The AI tsunami that is sweeping over us has some of us riding the wave and others thinking they will drown.
On one extreme, there are fears that AI will wipe out the human race as did Skynet in Terminator. Elon Musk, influenced by science fiction, has called AI a risk to the public and recently has sued OpenAI, makers of ChatGPT, for deviating from its original mission of using AI for the good of humanity. More alarm bells rang for AI used for nefarious or mischievous purposes, like deepfake photos, videos and, perhaps the most sinister of all, voices, such as kidnappers faking their victims’ voices in calls to desperate parents.
“AI will take my job” is a fear felt by anyone who holds their job in low esteem — a job that require little thought but much repetition. Examples abound. Uber drivers fear autonomous vehicles. Customer support workers are having their fears realized, as they are replaced by chatbots. Designers devoted to making 2D views, who should have gone extinct with the advent of 3D, will surely succumb to AI. Automated view creation is seen as low-hanging fruit by at least one CAD company. PR staff churning out formulaic press releases are sure to be replaced by ChatGPT or the like … or did that already happen?
“Hey ChatGPT, finish this building” was the message of a sign on a building under construction. Photos of it went viral. It may have given construction workers existential nightmares. Ironic, because it was the work of a recruiting agency that meant to show how important construction workers were and how AI could never replace them. The irony was lost on AI promoters, too, with its message taken to mean that AI had no limits and that the day was soon approaching when it could be used to create buildings.
Construction workers would feel no better were they to see buildings designed by Autodesk’s Forma. True Forma’s buildings are crude representations with little detail but could easily be seen as a frightening future to those doing low-level architectural work for those in the construction trades, for whom technology is even less familiar.
And then there are those riding the AI wave, waving to the crowd, preening for the cameras. These people fit into three classes:
- AI creators. These are the people who actually create meaningful AI applications. Sam Altman, CEO of OpenAI, is a creator. So, too, is Jensen Huang, whose company Nvidia creates the silicon that almost all AI runs on.
- AI pretenders. These people pretend to be using AI already or, under pressure to perform, to produce AI applications, offer unimaginative roadmaps to AI implementation in current products.
- AI promoters. Here are the people who hype AI. You know who you are. They wrote breathlessly of AI’s ability to detect cancer, create works of art, disrupt everything—but in a good way. When a CAD company announced they had software that could design a car, they fell for it.
Generative Design—Doesn’t That Count?
CAD vendors, who for over 10 years have given up on innovation, insisting that CAD was done, a mature and robust application with nothing more to offer. It was as if the ability to detail a design was the original intention of the MIT researchers who coined the phrase “computer-aided design” in 1959. What of the ability of computers to create new designs all on their own, rather than assisting with the detailing of an engineer’s design? That did not come up.
But when AI burst onto the scene with ChatGPT, even the most contented CAD CEO had to admit there was more to be done. With AI, the computer can actually aid in designing, not just in drawing and detailing as it has since its inception.
Eager to be seen riding the wave that may have caught them by surprise and called to task by investors who are eager to cash in on the AI boom, CAD executives have rushed out roadmaps. The message: AI under construction. Some have stretched the definition of AI to include past work.
We’ve led with generative design, one CAD leader said during a financial call — as if generative design was AI. He got away with it because financial analysts noted any mention of AI without questioning the definition of it.
Generative design, which relies on topology or shape optimization, may be artificial but is in no way intelligent. Watching generative design programs using topology optimization is like watching a baby stacking blocks. Most of the stacks fall over and the few that stay upright — let’s face it, proud parents — aren't very stable.
Topology optimization is good at creating whimsical shapes — shapes an engineer would never conceive of. And for good reason. Most of the shapes topology optimization creates are as ungainly as a tower of blocks made by a baby. With practically any load case, they will collapse. In addition, shapes optimized in this manner are unmanufacturable except by 3D printing—another technology that went through its hype cycle with no adoption by the manufacturing mainstream. Tying generative design to 3D printing is like tying two rocks together and hoping they’ll float.
The Pretenders
The excitement around generative AI, the type of AI that can generate more than the data it is trained on (like ChatGPT) has reached its all-time high according to Gartner, a leading industry analyst group. In its report on AI issued in July 2023, generative AI was at the very top of the 2023 Gartner Hype Cycle for AI.
Some technologies catch on quickly only to falter, unable to meet sky-high expectations, and then, despite their early-to-market advantage, never recover from the disappointment. This happened with Google Glass, the smart glasses abandoned as a consumer product and nowcling to life in limited production for industrial use. Similar fates fell on AR/VR headsets for industrial use and 3D printing, both technologies that have to pull themselves out of the Trough of Disillusionment.AI, after the ChatGPT phenomenon, may be at the Peak of Inflated Expectations. Autonomous vehicles, after Cruise’s fiasco in San Francisco, may be heading into the Trough of Disillusionment. Automated design is just getting started.
The Gartner report added to the pressure on CAD companies to revive their moribund products with AI. From winter 2023 on, conference after conference was dominated by talk of AI, every script was rewritten. AI was front and center on the main stage and in press conferences. CAD companies all would plead their cases — all saying they had been using AI for years, hoping we would tolerate the stretching of the now common definition of AI based on generative AI products like ChatGPT to include context-sensitive menus, UIs that infer relationships to existing geometry, topology optimization or some other cleverness dug up from the past shown as the future.
The pressure to announce something big in this heated race for AI implementation in software is understandable, as is the media's need to report on it. When have we seen this before. Oh, yes, during the Internet bubble when companies reframed themselves as dot coms. The media couldn't get enough of it.