The Role of AI in Design

Is AI in design just a myth? Will artificial intelligence ever be able to do or contribute to design? Right now the public face of AI is a bunch of dystopian Big Brother movies, and smart digital assistants that pull off the parlor trick of a machine being able to answer spoken human questions with (sometimes) relevant answers. You’ve seen the Big Bang Theory episode where Raj meets Siri? Kind of a cynical poke at automated assistants in general, but it was pretty funny.

I have a Google assistant in my house, and so far I have only used it to find my misplaced phone. My wife uses it to listen to aggregated news in the morning while she does other tasks. I can’t say that it’s extremely useful, but it does have its uses.

Games continue to be one of the biggest uses of computer technology, but they have helped pay for popularizing technology that has become useful in a number of other ways that maybe we couldn’t envision at the time. I think AI will have a similar growth curve. I’ve spoken with people directly involved in developing what might be some seriously powerful applications of AI, specifically with deriving patterns from large amounts of data. And not just for the silly stuff using it for targeted advertisement, but for finding causes and cures for disease.

It might help to define artificial intelligence. I roughly use a definition like: the ability of an electronic system to make decisions leading to a goal based on some form of input data and software/programming/hardware. The data could be in form of active sensors, or it could be a huge database. Learning seems to be a key requirement, which would mean it either adds to its database or adds new rules/conditions/decision making capabilities. Under this definition most computer programs would exhibit some level of AI, since they generally make decisions based on user input. Some software can learn, like CAM software that learns what options you prefer.

My real question in this post is whether or not this could be applied to design. Maybe I’m leaning on my knowledge of geekery in pop culture, but the example of Star Trek’s Data is an obvious nod to AI in the future. Data is able to do creative things like painting, music, poetry, etc. Maybe this is a realistic model, and maybe it isn’t.

A less obvious Star Trek reference is from the Voyager series, when the crew designed the Delta Flyer using voice commands and the Holodeck. It was done starting from just a hunk of digital clay, and then as people issued voice commands, the computer added or removed various aspects of the shuttle craft.

There were other episodes, from TNG, where the crew recreated a torture station from their dreams. As they told the computer to add sounds, lights, shadows, a table, an arm, they could change aspects of it. In both of these episodes, the design was collaborative, so multiple people were standing around giving the computer design related commands.

And then there was the episode where Geordi recreated a scene on the holodeck based on tricorder data, and extrapolated the position of an unseen body based on light sources and shadows.

Obviously, this stuff is way out in the future, but I can kind of see design headed that direction. Based on certain inputs or a previous design, fill these requirements in some way general enough to be understood by a computer. Yes, I think that much is possible, but not in the next 10 years. We seem to be stuck. I think we were all hoping for more from recent CAD starts.

Some time ago I did an interview with a well respected CAD executive, and at the end of another line of questions, I kind of ambushed him with this question about what he thought CAD had to gain from AI. He really wasn’t prepared for the question, so I didn’t include this bit with the rest of the original interview, but I thought his observations were relevant, and appropriate to the current discussion, so I’ve included it here.

=======================================================

What place do you think Artificial Intelligence is gonna play in the future of CAD?

<awkward silence, then uncomfortable chuckle> Choose a side. Do you support Elon Musk or Mark Zuckerberg on this subject? Technology plays buzzword bingo every day. In the end, there are still some core fundamental problems with most of the systems and processes that people use. We’re focused on the core of what people are doing in design and process flow. Do I think artificial intelligence will help? Yes. Are there neural nets that will automatically create the next thing? I don’t know. There was all the promise way back when of rules based engineering. The problem was the systems were too hard to program to create all the different rules. Can the systems be smarter about acting on the data? Sure. Are we using some of those things algorithmically to figure out how we manage performance, as we’re monitoring things, what parameters get driven by the equation to figure out when to spawn new instances, and when not to? There are some hard and fast rules, and there are some other ones that we can maybe optimize over time. You’ll see things introduced – or you’ll see the benefits of things. Hopefully you’ll never actually see the “AI” because our customers don’t really care about that. They care about what the results mean to them.

AI is extremely interesting, and possibly extremely overhyped right now. It may help with the following. You can envision a world with Alexa or Siri that as people use them more and more, you’re getting voice data patterns, and the systems get better and better at interpreting what people say. Except for my wife who has an accent. Do I think in the world of finding the right part or material or the right information about something … there may be a role to play here. I can imagine some applications in high performance engineering that is looking for something that can withstand some list of qualifications, maybe contradictory requirements like withstands heat, but is also very stiff, and not brittle when cold. I can imagine some kind of a role that is played in time. It’s just not immediately clear how it plays out in our world. Maybe I’m not smart enough to figure it out, but I think our customers are more concerned about what does it mean to me, and how does it impact my life and the way in which I use the system.

Make no doubt about it, I’m sure a lot of people are going to start touting a bunch of things about AI, but I don’t know how that’s going to impact CAD. Maybe when you collect a lot of information about say a building, and you’ve got all this data, you might be able to do some predictive stuff, like the Predix stuff from GE, that’s where you’re going to see some benefit. I think CAD users might benefit from the suggestions and diagnostics. I think in a CAD design aspect it’s not going to make that big an impact, but I can imagine that GE (and I’m not an aircraft engine design expert) gets all kinds of turbine sensor data over time. They can take that sensor data and be able to do some behind-the-scenes massive amounts of correlation related to repair or cycle times vs operating parameters and suggest maintenance based on events.

Right now, AI is such a hot topic that you’re going to hear a bit of technology looking for a problem rather than a problem looking for a solution.