interesting read. I agree with you but the problem we have today is that majority of people are ignorant about the use of technology that would allow you to be much better scholar than 99.999% of all scholars that lived before we had the Web or they are just too dumb to learn how to use it.
I also expect the side effect to be really bad or even fatal if they truly decide to go ahead with it.
In the end, I believe what is stopping us from advancing in a faster and better pace is not that we are not smart enough but other factors such as greed, corruption and wars.
I suspect that intelligence is a culturally dependent moving target. Most of our problems are social ones, and as such most of our ‘intelligence’ goes towards engaging with others. Even in the supposed objectivity of science, a lot of what we do is networking, keeping an eye on trends, and figuring out how to convey our results to other humans. As such, it is not clear to me that computers are an enhancement of intelligence as much as they are a way to alter the definition more.
Chess is a poor example for the superiority of computers over brains because it’s a very simple game from a computational perspective. Chess AI is very different from the thought processes of human players (turn prediction vs pattern recognition). It just so happens that for chess specifically, the AI approach works better.
Chess programs were historically used by AI research as an example for “being smarter than humans,” but that was first a misconception and then a propaganda trick. Computers are still fairly terrible at things that brains have evolved to do well, such as pattern recognition itself (where it can’t be substituted by something else) or navigating complex 3D environments.
However, all this actually reinforces your main point: if you want to enhance human intelligence it’s probably better to add complementary features (exact digital memory, high computational performance), rather than trying to improve on what the brain already does very well.
Should we try for higher intelligence? Sure. Why not? On the other hand, what does that mean?
Are we trying to raise the top end? In what way? Sure, we might come up with a list of attributes, but which matter? Which are most valuable? Do we know how to judge? Do we know how to design a better brain? Seems likely a lot of trial and error is needed.
Are we trying to raise the middle? Or eliminate the low end? Ethics aside, this might be easier. Very tricky ground.
Guess I am a bit more skeptical about the path to direct non-biologically enhanced intelligience. AI is still a dream. Direct mind to machine interfaces seem tricky. Decades away at least, probably several. The human brain does not have those sort of I/O ports. Might be an easier target for genetic engineering. Could the human brain be re-engineered to have structures suitable for machine interface? Could existing brains be retrofitted through gene therapy?
We have a long way to go. Easier to eliminate the low end, and boost the middle. Ethics aside.
Dominic Amannsays:
I think this view of intelligence is both incorrect and immaterial. One of the most important aspects of intelligence is our ability to make decisions.
This is very involved, and has little to do with how much we know. It has everything to do with how we weigh that knowledge, determine what is relevant to the case in point, and our personal history with similar decisions. Large databases don’t help very much except as a tool to providing our brains with a lot of information.
Some very recent research indicates that Glial cells have a direct contribution to intelligence (as evidenced by better decision making). Human glial cells directly injected into new-born mice have had statistically significant effect in increasing intelligence.
This is very extraordinary both from the perspective of trans-species transplant, and also from the fact that hithertoo, glial cells were not known to be a factor in intelligence.
Despite the ability to calculate 10k digits of pi, this does not seem to have enabled a broad-based revolution among the masses and lifted up millions of Archimedes. So just like always, that turns out to not be a significant limiting factor.
> What if, after a couple of generations (say 50 years), China could rise the average IQ of its population by 5 points using genetic engineering? How much would that matter? Probably not much at all. In 2060, the fraction of our intelligence that will depends on our brains will be tiny, assuming biological brains are even still relevant.
A singularity could simply fail to happen. Software could bog down under its own weight. The AI Winter showed that while Moore’s law marched own, it didn’t have to be accompanied by great progress in AI related software or ideas.
In which case 5 points could matter a fair bit. An exercise: 5 points is 1/3 a standard deviation. If China is at 100 now and shifts to 105 in the future, how many times bigger is the fraction of the population with IQ >150?
