Genetically Engineered Intelligence
There are a lot of unknowns about the future of intelligence: artificial intelligence, uploading, augmentation, and so on. Most of these technologies are likely a ways off, or at least far enough away to confound predictions. Genetic engineering, however, presents a very near term and well understood possibility for developing greater intelligence.
A recent news story published in South China Morning and discussed on Steve Hsu’s blog highlights China’s push to understand the genetic underpinnings of intelligence. China is planning to sequence the full genome of 1000 of its brightest kids, in the hopes of locating key genes responsible for higher intelligence. Behind the current project is BGI, which is aiming to be (or already is) the largest DNA sequencing center in the world.
Suppose that intelligence has a large genetic component (reasonable, considering estimates for heritability). Suppose that the current study unveils those components (if not this study, then likely another study soon, perhaps with millions of genomes). Then with some advances in genetic engineering China could quickly raise a huge population of incredibly intelligent people.
Such an endeavor could never be carried out on a large, public scale in the West, but it seems China has fewer qualms.
The timescales here are on the order of 20 years, which are relevant compared to most estimates for AI and WBE. More, genetic engineering human intelligence seems to be on a much more predictable path than other intelligence technologies. For both these reasons I think understanding, discussing, and keeping an eye on this issue is important.
What are the ramifications for
AI research? FAI? In particular relating to enhanced humans speeding further development
Whole Brain Emulation research?
Other technologies that may pose existential risks (nanotech, biotech, etc, especially in light of the fact that it may be China leading the way)?
The potential for recursive feedback? (Smarter scientists engineering smarter scientists. Less worrisome due to timescales)
Of course, there are a host of other interesting questions relating to societal impact, both near and long term. Feel free to discuss these as well.
Folks at the Singularity Institute have been watching the reproductive biotechnology area for quite a while, e.g. it plays a significant role in the Uncertain Future modeling app. Some relevant points:
Embryo selection technology (which is the only existing way to apply genetic findings on IQ) on its own has limited impact: if you produce a hundred embryos and have a perfect predictive model of intelligence, you could get a boost of perhaps a standard deviation for a typical couple. You can already get this benefit by using donor sperm and eggs, if one is willing to have genetically unrelated kids, and China doesn’t seem to have strongly encouraged doing so. Eggs from elite donor females are also scarce.
It gets much more powerful with the technology for artificial gametes, which let you make numerous embryos, boost the supply of eggs that can be obtained from a given elite donor to meet demand, and have multiple generations in vitro (create many embryos, cultivate artifical gametes from the best combinations, and repeatedly cross them to get what would otherwise take hundreds of years in only one). Artificial gametes are a pretty natural extension of stem cell technology, but people talk about having them in ten years, not the next couple years (when genomics will get to the point of plumbing the genetic architecture of intelligence and other traits conducive to scientific progress).
The expected combination of behavioral genomics and improvements in artificial gametes or genetic engineering technology makes the idea of lasting stagnation or social collapse look less plausible, and concentrates more of the probability mass for AI and WBE in this century than otherwise.
The increase of intelligence relative to other inputs in the scientific process makes it more likely that scientists and developers will see and respond wisely to risks, making it more plausible that AI development will proceed safely.
The effect of smarter political leadership will be attenuated and delayed: it takes time to rise to high positions.
Mass improvements of the electorate/selectorate will be even slower, as it will take a number of decades for enhanced kids to approach a majority, especially given likely limited adoption (even with big subsidies and measures like relaxation of the one-child policy for enhanced kids).
The potential gains from enhancement are much larger for those willing to have children genetically unrelated to them, or regions which encourage this.
Those working to reduce existential risk should shift intellectual effort at the margin to projects that are time-sensitive (e.g. the chance that AI or engineered extinction plagues will be easy to produce sooner than generally thought) and expect future enhanced folk to do more of the cognitive heavy lifting.
Efforts to transmit relevant information and values to young people/the next generation will become more important in the next 20 years or so. Ensuring that enhanced kids are exposed to ideas that encourage them to reduce existential risk as they grow up looks more important. It would be good to have a clear and accessible picture of the risks landscape available by then.
