Humans vs AI? | Seek n' Geek IV
Upper left: Hyper-efficient metal design from Arup Newsletter
Upper right: Evolved chair design from Autodesk, published in Fast-co Design
tldr; I think that the design methodology we're learning in 2.70 is increasingly relevant in a world of human + AI collaboration. Precision machine design thinking taught in this course teaches us as students to evaluate designs course to fine, and to identify the functional requirements, design parameters, methods of analysis, references (prior art), the design risks and clever counter-measures. In the future we'll increasingly use computers to run analytical models and generate design concepts, but the important decision making is still best suited to human intelligence.
Can robots do everything, even design bridge components and machines by themselves? “We’re on the cusp of a new era of design.”  “Generative design is so avant-garde that Wikipedia barely has 300 words on it.”  “The industrial designer of the future might just be a computer.”  Fast Co-Design and other design oriented websites have published a number of articles on algorithmic design, generative design, and generally the increasing influence of computers and artificial intelligence on design and engineering. A recent article in Medium called “The Alien Style of Deep Learning - Generative Design” elucidates a number of interesting questions for reflection for any student who wants to work an engineer and designer in the coming decades.
Here are the questions I’m asking and thinking about:
What is optimal design? How can it be created? How do we know it is optimal?
In many of the examples given in the articles, why are the solutions complex rather than sparse? Why do they look “organic” or “biological”?
Does our current education prepare us for our future work? Does it give us the tools, design processes, and important intuitions or do computers make it irrelevant?
Are independent computers truly the most effective, or are humans assisted by computer intelligence?
What is expertise? What is worth learning and what should we just let the computer know for us?
Image from Fastenal Company
Describing, the cover photo of the three metal cable anchors, Salome Galjaard, Team Leader at Arup an engineering consulting firm, says, “This is not only an exciting development for the construction sector, but many other industries as well. In the case of this particular piece, the height is approximately half that of one designed for traditional production methods, while the direct weight reduction per node is 75%. On a construction project that means we could be looking at an overall weight reduction of the total structure of more than 40%. But the really exciting part is that this technique can potentially be applied to any industry that uses complex, high quality, metal products.”
Trying to Make Algorithmic Design Accessible to More Designers
Autodesk’s Project Dreamcatcher questions, “What if a CAD system could generate thousands of design options that all meet your specified goals?”  Their design process starts with defining the challenge, generating the designs, exploring and selecting the best design, then actually fabricating the designs. Inputs into the software include natural language, image inference and CAD geometry.  The challenge is defined by design objectives like functional requirements, material type, manufacturing method, performance criteria, and cost restrictions. The design outputs rely on what Dreamcatcher calls “a procedurally synthesized design space.”  It would be fascinating to understand how this PSDS is formed and grown. This process although highly dependent on the algorithm for the design of the object still relies on human intelligence to understand the design inputs and to select the best design.
Chess playing is a sphere in which human computer competition, and now collaboration has been studied and developed since the mid 1960s. As computer capacity increased, the number of moves evaluated per second increased exponentially eventually enabling computers to consistently beat even grandmasters, but now new research into human-computer collaborative teams show that the combination and effectively using the unique advantages of both human thought and computer analytics is the most competitive. Using this more historical example to understand the current hype of generative design, I think we might experience a similar trajectory. Reference article by the BBC.
Line image from AlgorithmicDesigns.net Kaleidoscope Study
Face implant image from Wired.com article
There are a number of risks, or uncertainties associated with trusting computer design. How can a PE validate the design mathematically to put their stamp of approval? Can this generative AI software really take into account manufacturing methods and cost in creating designs? How will these designs account for parasitic forces? Many of the designs are difficult to create with traditional manufacturing methods, but rapid prototyping methods like 3D printing is great for low volume, high cost parts but at scale this still isn’t cost effective. The Autodesk team claims that “The Dreamcatcher system enables designers to truly leverage an emerging class of manufacturing tools that release designers from hundreds of years of predicating design decisions on tool based constraints.”
As seen in chess, human-AI collaboration if often the most powerful, because smart usage of the different types of intelligence lead to better results. One example is in medical applications where shapes and forms are already organic, current CAD technology is difficult to use. Algorithms and image processing can use pictures of people for reconstructive surgery providing significant benefit to the patient..
Analogous with other fields, it seems that human intuition leveraging AI platforms (like Dreamcatcher) will be a powerful combination in the future.
Extra Reading and References