This post is also available in: Italiano
Although today it is still perceived as something very abstract and theoretical, Generative Design is a creative theme on which the potential of the Manufacturing of the Future are finding their focus. Why are the expectations about it so high? Overturning the existing design paradigms in search of objective-based systems allows one to explore countless formal solutions thanks to an artificial intelligence that is able to considerably exceed the creative abilities of designers. This allows one to look beyond, to evaluate new possibilities in search of the limits that the technologies and materials can guarantee.
To learn more about these aspects, we got to meet one of the world’s leading experts in the field of Generative Design: Francesco Iorio, one of the development leaders of Project Dreamcatcher, the Generative Design platform that Autodesk has invested in for the technological future of design and production. Our reflection starts from this aspect in order to better understand the goals and ambitions of a project of this magnitude.
(F.I.) The beginning of this adventure goes back to five years ago. I received a request from the Autodesk leadership to come up with a project capable of sustaining an innovative vision both within the context of design and, of course, in the context of production. We started experimenting with a small team in Toronto. Together with my colleagues we achieved early encouraging results which convinced the leadership of the soundness of the new method. Today we have about 40 people working on Project Dreamcatcher including staff in the San Francisco and London offices. The team is growing and consists of individuals with a very broad set of skills. We have experts in mathematics, geometry, machine learning, mechanical engineering, materials science, structures, user experience, software development and, of course, design. We are constantly looking for other specialists.
(3DS) What is Generative Design for you?
(F.I.) To provide endless creative possibilities without losing sight of the most important objective, i.e. project feasibility. The aspects to consider are numerous. From formal ones, related to the consistency of the design that the system must be able to implement, to the more exquisitely material-related ones concerning its physical properties. Finally, we cannot ignore the characteristics of the production systems. Our goal is to infuse intelligence into software so that a designer can insert a data set, have the program process it and obtain proposals that match his requirements.
(3DS) The system is based upon artificial intelligence. How is it possible to train a Generative Design system?
(F.I.) The question arises especially on an objective basis. You go from having to communicate functions to having to communicate systems of behavior. The starting point is a requirement. This must be translated into mathematical formulas, a system of families of generators able to observe, interpret and propose solutions able to satisfy this requirement.
(3DS) In theory, everything is perfect. In practice, what are the technological limitations and difficulties that you must address on a daily basis when implementing such a complex Generative Design system?
(F.I.) In recent years there have been major advances in the field of artificial intelligence but the basic limitations have not yet been fully overcome. The underlying problems remains the same: the need to train the system with a huge amount of data and the fact that, as far as the current technologies are concerned, the system itself is not able to recognize whether a piece of data is substantially correct, complete, or not. Thus the initial selection of the data is essential, otherwise the system breaks down, becoming potentially ineffective in meeting the requirements with their simulations. Conceptually, machine learning is not as complex as it is to translate it into something that actually works. This is an open challenge on which there is still a lot of work to do. In general, of course, not only for the aspects that relate to Generative Design. What we do is only a small part of what is happening in the field of artificial intelligence.
(3DS) How is this complexity addressed and resolved?
(F.I.) By narrowing the scope, focusing heavily on observation. A Generative Design system has unlimited potential, but to train it takes a lot of concreteness. The first applications of Project Dreamcatcher, in fact, looked at circumscribed scopes. We have some case studies in design; we have designed an automobile frame from scratch and are currently cooperating with the aerospace industry. Thanks to this very selective approach we were able to get to the first results thanks to which it will be possible to gradually open the system to a wider network of partners and collaborators while waiting to finally democratize access to it.
(3DS) How does artificial intelligence take shape? How does all this richness become software that can be used by a designer who does not have specific IT skills, but who is rightly interested in aspects related to the design and production of new products?
(F.I.) The creation of the interfaces is the result of the work of a multidisciplinary team and of constant cooperation with designers. It is necessary to understand their needs and how they will evolve in the future, the development is constantly evolving. Behind the user interface there is a technological core that comes, as I mentioned, from an AI system trained on the basis of the observation of the technological history of a product, of the compatible materials and of the production processes. 3D printing has literally opened up a world of possibilities in this respect. And that’s not all. The formal synthesis and optimization required to meet the design requirements would be impossible without a production method based on Additive Manufacturing. The classic example that is used in the field of Generative Design is related to the formal synthesis of a chair. If we are able to train the system on the different types of chairs produced so far, provide it with all the specifications of the materials used and give it the structural constraints derived from the limitations of the production processes, we can obtain simulations that are truly useful to a designer’s work. According to this logic, Generative Design becomes a kind of design assistant which, thanks to its proposals, makes you think about the problem, focusing on aspects that the designer, even if extremely experienced, cannot humanly foresee. This is because the creative possibilities are virtually endless. The system acts as an assistant because it is not the one who defines the design choices; its task is to orient them properly, especially highlighting the advantages that one solution has over another in meeting a certain requirement. The most common result of formal synthesis is to obtain an object with the same mechanical properties as a traditional one, but cheaper and lighter thanks to the use of less material. And this is just the starting point. Today we don’t know what the limits might be including those from the point of view of production systems. 3D printing can benefit from generative design more than other manufacturing methods given the enormous, if not excessive, freedom of shapes available. At the same time, other manufacturing technologies, which today we call traditional and subtractive, could be revitalized if they were applied according to the procedures we are working on. This is because, although the constraints of shapes and materials are obviously more restrictive, they will be able to produce things that are currently outside the spectrum of human knowledge and intuition. In that context, generative design will increasingly find its own dimension.