Encouraging somewhat faster and more evenly distributed adoption of the enhancement technologies looks desirable. In Western countries, this means relaxing or avoiding restrictive regulations, spreading pro-enhancement memes, and the development of the personal genomics industry (companies like 23andme or Counsyl).
At any given time, there will be a 15-20 year lag between the ability distribution of enhanced young adults who have begun to contribute to research and economic growth and the distribution of newborns with the very latest enhancements. This looks like something that will boost growth substantially, but the big recursive effects look like they won’t have time to come into play until the second half of the century.
Genomics studies of intelligence may lead to the development of cognitive enhancement drugs on a faster timescale (although testing and approval still takes close to a decade, that’s less than growth to maturity). This is less certain (genetic variation affecting intelligence may mostly be rare idiosyncratic stuff, or only relevant in maturation and early brain development), but could be applied to a much larger population of existing adults, and children born without genetic enhancement.
Complete Genomics originally claimed they would sequence 10,000 genomes in 2010. In the article you linked to, they gave delays in establishing their main sequencing facility as a reason for promising only 5,000 for 2010. According to their wikipedia article, they only wound up sequencing 300 genomes in the third quarter of 2010. The hype in this area is usually one or two years premature (although that’s an amazingly short hype cycle nonetheless!), even with order of magnitude annual growth in sequencing.
Glad to see this is being weighed at SIAI. Some points:
If the enhancement is largely in China this seems unlikely. As an up and coming nation they are much less likely to take precaution when making moves to increase their own importance.
Things could be even worse if China adopted a state raised genius program, where they could mold the development of genius children from a young age (increasing their nationalism, for instance. Indeed, nationalism/obedience may be selected for genetically as well).
But is it feasible? Would the US ever allow full artificial gametes? Public opinion would be overwhelmingly against it. The only thing that would sway this, I think, would be seeing the full effect in China. Once Chinese children become overwhelmingly more intelligent than Western kids, the pressure to enhance would likely become much more poignant. No one wants to lose.
Such a scenario could leave the West 5 or even 10 years behind the curve. That’s a long time, especially considering this scenario is likely to play out 20 years from now when emerging technologies (AI, WBE, nano) may be on the verge of really taking of.
This is a reasonable strategy: focus on the scenarios where the confounding factor isn’t present. This is limiting though. Other strategies:
If enhanced scientists will be overwhelming Chinese in 20 years, then it makes sense to encourage a strong existential risk reduction atmosphere in China. It also makes sense to encourage as many risk savvy people as possible to get directly involved with Chinese research groups, especially those looking at AI and genetic engineering.
I suspect the forces at play here are potentially much more explosive. I think China knows exactly what it’s doing, and has strong political will power to get the job done. Once a genetic study has found a substantial amount of variation responsible for IQ, the pressure will be on to make use of that data. How much faster will research on artificial gametes go if the researchers are very well funded and don’t have any regulatory/ethical oversight?
(Note, also, that it’s possible that the brightest minds in the field will head to China once research begins to take off and it becomes clear the US won’t be taking part.)
I think the average IQ of politicians will not change much for the first few generations after the introduction of GE boosted babies.
Look at CA politicians column in the first table
Sure they are clever but they mostly seem to cluster in the 120 IQ range for functioning democracies, nearly regardless of the country mean. Also I need to crunch the data but my impression seems to be that some smaller countries tend to have usually smart leaders (Malta, Trinidad) compared to the mean of their countries. I can’t quite think of a major country that shows the same pattern.
However the ruling class of say Quatar or Saudi Arabia might be in for a leap frog.
I’m inclined to doubt that it’s the average of politicians that would make the most immediate impact, actually. Significantly stretching out the right-side tail ought to have practical consequences.
China is now basically run by technocrats—scientists and engineers. (A society run by such is part of many technologists’ fantasies, though I don’t see a massive brain drain to there.)
New Labour in the UK was run by some ridiculously intelligent and erudite people. I was a fan and even I would say they did quite a lot of demonstrating that intelligence means you can do much bigger stupid things.
There are really quite a lot of frighteningly intelligent politicians. I’m not sure just increasing politicians’ IQ will do much at all.