Paulsays:
There’s an interesting question implicit in your thoughts: Does technology add to our cognitive output or multiply it? Does Einstein with the internet further outshine his average peer, or does everybody approach him?
> This is very involved, and has little to do with how much we know. It has everything to do with how we weigh that knowledge, determine what is relevant to the case in point, and our personal history with similar decisions. Large databases don’t help very much except as a tool to providing our brains with a lot of information.
I think this is very wrong. One of the constant trends in studying human expertise and in AI is that to a large degree, making good decisions is due to a huge amount of data. You should check out the _Cambridge Handbook_ on the topic, but the basic point is: expertise develops slowly and gradually, scales with the number of examples processed, the faster feedback is the more skills develop (as expected, since supervised learning > unsupervised learning), case studies like chess masters show that they depend on thousands of games and fragments stored in long-term memory as a key part of how they analyze any position (rather than, say, larger working memory or deep evaluation of many plys), expert systems can reach near-human performance levels in domains by absorbing on the order of (just) thousands of items of expert-elicited data, and machine learning has observed that even simple algorithms can continue to increase performance if given huge amounts of raw data (most famously in Norvig’s paper ‘The unreasonable effectiveness of data’). Then you have the constant divide in the heuristics & biases psychological literature between ‘system I’ and ‘system II’ thought, where most thought and tasks are done via fast frugal system I pattern recognition (but how do you ‘recognize’ without knowledge?), and not slow serial system II thought where logic and reasoning are employed.
Anonymoussays:
This post is based on a totally biased observation. If you are not a racist, please first find out the facts of China using Eugenic. Yes. There is birth-control, the so called one-child policy. However, this policy results in more spoiled children than smarter children. The new government is considering to stop this policy.
At no point do I state that China is pursuing eugenics. I even link, in my first paragraph, to a post that says that this is not true.
What I can do is play thought experiments… what if China did it?
Anyhow, if you have a better reference proving that China is not pursuing eugenics, beyond the one I offer, please share it with us.
Doreensays:
It looks to me as though you are confusing intelligence with information. Simply being able to access vast amounts of information doesn’t actually improve your intelligence, just your education. There are entirely too many highly educated idiots to make that a worthwhile pursuit.
I am all for information, but the matrix within which the human being must function isn’t all about information, but the moral and long term affects of using that information in a beneficial manner.
Just one thing really, the Flynn effect(http://en.wikipedia.org/wiki/Flynn_effect) never came up in the discussion till now. Funnily enough the raw intelligence of the world has been increasing since the time the IQ test was invented. Effectively China, IF we gave in to the idea that they were artificially boosting pattern recognition abilities(who knows the true definition of intelligence anyway?), then they would have to also get ahead of the curve, or maybe they would just boost the Flynn effect itself… Who knows?
I was implicitly referring to the Flynn effect in the sense that by all accounts, we are getting smarter (measurably so) with each passing generation… and that’s not due to genetics!
Also, we don’t want to trigger the Eugenics Wars. Very important.
Matsays:
I agree with Doreen on this matter.
When it comes to the Flynn effect – wiki gives some ideas of what could be causing it. The fact that information are more available with technology can in great deal contribute to what and how we think, but saying that it is such a big factor in the intelligence means neglecting the nature in the nature-vs-nurture debate.
Indeed, there are big correlations between intelligence of children in studies involving the identical siblings raised in different families (they have the same genes). Something that should be taken into account.
When it comes to genes and intelligence – people with similar intelligence tend to stick together. A good question would be whether there is natural selection guided by our own choice of a partner. So, if we have a choice of a partner, do we peek the one who is a bit “simple”, the one that is “on the same level”, or the one that “amazes us”?.
Daniel Seboldsays:
I am disappointed in this discussion: is it possible to genetically produce a being intelligent enough to solve the problem of
how to manipulate the space/time fabric so that we may leave this planet and exploit other planets that are as of yet untainted by pollution?
interesting read. I agree with you but the problem we have today is that majority of people are ignorant about the use of technology that would allow you to be much better scholar than 99.999% of all scholars that lived before we had the Web or they are just too dumb to learn how to use it.
