Construction and AI, a bitter lesson that many will learn

One of the nice things about changing this blog to discuss construction over data centers is the audience is much broader. An example is the application of Artificial Intelligence (AI) to construction. If you Google “construction AI” you will have a wide range of companies and articles on the topic. If you talk to any big company in the construction ecosystem you will get a confirmation that they are working on the use of AI in construction. So all is good. Construction will soon have AI systems making construction better. Having worked on the construction problem for over 10 years I am more skeptical of what will be done. Why? One example is the very nature of how AI systems get developed. 

One of my super smart old friends shared this YouTube video.  How smart is he? He worked at Apple, Adobe, Microsoft, Google, Facebook, and startups. How old a friend? Have known him for 30 years.

The above video refers to this paper.

I find it interesting that so many times well written papers have the most interesting part at the end, but so many readers don’t make it to the end. Or by the time they get to the end they aren’t looking for the big points. At the end of this paper is the observation that the fundamental approach to AI is flawed. 

 “The second general point to be learned from the bitter lesson is that the actual contents of minds are tremendously, irredeemably complex; we should stop trying to find simple ways to think about the contents of minds”

This observation is based on studying 70 years of AI efforts and figuring out what works and what doe not. The YouTube video provides some nice graphics to illustrate this point with examples.

Now the biggest problem with doing this as this paper suggests is it says do not build your AI system with knowledge from your expert users. It feels good and is satisfying in the short term. Huh. What other things in life are bad for you that feel good and provide short term satisfaction, but plateaus and inhibits further progress. 

So what do you do? Take a contrarian approach. One based on search and learning that scales with computational capabilites. Huh what businesses have made a success based on search and learning? Google. Facebook. Amazon retail. They are all disrupting businesses with AI.

”breakthrough progress eventually arrives by an opposing approach based on scaling computation by search and learning. The eventual success is tinged with bitterness, and often incompletely digested, because it is success over a favored, human-centric approach.” 

Well off to Google Cloud Next 19 in SF. Ready to see what they are selling to win the battle against AWS and Azure.