My suspicion is that the truth is somewhere in between and the purple curve shows one possible guess (and it's no more than that). The general idea is that the organizing principles of intelligent thought will prove harder to crack than just producing the raw computing power, but on the other hand, several orders of magnitude of more Moore's law will be more dramatically transformative than the trend in the employment/population ratio to date would indicate. Additionally, of course, we will have the ongoing effects of China, India, etc, bringing a larger and larger fraction of their populations into the modern era, and this will have a depressing effect on US employment for several more decades (either we let our corporations continue to outsource to those countries, or we'll find our corporations put out of business by lower cost competitors, or we'll impose major trade barriers and invite corresponding retaliation, and any which way, it will tend to put pressure on US employment).
Now it seems clear that this trend is not going to run uninterrupted to zero in a democratic society. Other things being equal, machine intelligence has overwhelming economic advantages - the possibility of being readily specialized to task, the willingness to work night and day without demanding resource-intensive houses and cars in compensation, ability to be replicated exactly, etc, etc. However, people have an overwhelming political advantage - they can vote, and machines can't.
I expect in the short term (next decade or two), what we will see is an intensification of the trends of recent decades:
- Life will get better for most people in the upper class and upper half of the middle class. If you have a lot of assets, or a creative and intellectually challenging job, the technology will enable you and make your life better and provide you with cool toys and entertainment options. There will be increasing demand for your skills, and employment will not be a problem.
- In the lower half of the income/class spectrum, life will get steadily grimmer. There will be fewer jobs suited to your skills and abilities, and competition for the ones that there are will be intense. More and more people will fall out of the system and have to find alternative ways to survive (welfare, crime, institutionalization, disability, relying on family/friends, etc, etc).
- Competition for educational opportunities will become increasingly high-stakes, as education becomes more and more critical to have a chance at fitting into the system in a reasonably high-status way.
- Overall, psychological stress in the society will gradually increase.
- The basic assumptions of our culture - that status is associated with having income/assets, that hard work is good, that people living off some form of welfare are second-class citizens, that innovation is good, will not be widely challenged initially.
However, as times goes on, and as computer intelligence continues to get better and better and displace more and more people from the workforce, it seems that there will be fertile ground for political movements to address the situation. I see several issues that will build over time:
- The more-or-less unemployed underclass will get larger, is unlikely to be very content, and is potentially available to be recruited into political movements for change.
- Members of the elite will increasingly recognize the nature of the process, and realize that even though they are doing great, their children's future is at risk.
I assume that, with the exception of a few fairly bizarre computer scientists, most people do not wish to be functionally replaced with machines. I also assume that modifying people themselves to make them smarter will prove much less tractable than just making better machines (it involves brain surgery, long clinical trials to establish safety/efficacy, etc, all of which will make the product lifecycle turnover much slower than new computer chips or software versions, and at the end of the day, the pure machine is always going to be willing to work for cheaper, since no-one is going to build into it the desire for a big house in the suburbs and a fast car).
So people are, ultimately, going to want to preserve a human-dominated society, but clearly that is going to mean that, one way or another, at some point, we will have to agree to stop developing ever smarter computers.