This post is a mirror of a tweet I made on why I expect inequality to increase as AI progresses.

ai is a force of extreme empowerment. currently it seems like many AI tools will be broadly available for anyone to use. many people refer to this as the ‘democratization’ of technology.

however! the outcomes that occur when you give everyone access to something are nowhere near equitable. much of the US has access to gyms and everyone to the outdoors, yet not only is health and strength not attained by all, but it’s very clear that there are long tails in these distributions, and that small sections of the population are several times stronger than average as a result.

this phenomenon repeats itself almost everywhere you look. alcohol is available for any adult to drink regardless of whether you have a genetic history of alcoholism or your religion prohibits you from trying it. megapacks of oreos are available to us all even if we have sworn off of sugar and even if our maladjusted gut microbiomes provide such strong signals to consume them that we literally cannot hold ourselves back.

chatgpt truly does offer improvement to millions of people, but I already see a select few thousand that are using it to make themselves 6-7 figures a year worth in productivity increases, which is likely >100 times the median outcome. the tails of these outcomes, the post-pmf startups with ten cracked engineers in an SF basement using LLMs as a fish uses water, takes this even further.

you don’t have to be singularitypilled to see that AI is likely to significantly increase inequality – and certainly not just that of wealth, but of ability as well. although this is one of the reasons proponents of foss AI commonly cite to further their cause, they generally do a poor job at modelling the real world outside of other programmers. foss ai is likely to help the startups keep up with the incumbents, but it is not going to help normal people keep up with the world of technology which has already outpaced them so thoroughly.

further strengthening this effect, we often use this drastic increase in our productivity to create products which then decrease the ability of others. youtube shorts and tiktok are great examples of this, where the tail-end of successful and capable individuals in society spend their high amounts of agency and intelligence on crafting products explicitly designed to keep the other less fortunate cohorts of society drooling at their screens for hours a day.

it certainly rings true to me that it’s much easier to raise the ceiling than the floor, and that is generally why technology is developed the way it is – simply because there are not other paths societies can reasonably coordinate upon for eventually making things available to everyone.

but the real world is and has always been dominated by power-law distributions, and escaping them is not an easy task. even if you have ai. especially if you have ai.

this isn’t necessarily a doomer take – it is possible to have worlds with large inequities which nonetheless are still beautiful, thriving, and have constantly decreasing amounts of suffering over time. i won’t pretend to predict which timeline we have ended up in ourselves, but i would certainly be long anything from ‘chaos’ to ‘inequality of agency’ if i could.

i’d like to have also included that wealth isn’t fixed – so an inequality in wealth does not mean others are less wealthy in absolute terms, as it’s easy to create more. this is the primary mechanism by which our society has improved and i think the floor will continue to raise.