Tangerine is moving from idea and vision, to reality, thanks to the encouragement of many people like Mark Loomis. Mark’s been interested in ideas for many years. He sent me the following email the other day, which I greatly appreciate. With his permission I include it here in this post. Mark writes:
I read the attached article and it does a good job of differentiating generative design from topology optimization, lattice optimization and parametric design. However, a really interesting legal argument against the use of generative design can be found in the comments after the article. The response is also in the attached file.
His argument made me think about your discussions about the differences between models and drawings and why drawings are still relevant. Your point was drawings take a closer look at the details and that the drawing could be signed off as verified to be accurate. It’s very difficult if not impossible to do that with just a model.
The legal argument made a similar observation. If you have 1,000 designs to choose from the results of generative design, how do you verify your chosen design to be structurally sound if something goes wrong? What if one of the unchosen designs avoids the fatal flaw? How do you defend yourself legally if it can be proved you chose the wrong design? Generative design compounds the gap between models and drawings by expanding the model design space exponentially!
I’d love to hear your take on this legal argument.
Mark Loomis, PRINCIPAL | RLA | ASLA
MP STUDIO | LANDSCAPE ARCHITECTURE, LLC.
The article Mark refers to is: What Generative Design Is and Why It’s the Future of Manufacturing (by Ravi Akella), and the public comment beneath the article, by “MineCostMore”, is copied here:
“Call me old fashioned, but 10k iterations of a solution seems like a bigger problem than the one it solves, as well as a legal nightmare.
At the end of the day it still comes down to experienced Humans choosing the design and that introduces the same biases you’d get from someone going with what they know, but having to winnow through 9,999 potential designs to find the one they like (like being what they already know).
Reinventing the wheel is reinventing the wheel whether it’s automated or not.
But that’s an operational matter and I have to admit that I can see how generative design can be weaponized to buy time with supervisors and clients. I also think the potential exists to really milk RFP’s from government agencies especially for highly complex systems like vehicles and weapons. So there’s some good stuff there, depending on how you look at things.
But what scares me more than parametric randomization of design and inexperienced people approving those designs are the legal implications. Engineers are going to end up on the stand explaining why they chose Design 3045 over Design 7331 when 7331 clearly doesn’t contain the element that failed in 3045. Headlines will read “Fatal School Bus Wreck Ignored 9,999 Better Designs – 43 Killed”.
There’s a lot of elegance in engineered design which creates a strong layer of personal and corporate legal protection. A handful of design iterations can be defended to the nth degree by the engineers involved, but there’s no way for anyone to defend 10,000 design iterations which is what they’ll have to do. Whiteboards can be erased and beer soaked coasters are taken away by the barkeep, along with the freewheeling ideas written on them. But this article is talking about creating a massive body of evidence that can’t be deleted without, you know, destroying evidence.
I’m not sure the weight of the Engineer Of Record is understood here. My company structure protects much of my personal property, but not against criminal charges. Negligence resulting in loss of life means my new house is going to have bars on the windows and be in a really awful neighborhood and I doubt Autodesk is going to model the perfect escape for me.
You’d be mad to pit yourself against those odds. A single design out of (x) iterations that doesn’t contain the failed design elements is all a lawyer needs. Just one. No way.”
I’m no lawyer. But I am experienced with professional drawing and model production. I enjoy the comments above by “MineCostMore”. Absolutely, the “weight of the Engineer (and Architect) of Record”, though well understood by some, is poorly understood by others. He’s right.
It seems strange to say it, and yet it turns out to be true, that the meaning of, and the mechanism for clarifying professional claims of responsibility — over a project’s design and its resultant impact on public safety — is poorly understood, often, or understood not at all, even among people who work alongside these professions, both within design firms, and among various peripheral entities like software companies providing software for professional use.
The reason for the misunderstanding seems to me to be caught up together with basic misunderstanding of the fundamental nature and purpose, of two kinds of media, models and drawings. There are legal aspects of each media that should be obvious to everyone, but that, I think, in many cases, are not obvious. Why? Because what is simple and obvious, becomes clouded and obscure, for a number of reasons, among them: general enthusiasm for technology, and, more importantly, I think, the effect of familiarity.
It is well known, in studies of human psychology, behavior and perception, the human tendency to overlook, to not notice, to not reflect on, things that are too simple and too ordinary, too “right in front of our eyes”.
For things that are so common and so ordinary, in one’s everyday work, we tend to lose the ability to remember what they are, and why they are what they are. This seems to be the case with drawings. We tend to think very little about their functional purpose, while we think a lot about simply producing them, by tedious habit, “because we have to”. We don’t ask why we have to. It’s just our job. We get paid to do it, so we do it. We wish we didn’t have to do it, so we dream of a vague, and incoherent, future in which we don’t have to.