I also expect the side effect to be really bad or even fatal if they truly decide to go ahead with it.
In the end, I believe what is stopping us from advancing in a faster and better pace is not that we are not smart enough but other factors such as greed, corruption and wars.
good blog entry, keep them coming 😉
I suspect that intelligence is a culturally dependent moving target. Most of our problems are social ones, and as such most of our ‘intelligence’ goes towards engaging with others. Even in the supposed objectivity of science, a lot of what we do is networking, keeping an eye on trends, and figuring out how to convey our results to other humans. As such, it is not clear to me that computers are an enhancement of intelligence as much as they are a way to alter the definition more.
Chess is a poor example for the superiority of computers over brains because it’s a very simple game from a computational perspective. Chess AI is very different from the thought processes of human players (turn prediction vs pattern recognition). It just so happens that for chess specifically, the AI approach works better.
Chess programs were historically used by AI research as an example for “being smarter than humans,” but that was first a misconception and then a propaganda trick. Computers are still fairly terrible at things that brains have evolved to do well, such as pattern recognition itself (where it can’t be substituted by something else) or navigating complex 3D environments.
However, all this actually reinforces your main point: if you want to enhance human intelligence it’s probably better to add complementary features (exact digital memory, high computational performance), rather than trying to improve on what the brain already does very well.
Should we try for higher intelligence? Sure. Why not? On the other hand, what does that mean?
Are we trying to raise the top end? In what way? Sure, we might come up with a list of attributes, but which matter? Which are most valuable? Do we know how to judge? Do we know how to design a better brain? Seems likely a lot of trial and error is needed.
Are we trying to raise the middle? Or eliminate the low end? Ethics aside, this might be easier. Very tricky ground.
Guess I am a bit more skeptical about the path to direct non-biologically enhanced intelligience. AI is still a dream. Direct mind to machine interfaces seem tricky. Decades away at least, probably several. The human brain does not have those sort of I/O ports. Might be an easier target for genetic engineering. Could the human brain be re-engineered to have structures suitable for machine interface? Could existing brains be retrofitted through gene therapy?
We have a long way to go. Easier to eliminate the low end, and boost the middle. Ethics aside.
I think this view of intelligence is both incorrect and immaterial. One of the most important aspects of intelligence is our ability to make decisions.
This is very involved, and has little to do with how much we know. It has everything to do with how we weigh that knowledge, determine what is relevant to the case in point, and our personal history with similar decisions. Large databases don’t help very much except as a tool to providing our brains with a lot of information.
Some very recent research indicates that Glial cells have a direct contribution to intelligence (as evidenced by better decision making). Human glial cells directly injected into new-born mice have had statistically significant effect in increasing intelligence.
This is very extraordinary both from the perspective of trans-species transplant, and also from the fact that hithertoo, glial cells were not known to be a factor in intelligence.
Despite the ability to calculate 10k digits of pi, this does not seem to have enabled a broad-based revolution among the masses and lifted up millions of Archimedes. So just like always, that turns out to not be a significant limiting factor.
> What if, after a couple of generations (say 50 years), China could rise the average IQ of its population by 5 points using genetic engineering? How much would that matter? Probably not much at all. In 2060, the fraction of our intelligence that will depends on our brains will be tiny, assuming biological brains are even still relevant.
A singularity could simply fail to happen. Software could bog down under its own weight. The AI Winter showed that while Moore’s law marched own, it didn’t have to be accompanied by great progress in AI related software or ideas.
In which case 5 points could matter a fair bit. An exercise: 5 points is 1/3 a standard deviation. If China is at 100 now and shifts to 105 in the future, how many times bigger is the fraction of the population with IQ >150?
