of China was represented in 2018 by specific provinces/cities (B-S-J-Z = Beijing–Shanghai–Jiangsu–Zhejiang). These results are not necessarily representative of all of China.
- A few countries had issues with data quality or did not meet PISA standards in specific domains; those scores may be absent or flagged in the official reports.
- Many countries in Africa and some in other continents did not participate in PISA 2018.
- For the full official dataset, including standard errors, subgroup data, and any updated/revised numbers, visit the OECD’s PISA website.
Political Correctness, National Pride, and the Real Path to AI Leadership
It is about time to start acknowledging the obvious: If one wants to lead in the Artificial Intelligence “race,” investment in the fundamental factors that influence “plain vanilla” IQ is a prerequisite:
- Universal Education
- Public Health
- Economic Growth
Only then can we be assured that cultural factors (as well as subjective, pride-enhancing — but impossible to quantify — concepts such as “creativity,” “innovation,” “open-mindedness,” and a nation’s past, present, or future “greatness”) will be the sole remaining determinants.
Spontaneous $500bn or $1tr investments in AI initiatives, may sound impressive to the masses and/or the investors, but to subject matter experts are insignificant if they are not followed by investment in these three fundamental pillars. Soon enough, today’s technological investment of $500bn will have a value of less than $1bn. If this sounds like an exaggeration consider the following:
Back in 2012 Oxford used to have the fastest GPU (NVIDIA) based supercomputer in the UK. It costed about $3 million and was delivering the staggering speed of about 114 TeraFlops. It was a “toy” for exclusive few. All of us who had the pleasure to work on Emerald went through CUDA training in the summer, so that we could focus on CUDA learning undisturbed. Not only Computer Scientists, but Physicists, Engineers and even postgrads from the Physical Sciences (Medicine, Biology, BioTech) were invited to learn program this impressive beast.
Nowadays you can buy an 104 TeraFlop Graphics Card with less than $2000 or an 114 TeraFlop variant with $2500.
Epilogue
Believing one’s own propaganda is a dangerous mistake — especially in fields where collaboration is necessary, but one chooses instead to focus on competition.
Never forget: the act of marketing innovation is a world apart from truly innovating, just as the marketing of intelligence stands entirely distinct from genuine intelligence.
Footnote on “success”:
Success isn’t one-size-fits-all. Most humans need to see, hear, touch, smell… something to feel genuinely fulfilled — sometimes because it’s harder for them to recall and enjoy the mere “idea” of an object they can’t physically experience anymore.
Meanwhile, those with stronger abstract thinking skills can feel just as satisfied simply by reflecting on what others need to constantly sense or possess/own. Ultimately, success on the basis of intelligence alone can be defined in ways that may seem unimaginable to most, and doesn’t necessarily align with more common value systems.
One of the challenges of those who posses limited intelligence will be their (limited) ability to deal with those who do. Given the bad job that they had done in treating interesting humans in the past, it is easy to understand their fear of how an intelligence powered future will treat them and their own notion of “success” with a version that they cannot even sense.