(3DS) The definition of the requirement itself thus becomes a fundamental choice for the designer. How can the software help him to support his choices?
(F.I.) A project can be approached from several directions. So I think at least two aspects are fundamental. The first is related to the confidence in the machine and the software, which by now is well established in the modus operandi of design. This basis must necessarily be augmented by the component linked to intuition, to the perception of the possibilities. And this is the aspect that is becoming increasingly important to work on. The communication of the project’s data is fundamental. Computer graphics have reached a very high standard with regard to realism. The challenge that awaits it now is the visualization of the data. In order to choose, a designer needs to see, needs to have a tangible picture of what the possibilities are. The case of the chair, which we mentioned earlier, is deliberately aimed at achieving a single requirement. The goal that can be established by a Generative Design system is much more ambitious. We can get the sense of this if we think about a much more complex product, like an automobile, where there is the need to meet many requirements.
(3DS) The problem becomes almost more interesting than its solution.
(F.I.) The solution remains the goal, but to define new design horizons we need new problems to solve. For this reason, also in terms of software design, it is important to think it through when defining the lists of requirements in order to ensure practical support to a totally new vision of the configuration of a product. This approach is essential if the objective is truly the design of brand new solutions. This applies to many areas of design. Our intention is to begin this journey where there is the most predisposition and investment opportunities in research, such as the automotive and aerospace industries. The medium-term objective is to make the technology accessible to small and medium-sized industries, where the creative potential is enormous.
(3DS) As you rightly pointed out, the most fascinating aspect of a Generative Design system is probably that it has endless potential. The more the system learns, the more powerful its simulative capability becomes. Is this interpretation of the phenomenon correct?
(F.I.) Certainly. We are talking about enormous possibilities that we are building together. A Generative Design platform grows thanks to data from its partners, developing more powerful capabilities for simulation in order to meet requirements in an increasingly effective way. If the platform has an objective need for data to be processed, each partner receives a huge benefit from a large shared resource.
(3DS) We are looking at a real Cloud based application which, without the sharing of information on the network, would not have a reason to exist and which, thanks to the Cloud infrastructure, can generate benefits that would otherwise be impossible.
(F.I.) The future of Cloud computing should continue to position itself to gain exclusivity. To create an actual added value. Otherwise the investments in new infrastructure would be superfluous, or otherwise underused. In a transitional phase like the one we are experiencing there are technologies which are not yet ready and others that are but cannot yet be exploited in all of their actual potential. When we began a new Generative Design journey with Autodesk, a fixed point was precisely to try to work on a project that was not excessively limited by the current limits of the technologies and of the digital infrastructure. This is not to say that we were not realistic or, worse, that we wanted to be visionaries, I simply want to say that to try to innovate it is necessary to look beyond, being confident that the current limitations will be overcome somehow. We need to invest in this regard, with a vision that goes beyond the medium term. As an example, we could ask ourselves the following questions: what will the manufacturing scenario be like in ten years? What problems will need to be solved? And, finally, how can we support all this?
(3DS) The sharing of data necessarily involves a risk for any company, especially in the industrial contexts where patents and intellectual property are priceless. A Generative Design system like Project Dreamcatcher exists essentially on the basis of data sharing. How will you be able to overcome, and especially manage, these aspects in order to overcome the skepticism of companies?
(F.I.) The Cloud is indispensable in many ways, starting with the computing power needed to process a huge volume of data, such as that required for the simulations. At the same time, the accumulated knowledge will become an essential value, able to generate new expressions of design. In practical terms, we are hard at work to fine-tune policy criteria to enable each partner to choose what to communicate and share. Also in this context, the new business models will have to evolve, generating new problems. I am not a lawyer, so I have no first-hand experience in these types of situations in order to express a technical opinion. In the past there have been situations where certain limits seemed insurmountable. If we were to evaluate them now, in hindsight, we would probably smile. I am convinced that continuing on the path of research and progressive implementation of the application, technologies like Project Dreamcatcher will someday truly be in everyone’s reach. The benefits are so clear that finding solutions to problems including technical, legal and legislative are incentivized in order to permit its natural diffusion.
As an appendix to the thought and vision of Francesco Iorio, we offer one of the most interesting and inspirational contributions to understanding the creative possibilities that artificial intelligence systems will be able to support: the TED talk given by the futurist Maurice Conti in Portland, OR.
This post is also available in: Italiano