Two observations:
o #4 assumes a strong correlation between the thing being increased by intelligence-boosting biotech and the thing responsible for choosing to mitigate risks.
o #5 and #6 assume that voting, and similarly egalitarian power sources, will be the primary method whereby the intelligence-boosted minority will exert political power. (As opposed to, for example, making tons of money and buying elections outright, or constructing effective propaganda campaigns, or building giant lasers and threatening to write their names on the moon.)
I’d love to read the justifications for those presumptions if they are written down somewhere. (I do realize you can’t include all your reasoning in a blog comment.)
4 is based on the importance of simply noticing and understanding risks, and on data showing that increased intelligence and education are associated with more enthusiasm for public goods. 5 and 6 depend on the timescale, with the earliest enhancements being fairly limited.
我，作为个人来说，欢迎我们新的汉统治者。
Google translated this for me:
Groan.… :-P
See my 2007 article A Thousand Chinese Einsteins Every Year
What do people think are the odds that we’ll find specific genes that are markers for IQ?
(I assume that, just like with cancer, genes are “risk factors”, not on-off switches, and how they’re expressed depends somewhat on environment and behavior. But the equivalent of BRCA for IQ would still be really important.)
We already know a number of things about the genetic basis of intelligence—e.g. try searching for ASPM, NR2B, HAR1 and PYDN.
Current research suggests it’s hideously complex, and heritable characteristics with that level of complexity don’t tend to have simple markers.
then the Chinese may be disappointed.
I’m by no means an expert but I have studied a lot of the relevant fields to this topic in college. I know this thread is long dead but since it came up on the front page of google search of the title I feel the need to give my input. I was really into AI and studied computer science in college until I found out that Moore’s law is going to hit the atomic barrier before we have enough hardware by reasonable estimates to simulate a brain and there is no clear way to move forward (neither parallel programming nor quantum computing looks like it will save the dream without major breakthroughs that are by no means guaranteed to come within our lifetimes). The situation looks bad to anyone with a skeptical mind.
The same cannot be said of this genetically engineered intelligence idea. I switched to molecular biology and after two years of study it still seems just as possible especially given that in 2009 gene therapy was successfully used to introduce new genes into a genome to cure colorblindness in monkeys and a fatal brain disease in two human boys. Since you’re basically inserting a new gene into a random place in the genome, there’s always a chance you’ll insert into a cancer gene. That’s the bad news. The good news is that it works and can be done on adults. I would certainly volunteer for it if I was terminally ill. Also the cancer thing has a lot of conceivable solutions.
You mentioned recursive feedback. This is unlikely. While it is easy to find genes for intelligence that are already in the gene pool, inventing new ones is a whole ’nother ballgame. Not only would it require understanding the mind, it would also require us to invent new enzymatic proteins which, last time I checked, is still a computationally intractable problem. There is definitely a ceiling there that is likely unbreakable even by the first generation of superhuman geniuses. The good news is how potentially genius this first generation of engineered humans would be. It’s estimated that there at least hundreds of genes influencing intelligence each with multiple alleles. Notice how even 2^100 dwarfs the number of humans that have ever lived? You can safely bet that no optimal human in terms of intelligence has ever been born.
Even if you don’t buy this argument, citing that intelligence genes are simply additive and interact in such a complex manner that even China’s statisticians armed with supercomputers can’t find the optimal genome, you’d still have to concede that the current smartest humans are amazing beings that outperform the average or even the highly intelligent in remarkably profound ways.
In short, what you suggest can be done right now on adults in a highly sloppy and risky manner with some of the IQ genes we’ve already found and it’s efficacy would be highly variable, depending on the extent to which the gene’s influence is developmental.
Thanks for making this interesting and potentially important post! I got a lot out of Carl Shulman’s comment as well.
The ramifications are likely to be pretty negligible. Genetic engineering of the human genome is a difficult and slow process—and machine intelligence will likely zoom past human intelligence without genetic engineering of humans making very much difference.
To improve effective intelligence in the mean time, better options appear to include education—and preprocessing human sensory inputs using machines, and post-processing their motor outputs using machines—i.e. intelligence augmentation.
We haven’t done it yet, so it’s hard to say, but once we have hard data on alleles responsible for IQ, creating a genome with the desired alleles seems substantially easier than creating AI. Why do you believe otherwise?
