Genius quote by Einstein

“Everybody is a genius. But if you judge a fish by its ability to climb a tree, it will live its whole life believing that it is stupid.” - Albert Einstein


One of the differences between Einstein and Newton is that Newton didn’t really care that people liked his opinions. Einstein might not have cared either, but spontaneously Einstein’s opinions were very popular with the general public, whereas Newton’s opinions did not. If a person tries to please the public, or if he selects which opinions he will make public based on the expectation that those opinions will be well received by the public, he will be less likely that his opinions are good representations of reality. This is a clear example.

If all people really were geniuses, the term “genius” would lose its meaning, just as if all people were giants or if all astronomical objects were stars. The classification of an astronomical object as a star is not an arbitrary criterion. There are clear and strong reasons to distinguish a star from a planet or a quasar. A star has a core temperature high enough to produce nuclear fusion and shine substantially in the visible spectrum. The visible spectrum is not well defined, it could be somewhere between 360nm and 780nm or it could be 380nm to 740nm or something like that, and the distribution of wavelengths can also invade one of these bands a little bit, or not, beyond undergo slight variations, so some boundary objects such as brown dwarfs may behave like stars at times, but not at other times. But with the exception of these boundary objects, the vast majority will behave like a star or a non-star. The main determining factor is the mass, which in turn determines the core temperature and determines the spectral distribution of light emission. Jupiter, for example, emits a lot of infrared light, but not visible light (it just reflects), so it is not classified as a star.

For classification of giants the situation is similar. A person 2.30 m would probably be recognized by everyone as a giant, and a person 1.70 m would not be. But there is no well-defined height that separates the group of “giants” from “non-giants”, because there is no univocal criterion, as in the case of stars, of “emitting light”. So the criterion decision turns out to be arbitrary. It can be determined that over 2 m is giant, or over 3s is giant, or 4s, or any other value. It would also be possible to do a hormonal study and try to identify growth hormone concentrations in a certain proportion and use that as a criterion, or do a factor analysis of various physical and perhaps even behavioral characteristics and use that to stratify people into groups of “giants” and “non-giants”. Or even based on a genetic mapping, among other criteria.

Terman used IQ test scores for this classification, but there are serious limitations because the difficulty level of the questions is too low to correctly discriminate at the genius level. Tests at the time could measure up to about 135, maybe 140 (nominal ~197, like Harding, Langdon, Ward, Bryzman, etc.). Langdon and Hoeflin built new tests that allowed them to raise this threshold to around 160, maybe 165 (nominal ~193, like Rosner, Langan, Marilyn, etc., or ceiling 202 if using the Grady Towers standard). But these norms assume an inadequate assumption that scores are normally distributed. In any case, the tests of the 1970s-1980s began to measure correctly at a higher level and with this it became possible to get closer to making reliable diagnoses of genius.

Terman’s study of 1528 very talented children was a complete failure in the predictive power of these tests at very high levels, although it was successful in the 70 to 130 range. of the study were far above average in academic output and financial success. But it was tragicomic at the highest levels (180+) because no one selected by Terman won a Nobel or similar, but 2 children who were refused the study won Nobel Prizes in Physics, making it evident that the test grossly failed to predict very high levels of intellectual production.

I tried to promote some advances in this area, in 2000, with the Sigma Test, using problems other than those that can have algorithmic solutions, and with the Sigma Test standard, described in 2000 and applied in 2003. The real ceiling of the Sigma Test is controversial . Some people find it closer to 190 or 200, others find it closer to 165-170 like other high range IQ tests, just being different in the kind of thinking it requires, but not having a substantially different difficulty level. The nominal ceilings of high range IQ tests can be 190, 200 or even 250 like some Cooijmans tests, but this is a statistical inconsistency with the standardization method.

It would be okay to have a ceiling of 250 or 300 or 1000, as long as the score was not determined by a scale with forced fit to a normal distribution with a mean of 100 and a standard deviation of 15 or 16. If the standardization method consists of forcing the scores to an adjustment to the theoretical rarity level of a Gaussian distribution, then there are two points that need to be respected, the maximum IQ in a sample with 4000 elements will be close to 3.48 standard deviations or IQ=155.7. Of course, if this sample of 4,000 is selected, the situation changes, but the correct way to use this sample to estimate what the distribution would look like in a population of 8 billion is not trivial, and extrapolations need to be carried out very carefully. This has been done very wrongly, with bizarre results, and more and more bizarre over the years.

“In an ideal world”, standardization needs a large, randomized sample, preferably the entire population, or at least 10% of the population. But this is operationally infeasible. So you have to do what you can with what you have.

Second, the reliable ceiling is determined by the sample size. If you try to extrapolate anything, based on the fact that the sample examined is selected, the necessary adjustments are complex and there is a very high risk of causing very large distortions, reaching extremely inflated results. This problem is heavily criticized by Libb Thims, he makes heavy and incorrect criticisms, because he apparently believes that these errors are intentional (he accuses them of having commercial objectives). I don’t think they are intentional. There is a real and great difficulty in correctly measuring these things.

That’s why the “genius” rating based on clinical IQ tests is a joke. The “genius” rating based on high range IQ tests is limited, it depends on the test, it depends on the person. The best tests come close to touching the skill level that needs to be assessed, but there are factors other than difficulty that need to be considered. I’ll comment a bit on that a bit later on.

