Cooper Beckel
2023
-----------------------
There will always be an indivisible romance between Technology, Art, and War. We can admire one example, as ancient heating technologies developed they allowed people to fire ceramic pottery – an art form blossomed over the eras – and in the 21st century ceramic matrix composites adorn the armor of F-35 fighters. Lithography, the technical printmaking process, has not only carved images through generations, but it has also yielded the advanced semiconductor. It was into this discourse that I, at the time a fresh art school graduate, walked over five and a half years ago.
In 1939 General Motors sponsored ‘The World of Tomorrow’ at the World’s Fair, Deep Blue beat Kasparov in 1996, and in 2004 DARPA held its Grand Challenge; breadcrumbs in the tale of autonomous vehicles. When I joined that story in 2017 it was still in a chapter of academia-light, startup vibes, and move-fast feelings. The visions were clear and the money rolled in. The challenges were significant and exciting, and systems were built in days that urged cars to drive themselves on the roads of golden San Francisco. A scrappy student of the arts like myself was tickled to be involved; it was a time for ideas and they came from every direction. And the technology wizards worked their wonders, and the system evolved, and soon cars were driving with literally nobody behind the wheel on public roads.
When that milestone was reached and I took my first fully driverless ride, I was invigorated to feel the thoughts of thousands of people who I had worked with flowing through the silicon and wires and wheels. The car wasn’t driving itself, people were, even if their bodies were not in the vehicle with me. And this was a striking revelation: that in the era of Machine Learning the thoughts of hundreds of people can be synthesized through reward functions and fiber optics to execute decisions in the everyday. And this was a moment where my own thoughts began to evolve.
As ChatGPT and Bing make headlines, it should be clear that Machine Learning is entering the mainstream. As fleets of autonomous vehicles navigate American streets, it should be noted that some of these systems are becoming safety critical. Indeed, this is a networked America, a world with cascading systems of systems where thoughts from the past are digitally retained and creating the future. It is with this in mind that I recently became fascinated by the aviation industry. That is an industry which lives and dies by fault-attribution; a global industry where language matters and in which systems transparency is the minimum. Read a crash report from the French BEA and you will begin to realize why commercial aviation is successful. Theirs is a worldview; if a failure happens once it should never happen again. And it is in this paradigm that I believe ‘move fast and break things’ dies for many of the companies who have taken on stewardship of advanced and evolving Artificial Intelligences.
This then is an era where language matters, even when social media erodes our lexicon, and where “attention is all you need”. As the ‘data guy’ by day I begin to demand a level of specificity when discussing layered AI subsystems; I need phraseology to be as specific to me as it is for the pilot of a 737. A pilot does not need to see the entire airport, but they must have a correct mental model and correct words must be spoken to them by air traffic control. Replace the pilot with their simulacra and the same is true, a safety critical system must be trained to think critically with ideas that are specific and terms robust; and we must have the vocabulary to describe problems when things go wrong.
The technology industry must be able to learn from its aviation counterparts in this regard. Equally the technology industry should come to see artists not as makers of images, but as shepherds of perception, or aviators in artificial intelligence, whose models of reality might best reflect human values.