It’s a strange state of affairs. The discourse around software technology, as a tool that produces various media used in the built-environment professions, is held back from productive development, as is software itself, because of inadequate recognition of some fundamental aspects of how these professions function, and the role of different kinds of media in serving this function.
In domains where public health and safety are at stake, the exercise of professional services — by architects and engineers, who are registered and certified by the state to practice — is reliant on clearly controllable communication, and therefore on media capable of clearly controlled expression. I’ve blogged about this before, for example here: Drawings Support Thought and Understanding, and here: 3D is not ‘2D 2.0’ .
There are extremely important different and definitive characteristics of two primary kinds of media in the AEC profession respectively: drawings and models. In the rest of this post I will:
- describe the differences, defining two distinct and unique media, each with their own distinctive purpose, by describing the scope and limits, as media, of drawings and models respectively.
- discuss two different fundamental problems innate to all kinds of models, generally, (including generative models) and finally I’ll…
- add a brief comment about the point above regarding generative design specifically.
Scope of “Models”
Let’s keep to basics. How can we describe the scope of the meaning of the word “model”?
Models occupy space. They are spatial. No matter the type — mental, physical, digital vector, digital point cloud, digital mesh, digital image array, digital hybrid environment — models, like the real world, are environmental. In a sense, they’re wide, expansive, whole environments. I think that’s an effective definition, essentially.
Limits of “Models”
Like yin and yang, scope always goes hand in hand with limits. The major limitation of models arises from the fact that models have better-explored regions and lesser-explored regions. Better developed regions and lesser developed regions. Clearer and fuzzier regions. Areas of higher confidence and lower confidence.
The trick is to know where are the high confidence, clearer, more articulate, better-elaborated regions and conversely where, elsewhere, things are fuzzier. This is a fundamental limitation of models. A model, by itself, gives no indication whatsoever of where we should have higher or lower confidence. It should be obvious, then, that this characteristic of models is a fundamental limitation, a limit with real and substantial practical (and legal) impact.
Let’s move to drawings now, which, as all things in this universe, have both scope and limits.
Scope of “Drawings”
In this post I’m building the argument that drawings are the ideal accompaniment to models (like sound infused into silent film), and not because I want to make that argument, but for a better reason: it’s true.
Models are wide, expansive, environmental, whole things. Drawings are nothing like this at all. On the contrary, and as polar opposite, drawings are narrow and focused; they embody the act of taking a closer look.
Fundamentally, drawings and models, as media, have essentially nothing in common with each other, but rather, instead, they are a true pair, not a pair of convenience or accidental proximity, but a pair born of absolute necessity. You can demonstrate this to yourself in a few seconds. Observe the environment in which you currently stand, or sit, or whatever you’re doing. Try to understand the environment effectively, in order to do something, like take a seat, or walk through a doorway, or pick up a glass of water.
The environment is all around you. How do you understand it and act within it? Well, you begin by taking a closer look… here, and here, and here, and there. This is fundamental. Thinking doesn’t happen without it. Understanding will not develop, without it. I discuss this more, below in the section on problems.
This act, of “taking a closer look” sufficiently describes the scope of drawing as a medium, although additional important notice may be taken of the role that “taking a closer look” plays in articulating the clearly controlled communication on which the architecture and engineering professions rely.
Also note that the scope of drawing is innately tied to its corresponding limits.
Limits of “Drawings”
Fundamentally, the act of taking a closer look is meaningless, unless there is a wider environment within which the closer look occurs. This is uncontroversial. If there is no forest, then there is no “taking a closer look in a forest”. If there is no mental model — or ‘mental and digital’ model — of a building, then there is no “taking a closer look within a building model”. This well and truly does mean exactly what it sounds like it means: without a model (at least an imaginary one, a mental model), drawing is not possible. Sure, there may be found some exceptions to this, in some kinds of experimental abstract art, but in the AEC industry this holds.
I believe it suffices to say that no person, author or viewer, has ever understood a drawing, not in a meaningful way, without engaging in the mental activity of instantiating that drawing in-situ within a mental model of the wider whole of the environment within which the drawing gains its meaning. It’s interesting that conversely, there is bi-directionality. A drawing gains its meaning, when it is put into a model, and likewise, a nascent drawing, one being authored, comes into being from, out of, a model.
Both while authoring a drawing, and, after authoring, as others are viewing a completed drawing, meaning comes from a mutual interdependency and interplay, in the mind, between drawing and model; each derives its meaning in interplay with the other. To make sense of a drawing, viewers imagine the drawing in its place within the wider whole of an environment, within a model, or portion of a model. And the reverse: to make sense of a spatial environment, modeled or real, people engage in the act of taking a closer look, within it, the act embodied traditionally in the medium of drawing.
