You would not be reading this if it were not for intelligence. The 600 million year journey from our amoebic ancestor’s first nervous system to space travel has been driven by the development and application of intelligence. Luck aside, every major success and failure that has ever occurred is a consequence of the presence or absence of intelligence. The future of humanities happiness and mere existence rests upon having the intelligence to solve problems from pandemics and supervolcanoes to cures for cancer and maybe even death itself.
It is for these reasons I strongly believe that the two most important causes of our time are solving artificial intelligence and maximizing human intelligence. By solving artificial intelligence I mean having the technological capabilities to create an Artificial General Intelligence (AGI) and, even more importantly, find watertight ways to actually control it. By maximizing human intelligence I mean finding ways to allow humans to cognitively achieve more. This includes raising IQ scores, as well as increasing memory, rationality, motivation levels, or anything else that allows humans to think and act in smarter ways.
In a rank ordering of these causes by importance, artificial intelligence must come first. This is because, if we successfully create a friendly AGI that we can control, it will be so much smarter than anything else in existence that it will be able to solve all other problems for us. This includes solving problems that are simply beyond our capacities, as well as problems that are so difficult and costly as to entail little expected return on investment – problems that are technically solvable but where incentives are misaligned. Basically, having a friendly AGI is the magic bullet for everything else. (I will cover the reasoning behind this in further detail in another post). In the meantime, read this for the most eloquent and lucid explanation of why. Read this for the interesting and wise reasoning of Nick Bostrom. Or read our homemade calculations (link coming) for why creating safe AGI is the most impactful cause to work on and donate to.
However, while solving AI is the most compelling end goal, the best means to this end may be through boosting human intelligence. A pretty compelling argument can be made that we will not be able to successfully create a friendly AGI that doesn’t kill everyone without more human intelligence and its beneficial knock-on effects. Increased human intelligence will both fend off other existential risks before we create AGI, and increase the potential that we actually create a safe AGI that doesn’t kill or enslave everyone.
One of the clearest examples of the potential for heightened human intelligence to thwart other existential risks is through increasing the wisdom of national and geopolitical decision making. Major political follies like the election of Trump and Brexit would fail to occur with a smarter general population. Even if these decisions were the right ones – which it seems to be increasingly apparent they are not – they were made for absolutely wrong reasons… Ignorance trumped intelligence. Enhancing the intelligence of a country’s population will increase stability, innovation, and economic growth. This directly and indirectly reduces the chance we all die before having the chance to create friendly AGI.
Concrete findings on the importance of intelligence towards the aforementioned metrics of a country’s success can be found by reading this, this, & this along with this example of the effects of raising IQ in the US outlined in The Bell Curve and later validated by other authors (heard through Gwern):
[If we boosted IQ in the US by a mean of 3 points back in 1994]: “For starters, the poverty rate falls by 25%. So does the proportion of males ever interviewed in jail. High school dropouts fall by 28%. Children living without their parents fall by 20%. Welfare recipiency, both temporary and chronic, falls by 18%. Children born out of wedlock drop by 15%. The incidence of low-weight births drops by 12%. Children in the bottom decile of home environments drop by 13%. Children who live in poverty for the first three years of their lives drop by 20%.”
Finally, as Nick Bostrom points out in his book Superintelligence, the way to successfully control the AGI is only through having more intelligent people working on it. Unlike with most other things where we can gradually improve with trial and error, we only get 1 shot at making AGI right. Only more lucid thought and wise solutions will be our salvation. The sole source of this is through greater quantity and quality of intelligence.
