The AI Threat Isn’t Skynet – It’s the End of the Middle Class

I'm working on a machine learning product for conversion optimization, https://improve.ai.

Despite my initial fears, I'm noticing that, like any tool, it simply allows the user to do more with less. My product isn't going to replace conversion rate optimization consultants - it's just another tool they use to provide more value to their clients.

If the barrier to using machine learning were high, then I would be concerned that the benefits would only go to few. But since very powerful APIs will be increasingly cheap and democratized, the barriers will be quite low to reap the benefits of these technologies.

We often get caught up in the technical minutiae of machine learning, but you already don't need to be a deep learning expert to take advantage of, for example, cheap speech recognition APIs. This lowered barrier to entry will only accelerate.

But, then again, I'm an optimist. :-)

  > In English-speaking countries, a blue-collar worker is
  > a working class person who performs non-agricultural
  > manual labour. Blue-collar work may involve skilled
  > or unskilled manufacturing, mining, sanitation, 
  > custodial work, oil field work, construction,
  > mechanical maintenance, warehousing, firefighting,
  > technical installation and many other types of
  > physical work.
  >
  > In contrast, the white-collar worker typically
  > performs work in an office environment and may
  > involve sitting at a computer or desk.

Computer programming is currently white-collar work. At what point are we going to go outside and begin to do it with our hands?

Also, the nature of programming is the ability to automate away white-collar work. We're probably going to need new categories, once many of those are gone. We might end up with engineers, a service-worker class and a managerial/executive class.

I think there's more to conversation rate optimisation than randomly testing variations of a greeting. The price testing you mentioned looks interesting, but a good CRO consultant will do research and figure out the objections manually. Then, instead of having to test an infinite amount of variations with millions of visitors to reach statistical significance, they can test one or two. I call what your tool seems to do "throwing darts in the dark". And if I've completely misunderstood your product, perhaps you've got some optimising to do ;)

Throughout the 20th century new technology has been steadily increasing the value of human labour, but some forms of human labour have increased more than others, and this imbalance has caused a market distortions.

Technology has done a great deal to make railroad engineers more valuable for instance, because one guy could suddenly move hundreds of tons of goods. Used to be it took an army to do that.

On the other hand we have plumbers, and the amount of work a plumber can do has changed little in a century. There is no plumber who can to the work of a hundred plumbers from 100 years ago.

This causes a market distortion is what is hurting so much of what America calls the middle class. If we deploy AI well, it may correct some of these market distortions. A plumber with a team of robots and an AI assistant will all of the sudden be able to do enough work to justify the cost of his labour. Maybe. If one were to entertain the notion that there are certain jobs we haven't automated enough, which hurts their ability to compete with the value we get from low cost goods and services.

Each time this conversation comes up I notice people often like to suggest that history is just repeating and we have seen this before. I think its important to recognize patterns in cycles but its also important to be open to the possibility we've only seen local cycles and larger global cycles are still revealing themselves. In other words is AI really analogous to shifts we've seen before?