There’s an interesting question implicit in your thoughts: Does technology add to our cognitive output or multiply it? Does Einstein with the internet further outshine his average peer, or does everybody approach him?
> This is very involved, and has little to do with how much we know. It has everything to do with how we weigh that knowledge, determine what is relevant to the case in point, and our personal history with similar decisions. Large databases don’t help very much except as a tool to providing our brains with a lot of information.
I think this is very wrong. One of the constant trends in studying human expertise and in AI is that to a large degree, making good decisions is due to a huge amount of data. You should check out the _Cambridge Handbook_ on the topic, but the basic point is: expertise develops slowly and gradually, scales with the number of examples processed, the faster feedback is the more skills develop (as expected, since supervised learning > unsupervised learning), case studies like chess masters show that they depend on thousands of games and fragments stored in long-term memory as a key part of how they analyze any position (rather than, say, larger working memory or deep evaluation of many plys), expert systems can reach near-human performance levels in domains by absorbing on the order of (just) thousands of items of expert-elicited data, and machine learning has observed that even simple algorithms can continue to increase performance if given huge amounts of raw data (most famously in Norvig’s paper ‘The unreasonable effectiveness of data’). Then you have the constant divide in the heuristics & biases psychological literature between ‘system I’ and ‘system II’ thought, where most thought and tasks are done via fast frugal system I pattern recognition (but how do you ‘recognize’ without knowledge?), and not slow serial system II thought where logic and reasoning are employed.
This post is based on a totally biased observation. If you are not a racist, please first find out the facts of China using Eugenic. Yes. There is birth-control, the so called one-child policy. However, this policy results in more spoiled children than smarter children. The new government is considering to stop this policy.
@Anonymous
At no point do I state that China is pursuing eugenics. I even link, in my first paragraph, to a post that says that this is not true.
What I can do is play thought experiments… what if China did it?
Anyhow, if you have a better reference proving that China is not pursuing eugenics, beyond the one I offer, please share it with us.
It looks to me as though you are confusing intelligence with information. Simply being able to access vast amounts of information doesn’t actually improve your intelligence, just your education. There are entirely too many highly educated idiots to make that a worthwhile pursuit.
I am all for information, but the matrix within which the human being must function isn’t all about information, but the moral and long term affects of using that information in a beneficial manner.
Just one thing really, the Flynn effect(http://en.wikipedia.org/wiki/Flynn_effect) never came up in the discussion till now. Funnily enough the raw intelligence of the world has been increasing since the time the IQ test was invented. Effectively China, IF we gave in to the idea that they were artificially boosting pattern recognition abilities(who knows the true definition of intelligence anyway?), then they would have to also get ahead of the curve, or maybe they would just boost the Flynn effect itself… Who knows?
Gattaca all over again?
Gattaca (Movie)
@Anirudh
I was implicitly referring to the Flynn effect in the sense that by all accounts, we are getting smarter (measurably so) with each passing generation… and that’s not due to genetics!
Also, we don’t want to trigger the Eugenics Wars. Very important.
I agree with Doreen on this matter.
When it comes to the Flynn effect – wiki gives some ideas of what could be causing it. The fact that information are more available with technology can in great deal contribute to what and how we think, but saying that it is such a big factor in the intelligence means neglecting the nature in the nature-vs-nurture debate.
Indeed, there are big correlations between intelligence of children in studies involving the identical siblings raised in different families (they have the same genes). Something that should be taken into account.
When it comes to genes and intelligence – people with similar intelligence tend to stick together. A good question would be whether there is natural selection guided by our own choice of a partner. So, if we have a choice of a partner, do we peek the one who is a bit “simple”, the one that is “on the same level”, or the one that “amazes us”?.
I am disappointed in this discussion: is it possible to genetically produce a being intelligent enough to solve the problem of
how to manipulate the space/time fabric so that we may leave this planet and exploit other planets that are as of yet untainted by pollution?