I agree. The question isn’t will genetic engineering allow us to keep pace with machines. The question is how will genetic engineering affect the path taken toward the development of machine intelligence?
Combining existing alleles is relatively easy, yes. That happens by chance anyway, once in a while. However, with human development taking a couple of decades, that leads to a pretty slow build-test cycle. I figure we will probably have machine intelligence quite a while before the very first generation reaches maturity—in which case the significance will likely be extremely low.
Gene therapy is a bit different—though that has some problems of its own.
The benefit of engineering is that instead of a great combination happening once in a while it could very quickly happen 10 million times in just a few years.
The benefit of whole genome studies is that you don’t need to do multi-generational studies. You already know which allele combinations are gold. A single generation is enough.
What are your time estimates for AI? Certainly AI could happen in less than 20 years. It could happen tomorrow. But if there is a substantial portion of the probability mass sitting past 20 years then I think it behooves us to seriously take into account the possible effects of genetic engineering.
You can do some things with the existing variation. Not very much compared to what is possible with machines, though.
My time estimates have some probability mass past 2030:
http://alife.co.uk/essays/how_long_before_superintelligence/
I doubt there will be many adults who have been born with significantly engineered brains for a while past that—due to techincal problems, regulatory problems, etc. Anyway, I am not counselling ignoring the possible effects of genetic engineering—just estimating their probable impact on the run up to machine intelligence. As I said, too little, too late—I figure.
What’s important is how much compared to existing humans. The difference between a world class scientist and an average scientist is very important. That difference, multiplied by a thousand or a million individuals, can make a landslide difference in outcomes before AI has arrived.
It doesn’t seem regulatory problems will be an issue in China. Do you have a reason to believe otherwise? There will likely be technical problems, I agree, but do you really believe those problems are anyway near the level of difficulty of creating AI? Not to mention there is going to be much more money and effort spilled into genetic engineering research.
I think you must have a farther-out estimate of when machine intelligence is likely than me. There isn’t likely to be much time for germ-line genetic engineering of humans, I figure. So, on my timeline, it probably won’t make a landslide difference. It will probably make no difference at all.
China need more R+D at the moment. They won’t come into the picture for a while, regulations or no.
I don’t agree with you about how easy messing with the human genome is or about these projects’ respective funding, either.
We have made great progress augmenting human intelligence with machines. We have made negligible progress improving human intelligence by genetically engineering humans. The advantages of using machine augmentation will continue to grow—and will have become enormous by the time genetic engineering eventually gets around to producing some adult humans.
I don’t place AI so far out. I would be surprised if it took longer than 30 years. I think where we differ is for estimates on how hard genetic engineering is.
It’s true we haven’t had any success genetically engineering human intelligence. Nonetheless the tools we now have available are much more powerful. We simply couldn’t do Genome Wide Association Studies until recently, and still don’t quite have the ability to do them on huge scales. But it’s very likely that within just a couple of years million person studies will be happening.
It’s possible that such studies won’t turn up much. It’s also possible that huge numbers of alleles will be identified with statistically significant effects on IQ.
As Carl mentioned, constructing artificial gametes is also currently a limiting factor. But there is clearly a lot of work happening here. Craig Venter’s group, for instance, just constructed and rebooted the first artificial genome. That genome was much shorter than a human genome, of course, but it’s clear the fundamental idea is sound.
In my opinion these are the only two missing ingredients. That is: successful genome wide studies and the technology to make artificial gametes.
If these ingredients become available in the next couple of years, then it seems likely that large numbers of very smart adults will be around in about 25 years. Just in time to potentially make a huge difference in AI research.
I put a decent chunk of probability mass on AI occurring in less than 25 years, in which case none of this would matter. But I also put a decent chunk on 25+ years, and likewise a very good chunk on successful genetic engineering in the next 5 years (which, again, is where I think the core of our disagreement is).
Designing one new gene would be far beyond our capabilities. But choosing a set of good alleles based on genome-wide association studies would be trivial. Combining them together in a gamete would be difficult with current technology, but may become easy quickly. GM crops exist with IIRC at least 6 stable transgenes.
Terrance Tao, run and hide!
Terence Tao is not Chinese in that sense.