Beyond being a means to the end of AGI – which will then solve every form of human enhancement that we haven’t – increasing human intelligence is a worthwhile end in itself. This is because of the “IQ halo effect” whereby those who have a higher IQ also typically experience an increase in other positive outcomes, ranging from reduced likelihood of committing a crime to being healthier and living longer. I have included two descriptive tables below that show how IQ influences different life outcomes:
And this chart found on pg406, Strenze, Intelligence and Success, ch25 of Handbook of Intelligence Evolutionary Theory, Historical Perspective, and Current Concepts, ed Goldstein et al 2015. (Also found through Gwern if you haven’t realized it yet you should really start reading Gwern… Also sorry for how bad this table looks… This will be fixed soon.):
“Table 25.1 Relationship between intelligence and measures of success (Results from meta-analyses)”
“r correlation between intelligence and the measure of success, k number of studies included in the meta-analysis, N number of individuals included in the meta-analysis”
|Measure of success|
|Academic performance in primary education|
|Job performance (supervisory rating)|
|425||32124||Hunter and Hunter (1984)|
|Job performance (work sample)|
|36||16480||Roth et al. (2005)|
|Skill acquisition in work training|
|17||6713||Colquitt et al. (2000)|
|Degree attainment speed in graduate school|
|5||1700||Kuncel et al. (2004)|
|Group leadership success (group productivity)|
|14||Judge et al. (2004)|
|Promotions at work|
|9||21290||Schmitt et al. (1984)|
|Interview success (interviewer rating of applicant)|
|40||11317||Berry et al. (2007)|
|Reading performance among problem children|
|8||944||Nelson et al. (2003)|
|Becoming a leader in group|
|65||Judge et al. (2004)|
|Academic performance in secondary education|
|Academic performance in tertiary education|
|Having anorexia nervosa|
|16||484||Lopez et al. (2010)|
|Research productivity in graduate school|
|4||314||Kuncel et al. (2004)|
|Participation in group activities|
|Group leadership success (group member rating)|
|64||Judge et al. (2004)|
|Popularity among group members|
|19||2546||DeNeve & Cooper (1998)|
|Procrastination (needless delay of action)|
|7||6062||Griffeth et al. (2000)|
|Recidivism (repeated criminal behavior)|
|32||21369||Gendreau et al. (1996)|
|Number of children|
|Traffic accident involvement|
|10||1020||Arthur et al. (1991)|
|Conformity to persuasion|
|7||Rhodes and Wood (1992)|
|8||2548||Bourhis and Allen (1992)|
|18||Woodberry et al. (2008)|
Do note that some of these sources are quite old and look into a particular correlation yourself if you plan to cite or use it. For example, more recent studies find a decent positive correlation between IQ and physical attractiveness, something which is here shown to have a correlation of -0.04. (For example this, aside from its bold title, the study does a good job describing the difficulties of measuring IQ to attractiveness correlations. This is due to it being well written, shitting on many other similar studies, and having very poor statistical significance itself).
Meanwhile, there are also some new correlations that have also been found. For example, Scott Alexander in this piece presents studies which show “a person with IQ 115 (85th percentile) is 20% more likely to survive to age 76 than an average person with IQ 100.” and moreover that this finding is not due to having more money or being in a better environment but instead that “95% of life span intelligence correlation is due to genetics“. Alexander also notes a positive correlation between height and IQ while providing numerous studies which show there is a correlation of around r = 0.23. I also easily found one for obesity, and it seems that both are almost entirely genetic in origin. A significant correlation also exists for one’s creative potential. As I go into detail later, this correlation is particularly important when it comes to why intelligence matters for creating a better future. This well written paper finds that having an IQ above 89 was a necessary threshold for the quantity of ideas one can generate, 104 for the quality of ones best ideas, and 119 for the average quality of ones ideas (this finding was not statistically significant like the former two but it almost was at p=0.14 and has consensus from other studies around it). Above these thresholds of intelligence, personality factors matter far more but when looking at creative achievement rather than merely potential we find that there is no threshold and that it instead the scale of ones creative accomplishments continues to increase linearly with IQ.
A bar graph showing chances of being obese at age 51 with significant linear drop offs with IQ.
Different scatter plots using segmented regression to find thresholds for IQ needed for certain metrics of creativity. Scatter plot a (top left) is the 89 IQ point threshold found for the number of ideas that one could come up with when asked a “divergent productivity” prompt such as “name all the things you can do with a hairdryer”. Plot B looked at the quality of those ideas generated taking the best two that each person came up with. Plot C looked at the average quality of all the ideas generated by each individual. Plot D looked at actual creative achievements by people such as creating art or writing a blog.
Last, but certainly not least, IQ even correlates with sperm quality! This study finds a correlation of between 0.14 and 0.19 between difference “key indices of semen quality”.
I understand that the line of reasoning I have presented is very abstract and qualitative. It would be incredibly hard to add probability estimates for any one of these claims especially when even the best experts fail to predict something like when AGI might be created or the probability we succeed in designing it correctly. While I cannot provide numbers, I hope that one can agree with the chain of reasoning made and begin to consider just how impactful artificial intelligence and human intelligence are. Try to come up with any other cause areas that can be even nearly as compelling.
A classic response to my assertion that everyone reading this should consider working on AI and Human Intelligence is this: “If everyone started working on AI and Human Potential then who would be creating the next Game of Thrones for us to enjoy? MY LIFE DOES NOT HAVE MEANING WITHOUT GAME OF THRONES!!!”. This is fair enough rebuttal until one acknowledges that only a small number of people will ever read this article and that we are operating in the status quo of virtually nobody working in these spaces. How often does the general public have conversations about the implications of creating a super intelligence and how to control it? What about education policy and how to actually make people more intelligent? How many people think that if only we raised education spending and reduced classroom sizes in the US then our kids would get smarter? There are enough painters, singers, and film producers to already create anything your mind would or wouldn’t want to imagine. Artificial and Human Intelligence involvement is how you can make an impact where others haven’t.