Let’s elaborate more on a discussion of significant problems of each of these media. The major problem with drawing is that drawing requires a model for meaning to be developed. The reverse is true for models. This of course suggests the possible value of drawing-model fusion. More on that later, after a more detailed look at the problem of model media. There are two primary types of problems with models. The first type of problem is a hard problem, a major practical problem. The second type of problem is worse, a more fundamental problem that goes right to the core of the nature of human cognition.
Two problems innate to models:
There are two types of problems with any kind of model — mental, physical, digital vector, digital point cloud, digital mesh, digital image array, digital hybrid environment. All of these, whether they are produced by generative design or not produced by generative design, share certain basic definitive characteristics, the scope and limits that are characteristic of their media type: model. And so they all exhibit the same problems.
Models: Problem #1
The first problem innate to models, that is, innate to models standing alone unassisted by any other media, is that they’re fuzzy, of uncertain quality in many regions, while at the same time they are in fact “good enough” in other regions. The problem is knowing the difference.
Models, by themselves, reveal nothing about this. They enable/allow no certainty whatsoever and make no affirmation, of being “good enough”, anywhere, no matter that in some areas, in some regions, they in fact are “good enough”. That is, in some regions, they’re satisfactory of the relevant professional standard of care.
It takes only a second to recognize that this kind of fundamental opaqueness, doubt, confusion, fuzziness, lack of essential clarity is untenable, full stop, in professions reliant on clearly controlled communication. There is no negotiation on this point and there never will be. Models simply provide no legal ground to stand on whatsoever (except in a few specialized exceptional cases). Not by themselves.
Let’s talk about what clearly controlled communication, about spatial assets like buildings and such things, actually looks like in the AEC professions. Clearly controlled communication exhibits precisely the following kind of clarity and affirmation, or assertion:
“Here, I’ve taken a closer look; at this location, here, I’ve made clear my communication of something that matters; I want you to look here, at what I’ve shown, and I want you to look here, for two reasons. 1: what I’ve shown here matters, and 2: I’ve shown what matters here in a way that’s understandable, that makes clear what matters. And, following from this, or integral to it, I affirm, or rather, I assert, that everything that should be shown here, is in fact shown here. And I further affirm that nothing that matters, here, is missing.
Professions reliant on clearly controlled communication are reliant precisely on this. It cannot be otherwise, no matter the inscrutable pronouncements one hears commonly in what has become a strangely incoherent discourse surrounding AEC software technology these days, and for many years gone already.
The necessary clarity comes very simply from knowing the difference between everywhere and somewhere. An architect or engineer may in the course of a project make controlled communications describing thousands of specific somewhere’s…, here’s. And still, these thousands of drawings, though many, are finite in number, much much much less than everywhere, and therefore their clarity and affirmation is manageable. The author of these communications, because they are finite in number, has the required time to review each of them carefully, to study, develop, improve and complete them, and to issue them with a signature asserting that each of them meets her/his professional standard of care.
So what happens to professions reliant on clearly controlled communication when their scope of communication expands from a finite number of somewhere’s, to everywhere within the entire spatial environment of a proposed project? Let’s take a look at what happens to clearly controlled communication when that happens. How is one going to be able to assert that:
“…everything that should be shown everywhere, in the entire space of the project, is in fact shown everywhere. And nothing that matters, anywhere, is missing.”
This is a fundamentally different task and achieving it is an absolute absurdity. And remember this is the minor problem among the two problems of models. Entertainment of such absurd ideas as this must be brought finally to an end, after decades already of incoherent rubbish common in the standard discourse. Note the rest of it:
“Everywhere, I’ve made clear my communication of what matters, which is everything everywhere; I want you to look everywhere, at what I’ve shown everywhere, for two reasons. 1: Everything everywhere matters, equally, and 2: I’ve shown everything everywhere (even though that’s impossible) in a way that’s understandable, that makes clear what matters, which is everything everywhere
(This is in John Cleese territory of the absurd)
Does it make any sense at all to say that now that we have modeled digital environments, that therefore we can discard and dispense with the act of taking a closer look (at those environments)? No. It is an homogeneously absurd notion, with not a single valuable nugget of truth or meaning in it.
In terms of legal repercussions, on professions, the difference between drawings and models, in terms of facility for providing clearly controlled communication, is as profound as the difference between the words “somewhere” (or “here”) and “everywhere”. To equate these, and/or, to suggest that it’s desirable to replace “here” with “everywhere”, is really to indicate that you’ve slipped far away from understanding what’s going on, that your thought process has slid off the rails.
