What 3D Printing Teaches Us about the AI Hype Cycle

April 09, 2024

When I was in law school I was heavily involved in the Connecticut Center for Entrepreneurship and Innovation. Through UCONN’s Innovation Accelerator program I worked with companies looking to build market plans for their product ideas. The goal was to get an investment from an angel investor. On the other end of the spectrum I was a student lawyer in the Intellectual Property clinic, helping secure trademarks and patents for start-ups. I know entrepreneurs, their personality, and the hype cycle that fuels job creation but blinds folks to challenges in creating products and businesses.

By The Numbers

In 2013 the hype around 3D printing reached a big peak:

You can see the hype in May followed by a quick taper. That level of interest did not arise again until 2018. I remember those days in 2013. 3D printing was going to save our economy, we were going to start manufacturing in the US again, and maybe we would all own our home version of a replicator from Star Trek.

The Reality of 3D Printing Today

Over the weekend, I picked up my first 3D printed item from the Boston Public Library’s Business and Innovation Center. Made using a little bit of an older Makerbot I discovered what I imagine many other 3D printer users discovered years ago. Instead of an instant pristine product I had something that resembled the object I wanted but had to rip out the extra plastic that supported the build. Instead of a smooth surface I could see the strands of 3D print filament. Technical challenges with this older printer meant the flat portion of the object curved. It was a neat idea and the object worked, but it required some love before it was usable. All things being equal, I’d prefer to have just purchased a mass manufactured version. It just does not exist yet.

With hindsight it is easy to see what happened with 3D printing. The shine wore off, and in 2016 WIRED penned a post-mortem on Makerbot’s fall. Early product-market fit with hobbyists, enthusiasts, and power users did not translate into mainstream adoption. 3D printing never died, and quality of the printers and goods continues to improve, along with interest in it, but it never revolutionized society the way the iPhone did.

Progress and Promises in AI

History does not repeat itself, but it does rhyme. In the case of AI specialized applications like tracking hands to control a VR headset, reading the road to drive a car, or generating fake audio or video of real people work great now. Unsatisfied with these advances prognosticators claim AI will take your jobs and destroy humanity.ChatGPT claims to pass the bar exam. I have thus taken it seriously and thrown batteries of tasks and tests at ChatGPT only to watch it belly flop.

LLM Disappointments

Where does ChatGPT fail? Creating a configuration file for fish shell based off documentation from a command line tool did not yield a usable product from any existing commercial AI chat bot. They all refused to make me an iCal formatted training schedule for the BAA 10K. LLMs are a great leap forward in relevancy and research tooling, with caveats on hallucinations, but are not great at reasoning and applying abstract patterns to new problem spaces. Increasing the size of ChatGPT or other LLM training sets is merely putting more smoke and mirrors in front of the illusion of intelligence.

A Critically Honest Approach to Hype

What should a citizen in a voting booth or an executive in a large organization take away from this?

  • Avoid the temptation to pick a side in the hype cycle, critics and boosters tend to exaggerate and embellish their claims.
  • Explore without over-investing: Elon Musk bet on automating car building with robots and expensively learned it was more reliable to use people. The technology wasn’t there yet.
  • Ground truth claims that seem too good to be true.

Want to get posts like this in your email?

This work by Matt Zagaja is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.