First, I’d like to finish my analysis of Einstein’s statement: saying that everyone is a genius is a convenient statement for someone who intends to run for president or gain popularity in some way, but giving the word “genius” such a trivialized meaning is not it is neither conceptually useful nor etymologically appropriate. The efforts of Terman, Wechsler, Pintner, Levine, Johnson, and others to stratify intellectual levels into reasonably well-defined ranges based on test scores are interesting and may have some use, including for special education, for diagnosing some diseases and others. purposes. But there are two major problems.

The first is that the difficulty of questions on traditional tests used in clinical practice is a joke if used for IQs above 150, and even high range IQ tests do not measure correctly above 165 or 170. In an article by Garth Zietsmann, in which he presents a standard estimate for the Power Test, he states that the average IQ of a Ph.D. from MIT is about 144 and from a Nobel laureate in Science it is 155. I wrote an article in 2001 contesting this assessment, which seems far from reality. Maybe the average IQ at MIT is really close to 145-150, but the average IQ of a Nobel laureate in Physics or Chemistry is usually way above 155, maybe close to 170 to 180. The problem is that most tests used to evaluate these people (future Nobel laureates) have a nominal ceiling close to 160 and an actual ceiling close to 135. So even if they get all the questions right, the maximum score is 160, and as many tests (WAIS, for example) include knowledge questions , there is a risk that the person will get some answers wrong due to lack of knowledge (little relationship with intelligence) and have less than 160 or even less than 150, or make mistakes for some other reason (for looking for more creative solutions, for overestimating the question etc.), as in the case of Feynman, and these people may score well below the ceiling on the test used, even though their true IQ is much higher than the ceiling.

If Nobel Prize winners in Science were evaluated with psychometric instruments properly standardized, with an adequate level of difficulty and with challenging and stimulating problems that they would feel motivated to work hard and solve, I find it difficult that the average IQ of Nobels in Science would be below 170.

So perhaps a realistic and reasonable cut-off point for separating genius from non-genius is somewhere around 165. This brings up another important question cited by HoonYoung Kim: whether a genius is someone who scores high on a test or is someone who has produced new knowledge in a relevant area. I think both are, but while the person only scored on a high range IQ test, this indicates latent potential for genius, while innovative intellectual output is concrete evidence of genius.

Some high range IQ test problems are quite difficult and possibly suitable for measuring intelligence at the real level of genius. But there are some important differences between these problems and the kind of problems that geniuses solve.

  1. If you miss 10% or 20% of the items in a test, your score will certainly suffer greatly. But if Aristotle, Newton or Einstein get 50% of the problems they work on wrong, that doesn’t negate the importance of the problems they get right.

  2. In a test you cannot choose which questions you would like to work on. There is a predefined list of tasks and you have to solve it. But often you don’t like various items and you don’t work passionately and doggedly, you don’t do your best at it. If it was a much more difficult problem, but you enjoyed working on it, it could have turned out much better.

  3. In real-world problems, the person usually does not have a well-defined statement telling them what they need to do, nor is there a statement providing the numbers they need to use, nor how they should organize the information in the statement to elicit an answer. , or a ready-to-apply formula. The person needs to find out what should be done, he needs to organize the structure that represents the problem, test some possibilities until he finds the appropriate structure, plan the tests to verify how this structure should be, often the solution is neither complete nor totally correct. , is just a good approximation, but is sufficient for some purposes. There is often no single answer, or a “right” answer. The problem can have many useful answers, some of which are more useful, more economical, more efficient, etc. The real jungle is much more complex, and some problems have no end, it’s not something where you come to a result and say “done, I’m done”. It is a continual search for successive improvements, like Aristotle’s Physics, then Buridan, then Galileo and Newton, then Einstein, Planck and their friends, and this is certainly not the final chapter of this story. It’s just the stage we’re living in, where our mathematical models for reality try to fit the level of precision and accuracy of the instruments available. But there are already some inconsistencies and in a few years or decades another revolution will be needed to adjust the model (or models) to the experimental data.

If an intelligence test is not designed with all of this in mind, the test may measure well up to a point, but above that point it will be a caricature that does not reflect intelligence itself. Current tests may work well for IQs up to 165 or 170, perhaps too optimistically to reach 175, but they hardly ever go beyond that. Some people tested with very high scores may actually have 195 or 200 or higher, because just because the tests don’t measure correctly doesn’t mean the people tested aren’t as bright. It just means that the tests don’t provide enough evidence for that skill level as long as the questions don’t match that intelligence level.

So we don’t have an appropriate criterion for unequivocally determining whether a person should be given the attribute “genius”, although there are some reasonable subjective criteria and some reasonable objective criteria. And whatever the criteria adopted, this statement of Einstein’s does not harmonize with the reasonable meanings of “genius”. I don’t think Einstein really thought that, but if he did, he was wrong. I think he said that because he was a pop star, charismatic, likeable, and that kind of statement is consistent with his personality, it gains popularity because it satisfies people’s fantasies with a sweet illusion that doesn’t have to be realistic; just be convenient and pleasing to the eyes of the crowd.

Note: My origital text is wrote in Portuguese. This is an automatic translation from Portuguese to English. Sorry if it got too bad. I’ve tried to select a few words that avoid distortion, like “astronomical object” instead of “astro” because Google incorrectly translates “astro” to “star”, but it’s possible I have let slip some mistakes.