Note also that both of these things have to do with increasing intelligence; just in very different ways. If it is not already clear from the reasoning I have provided why increasing intelligence is so compelling, it can be spelled out with the words innovation and multiplier effects.
For those who haven’t read Peter Thiel’s short book, Zero to One, I would recommend you get around to it. For those who have, I hope we can agree that its main thesis, that going from something not existing, to the very first version of it (literally from 0 to 1) is where the most value is. In other words innovation is incredibly valuable and is the best way to create new value for everyone to enjoy. It is innovation that leads to breakthroughs that fundamentally change the ways in which we live. Without getting too deep into this (read the book!) just think of the way economic growth driven by globalization and technology has reduced global absolute poverty from 42% in 1981 to below 10% now.
As for multiplier effects, boosting intelligence of any kind is the ultimate form of the multiplier effect. Similar to AI but in a less extreme way, making others smarter can allow them to achieve everything that you would have as an individual and more. For example, one of the driving forces behind me spending time writing this post rather than spending the time doing more research into human intelligence enhancement is because if I can get even one person to change what they are working on towards and focus with me on this space, then the return on my time will absolutely be more than worthwhile.
I hope I have not made it seem like boosting intelligence is an unquestionable good in every scenario. I have already tried to emphasize that in the case of Artificial Intelligence we must absolutely make sure that the AGI we create has the best interests of humanity at heart. I think there is much more of a debate to be had about the net benefits of making only a small subset of people ridiculously smart (I will be exploring this in a future post). However, there is much less viable contention around whether or not we should raise the mean intelligence of the human population. This is for the reasons previously given: more economic, political and social stability, increased innovation, support for the successful creation of a safe artificial intelligence, and more people suffering less in their everyday lives.
So, if I have convinced you that artificial and human intelligence are the two most exciting spaces to work in, you must consider which to be more involved with and in what way. Fortunately, these two causes also, broadly speaking, require different skill sets meaning no matter who you are, reader, there is a way for you to play a part in all of this.
(Below I go into ways that you can work directly in these most impactful and exciting of spaces. Bear in mind first and foremost your own skill set and motivations. There are other career tracks not mentioned where you can feasibly incorporate both AI and Human Potential in some way that will have an impact. For example, by having a deep understanding of both of these concepts and then going and being a politician. There are also many other domains where people can do huge amounts of good. However, in the long run all of these other problems can be subsumed by investing now in superior intelligence.)
So, which one should you work on? My best advice is to try both but ultimately it will come down to how well you intuitively understand math. (Here are two interesting pieces on how mathematical ability is reasonably inflexible (h/t Scott Alexander). You either have it or you don’t). If you understand numbers intuitively and enjoy working with them, then AI is the path for you. I tried this path and managed to take Intro to ML with Andrew Ng and play around on Kaggle for a bit before realizing that, while I could be a code monkey and run basic algorithms, I was never going to create new algorithms that changed the world. Here is a great track for how to get good at AI and Machine Learning (the hottest part of it currently) that a friend who is quickly rising the ranks at some of the world’s best AI research institutions has shared: (link coming).
If math is not your thing (join the club!) then maximizing human intelligence might be the track for you. Human intelligence is a very nebulous term and everybody’s definitions of what intelligence is varies. Taking my earlier definition of “finding ways to allow humans to cognitively achieve more” there are many different facets of this which include but are not limited to: IQ, memory, information processing, external systems (like the internet), personality types, and education.
So which of these mechanisms to boost human intelligence is the most promising? Well this is the very question that I will be setting out to answer over the course of writing on this blog. However, until I have more details to specifics to share I will provide a few hints and a framework:
I really like the framework that 80,000 hours uses to think about what opportunities are the most promising: 1. What is the scale of the problem (if it is solved what will the impact be)? 2. How many people are currently working on it (what is my marginal impact as an individual working in this space)? 3. What is the chance that it can actually be solved (how much like rocket science is it)? In addition to these and tied to point #2 is the question of what the current bottlenecks are in the field or a particular problem?
For example, when considering how to make people smarter, it is becoming ever more clear through both my research and personal experience that traditional education is neither the bottleneck nor the most promising to actually make a difference. Of the already very small subset of people who are interested in helping people become smarter (that is the point of education, isnt it??) over 99% of them will be working on some kind of fluffy education reform or new ed-tech product that is going to revolutionize the world. Meanwhile, thanks to this thing called the internet, access to educational material is at the fingertips of everyone if they have the motivation to pursue it. The real bottleneck is in how to actually raise human intelligence through looking at things like nutritional interventions, genetics and others.