Models: Problem #2
The second problem with models is a harder problem, more fundamental. It goes right to the nature of human cognition itself. Legal problems by comparison are simpler. The first problem, lack of clarity, obfuscation of what matters, inability to affirm or assert where a model is “good enough” and where, conversely, it isn’t, is, compared to the second problem, minor. The second problem is (even) more basic.
Understanding comes of course from thinking. And where the basic dynamic of thinking is undermined, understanding fails.
What is thinking? That’s a question for scientists and philosophers, a question that has yet never been satisfactorily answered, and probably never will be. But we can answer a much simpler question:
What is, in our everyday experience, the basic observable dynamic of thinking?
I’ll assert that the answer looks something like this:
- Whole things whole, we can’t understand
- We must narrow to simplified essence to understand things
- And, (2) makes no sense without (1), and (1) makes no sense without (2).
With this simple recognition we’ve hit on something essential about the nature of thinking itself.
Thinking (and understanding), not only involves, but essentially is, a bouncing back and forth between (1) and (2), a ping-ponging between things as they are, and simplified essential representations, between wide environmental whole, and the narrowing act of articulate focus, the act of taking a closer look.
Between these is where thinking happens and understanding grows.
I have some stories to tell about this, from daily observation and from my own professional work producing drawings and models. Perhaps you’ll find these stories useful:
These two problems suggest an inevitable fusion, of drawings IN models. They also suggests the possibility of evolution yet to come, in the expression and form of “drawing”, a re-envisioning of what “drawing” can be.
Generative Design Models
The points made by “MineCostMore” in the comment copied at the top of this post sound valid to me, not as a rejection of generative methods of modeling, but more narrowly as a new kind of liability worth addressing.
One aspect of the problem identified is simply the generalizable inadequacy of models, alone, in supporting clearly controllable communication.
This part of the problem can be mitigated by further development and evolution of the medium of ‘model’ itself, such that a model will incorporate within itself exactly what is needed for providing the necessary clarity.
From Tangerine’s web page:
Tangerine’s Media Innovation Spec 2018 specifies a next-generation evolution, of the expression of the essential function of “drawing”, through equipment, which we detail in the spec, for “taking a closer look“, built-in within modeled environments of any kind. The spec is now available for those wishing to make use of it within their own software development organizations.
The other problem, that of lawyers indicting engineers for generative designs not selected, is, I agree with Mark Loomis, a similar problem, a related problem, one for which I’m at a loss to answer and that will require further thought.
Perhaps the answer will be found in framing that problem as it actually is, as a problem that closely mirrors the basic nature of thought itself, a dynamic back and forth, between everything that is possible, and, narrowing down from everywhere and everything, to somewhere and something, continuously ping-ponging back and forth between these poles, in order to produce, ultimately, meaning and understanding.
In the case of generative design producing hundreds, or millions of possible iterations, well and truly surpassing the ability of any engineer to carefully evaluate all the relevant characteristics of each, perhaps what is going to be required, legally, is that the engineer be at least able to demonstrate, when challenged:
- that, using best available automated evaluation tools that evaluate against mission critical criteria, the engineer exercised due diligence, meeting standard of care as it will be defined for generative design scenarios, in narrowing from all possible solutions, to a manageable set of eligible candidate designs, and
- that for the selected (developed) design(s), that clearly controlled communication is maintained.
In some cases, then, “generative” methods of modeling may present a specialized new kind of problem, a new kind of liability worth addressing carefully. The result of (1) above is a set of design iterations vetted by the engineer. Could there be transparency in the vetting process? Could the process involve a legal means for the deletion of all other design iterations from further future investigation?
Generative modeling though is a subset of “modeling” generally. All kinds of models, including generative models and all other types of models, produced by any means, share basic definitive characteristics:
- They’re expansive, environmental. (scope)
- They occupy space. (scope)
- Their quality and validity varies throughout that space. (limits)
- They present, by themselves, no clarity regarding where their quality is high; they offer no confidence in knowing where they’ve been determined, and affirmed, to be “good enough”, and where therefore they may not be. (limits)
The Right Pair, in Fusion
In an elegant demonstration that models and drawings are a right and inevitable pair, the limits of drawing are mitigated by the scope of models, and vis versa, the limits of models are mitigated by the scope of drawing. It’s such a fine pair in fact that fusion is not only suggested, it is, like the fusion of sound into film, imperative, certainly so in cultures seeking effective, practical, powerful and inspiring development and evolution of both computing equipment (hardware and software), and media itself.
Drawing-model fusion is going to be a significant driver of innovation that matters.
Kindly have a look at Tangerine’s website for our past work on drawing-model fusion: Earlier Media Innovations (2012), and our proposal and specification for new work, Tangerine Media Innovation Specification 2018, that obviates and surpasses that earlier work, and its patents.
Please feel free to express your views on these issues in the comments section below!