Epistemic status This post is a “plausible conjecture” alternative to Eliezer’s take on clickbait. My arrogant epistimology thinks it’s substantially closer to the truth than Eliezer’s version.
In the 1950′s there were a similar number (order of magnitude) of humans, and they did a similar amount of socializing as today. This would suggest that a similar quantity of memetic optimization was going on then. Any universally popular meme would have risen to popularity. The difference between then and today was that then people usually only communicated with people geographically local to them. With a small number of newspapers being widely read. Whereas today, people communicate with like minded individuals around the world regularly.
Suppose that different humans have different selection criteria when deciding to share a meme. In 1950, a meme had to appeal to a broad section of society to be spread. Not 100%, more like 25%, the small number of different newspapers, and small numbers of local people to communicate with still let memes specialize along a few socioeconomic lines. Eg Catholic memes vs Protestant memes in Northern Ireland, or Liberal vs Conservative memes in many places. Each newspaper had a side, and many of your neighbors were on your side.
Nowadays, memes can specialize to focus onto tiny subsets of the population. Given that a tiny fraction of the population think in a particular and unusual way, they can gather together online, and share memes optimized exclusively to them. This produces many internet subcultures. Within each subculture, the selection pressure isn’t that strong, there aren’t that many model railway buffs or code golfers or … But the memes are optimized to a particular way of thinking, not the combination of several.
In some circumstances, Schelling points create an averaging effect in the 1950 case. Suppose an issue, say cats vs dogs, is sufficiently minor that newspapers are not split into a pro cat paper, and a pro dog paper. Then if a newspaper says anything excessively pro cat, the dog fans will call them out on it, and possibly stop reading the paper. Likewise if the paper is pro dog. So the paper finds a Schelling point, where neither side are upset enough to cause a real fuss. Note that this process is only truth seeking to the extent that cat supporters will let valid pro dog arguments slide and call out invalid ones.
The main effect of filter bubbles is increasing the variance in the meme pool. The biologists can stop arguing with creationists, and get down to sorting out the details of kin selection or whatever. The creationists can stop having to pedal creationism to the unconvinced and can get together to work out the difference between micro-evolution and macro-evolution.
Optimizing for a combination of A and B can produce more of both than choosing which to optimize for at random. If we have memes that are (A=10, B=0) and (A=9, B=9) and (A=0, B=10), then the middle one is likely to spread in a world without filter bubbles, but the extremes could spread in bubbles optimizing only A and B respectively. If A=sanity, then the average sanity could fall due to the optimization for sanity being focused into one place, and diminishing marginal returns on sanity for optimization.
Other effects on the quality of discourse could include stupid people having an easier time voicing their opinions. Eliezer says that the quality of internet discussion has degraded from 2002 to 2017. (I wasn’t old enough to use the internet in 2002, so can’t confirm or deny this.) According to these sources, under 10% of the world was online in 2002, facebook and twitter hadn’t started yet. In short, getting on the internet required more technical competence, the hardware was more expensive, and there was less to do there. The typical internet user was moderately well educated and smart. The typical newspaper journalist was also moderately well educated. Any reduction in quality would seem to be from uninformed people being finally able to tell the world why the earth is flat.
If you think that the world needs a few highly sane people, not many slightly sane people, then an aggregation into a few groups of sanity is beneficial.
Under the hypothesis that clickbait is destroying intelligence, the existance of less wrong, and places like it, is highly surprising, under a segregation hypothesis, its expected that the most rational people clump together.
Clickbait might not be destroying our general Intelligence
Epistemic status This post is a “plausible conjecture” alternative to Eliezer’s take on clickbait. My arrogant epistimology thinks it’s substantially closer to the truth than Eliezer’s version.
https://www.lesswrong.com/posts/YicoiQurNBxSp7a65/is-clickbait-destroying-our-general-intelligence
In the 1950′s there were a similar number (order of magnitude) of humans, and they did a similar amount of socializing as today. This would suggest that a similar quantity of memetic optimization was going on then. Any universally popular meme would have risen to popularity. The difference between then and today was that then people usually only communicated with people geographically local to them. With a small number of newspapers being widely read. Whereas today, people communicate with like minded individuals around the world regularly.
Suppose that different humans have different selection criteria when deciding to share a meme. In 1950, a meme had to appeal to a broad section of society to be spread. Not 100%, more like 25%, the small number of different newspapers, and small numbers of local people to communicate with still let memes specialize along a few socioeconomic lines. Eg Catholic memes vs Protestant memes in Northern Ireland, or Liberal vs Conservative memes in many places. Each newspaper had a side, and many of your neighbors were on your side.
Nowadays, memes can specialize to focus onto tiny subsets of the population. Given that a tiny fraction of the population think in a particular and unusual way, they can gather together online, and share memes optimized exclusively to them. This produces many internet subcultures. Within each subculture, the selection pressure isn’t that strong, there aren’t that many model railway buffs or code golfers or … But the memes are optimized to a particular way of thinking, not the combination of several.
In some circumstances, Schelling points create an averaging effect in the 1950 case. Suppose an issue, say cats vs dogs, is sufficiently minor that newspapers are not split into a pro cat paper, and a pro dog paper. Then if a newspaper says anything excessively pro cat, the dog fans will call them out on it, and possibly stop reading the paper. Likewise if the paper is pro dog. So the paper finds a Schelling point, where neither side are upset enough to cause a real fuss. Note that this process is only truth seeking to the extent that cat supporters will let valid pro dog arguments slide and call out invalid ones.
The main effect of filter bubbles is increasing the variance in the meme pool. The biologists can stop arguing with creationists, and get down to sorting out the details of kin selection or whatever. The creationists can stop having to pedal creationism to the unconvinced and can get together to work out the difference between micro-evolution and macro-evolution.
Optimizing for a combination of A and B can produce more of both than choosing which to optimize for at random. If we have memes that are (A=10, B=0) and (A=9, B=9) and (A=0, B=10), then the middle one is likely to spread in a world without filter bubbles, but the extremes could spread in bubbles optimizing only A and B respectively. If A=sanity, then the average sanity could fall due to the optimization for sanity being focused into one place, and diminishing marginal returns on sanity for optimization.
Other effects on the quality of discourse could include stupid people having an easier time voicing their opinions. Eliezer says that the quality of internet discussion has degraded from 2002 to 2017. (I wasn’t old enough to use the internet in 2002, so can’t confirm or deny this.) According to these sources, under 10% of the world was online in 2002, facebook and twitter hadn’t started yet. In short, getting on the internet required more technical competence, the hardware was more expensive, and there was less to do there. The typical internet user was moderately well educated and smart. The typical newspaper journalist was also moderately well educated. Any reduction in quality would seem to be from uninformed people being finally able to tell the world why the earth is flat.
If you think that the world needs a few highly sane people, not many slightly sane people, then an aggregation into a few groups of sanity is beneficial.
Under the hypothesis that clickbait is destroying intelligence, the existance of less wrong, and places like it, is highly surprising, under a segregation hypothesis, its expected that the most rational people clump together.
https://www.internetworldstats.com/emarketing.htm
https://www.inquisitr.com/830664/the-history-of-social-media-when-did-it-really-begin-you-may-be-surprised-infographic/