# Any one into Generative Design ?

“Medieval methods” is condescending as it is inaccurate. Industrial Design methods date back to the 1930’s, and many of our current methods (ethnography, personas, usability, rapid prototyping) came from the last 20 years. So let’s come up with a better title.

What do you mean by judging on 100 designs? If I enter this contest, I’ll probably create only a handful of good quality designs vs. your hundreds of random designs. So let’s say “limited to 100 designs” per entry.

This is going to be a 1 hour challenge start-to-finish right? So that means all the GD programming and output has to happen within that timeframe. Right?

For me the link is clear between the two. The claim is made that the design process or parts thereof can be broken into discreet chunks and then made into a formula. If this can be accomplished, then the result can be:
a. Covered in a friendly UI and sold as an app for individual users to generate designs. The result will be x number of designs based on the input parameters.
b. Iterated on by a machine. The result will be x^n number of designs based on the input parameters.

Suppose there are a set of three OLED monitors sitting in Nike marketing and sales HQ, one shows the minute by minute user generated shoe combination’s from the website and stores, and one shows the result of a machine sifting through combination’s of the variables. The third monitor is connected to the Nike design department. I put my money on the design department to come up with the designs that will have resonance with the sales and marketing.

The results of monitors one and two, user design and gendesign, would be hard to differentiate.

Let’s not prejudice the contest with the title, let’s instead call it “2010 Designer Methods vs Theoretical Approaches by Computer Programmers to Solving a Non-Existent Problem with Machine Logic and Human Sorting of the Results”

You propose that, in order to level the playing field, that we use black and white line drawing generated in both cases by the machine. In effect, reduce the human element out of the equation, no sketches, no hand models, no mockups, no gray areas of inspiration found in the edges.
This betrays your lack of understanding of how humans create and evaluate a design. I suggest that you tailor the output of your machine program up to human expectations and compete on a modern level playing field. Since I would like a real competition, here is a clue to that, teach your program to sketch. This is already possible with various tools the easy one being SketchUp. Study the reasons why a sketch is better than a rendering.

I doubt you will.

cg, that is of course impossible. My personal “design programming” has been thirty plus years in the making, and I apply that in each hour I work. Since I believe I know both sides of this equation pretty well, and retain my confidence in the human creative process, we could suspend both the time and money limitations on the GD side without concern.

I’d be happy to give my opinion, but how it’s done here is that everyone discusses the outcome

Believe me, I’ve embraced the CAD revolution, ways of doing things outside the traditional ID, and am open to new tools… though lately I’m trying to pull back and get down to thought behind designs more. The things written above - trends, emotion, cultural relevance - are difficult to understand (at least for me), let alone capture and I can’t begin to guess how you’d include that in the programming

How would you aim at the spirit of a brand in something like a perfume or whiskey bottle? It’s a lot more than shape I think

The answerer is simple. You don’t even try. The complexity of it is beyond the reach of current computational process. Generative processes is a dumb assed process but an immensely powerful one. It can create great results, as nature does; provided the initial models are built sensibly (with some design intent) and selection is made intelligently. So, finally its about choice. It’s designing by selection.

Now for the contest. The purpose of this test is not to imitate, compete against or test against the steps involved in the medieval processes. The end results need to be compared, not by designers, but by those whom it is intended for - consumers. This way we can avoid the prejudices of designers (including the search for pencil marks) and asses if it is capable of producing similar end results.

I suggest a “Turing test” http://en.wikipedia.org/wiki/Turing_test/ . This will involve the comparison of machine generated and designer generated end results using a blind test method, where 100 or so designs can be shown to unbiased parties who will be asked to identify the ones that they like. If they choose the ones generated by medieval processes then we know that it is superior design method. If they score equally then both processes are equally good or bad.

I believe that this would be the recommended scientific test for such a comparison. This will help us judge end results. We can compare time taken separately.

SK, As yo has asked several times, I just read through again and do not find a response from you.

Wow, I just read through 206 replies.

It appears from this thread (and the similarly entertaining “designers vs. engineers” thread Do engineers really "design" anything?) that the industrial design process is not understood. Industrial Design is about human and technological interactions.

I see Generative Design, based on the examples in this thread, as Generative Styling.

All the examples of GD shown earlier do not show any consideration between the human/ technology interface. Using the mp3 example they have the same five buttons moved around a bit, hundreds of times. The value of design is observing how people use mp3 players, in what context, and working out solutions based on what people want/need. Do they use it in the left-hand or right-hand? one-handed with their thumb or in both hands like a game controller? where do they listen to mp3 players? what do they listen to? and a million other possible questions based on observation, intuition, personal experience, book smarts and so on. This is what determines where the buttons go, how many, what they do, what they are made of - not a SolidWorks design table.

Styling is important too, people like beautiful things and beautiful things work better (see Donald Norman).

I can see a role for GD, but at a superficial level. For industrial design GD is the Bat-computer, you ask it an obvious question and it spits out the correct answer:

ahhh come on man, is a lesson in contradiction 101?

Now for the contest. The purpose of this test is not to imitate, compete against or test against the steps involved in the medieval processes. The end results need to be compared, not by designers, but by those whom it is intended for - consumers. This way we can avoid the prejudices of designers (including the search for pencil marks) and asses if it is capable of producing similar end results.

You want to reach an end result so the target user can compare and pick which one is best. Fair enough a good test.

In order to reach an end result to do that you need to under go a ‘process’ .

As each end result is a derivitive of their individual process ( the rather rudely termed “medieval” and the generative) you are also comparing indirectly each process as to which one was more effective. (This is how processes evolve)

The test would be good idea, but what brief? I cant see (for arguments sake) generative design at it’s current state of evolution solving an open brief of how to reduce cases of tinnitus among 16-35 year olds. I would assume it would have to be (based on the rather generic MP3 and spoons) a styling job. How about pitching a new concept phone for Nokia, taking their brand DNA for an emerging market …lets say India (as an example), with a time frame of a 1-2 weeks.

Also back to your point of generating 100 designs for consumers, which seems to be one of the strengths of generative design… what is the point? It is not possible, especially for a consumer, to evaluate so many. On average from my experience of user feedback sessions at our firm most people can hold their concentration for just over 25-30 minutes in an evaluative environment, thats after spending the first 5-10 minutes breaking the ice :p. stick 100 designs infront of them and they wouldn’t know where to begin.

I think the underlying response to this whole thread is the following. We have agreed the merits of generative design, as a tool. CG and Brook suggested some great examples with regarding to internal part structures, ribbing bosses etc… I am amazed the thread has gone any further. The issue with this “heated” discussion I think is your repetitve advocation of our “medieval” methods being able to be replaced by punching in a set of parameters and sitting back and watching 1000’s of designs generated.

With regards to Yo’s question still being unaswered you can find some answers by searching on Linkden as I did, it would be nice to hear from SK though, but seeing as it has been repeatedly ignored I would say it is safe to assume no. This would be a good case study for the feedback and context post :p.

I dont think anyone hear is worried about generative design… we only need to be worried when generative design develops into skynet.

So perhaps human designers fag out after hand full of designs ? perhaps this a limitation ?

I suggest that we retain “medieval methods” as it symbolizes a period in design history where design was though to be purely a result of designers mental deliberation; " curious tools" because, its an equally silly concept.

One person need not judge 100 designs. I suggested a sample size.

The reason for suggesting 100 designs is because there are billions of distinctively different people in this planet and their preferences differ. Also, to avoid the satisfaction that yo would derive in comparing what it generated with what he is able to produce and to able to pass a personal judgment on; perhaps on behalf of all IDers. I have no time for such entertainment.

What I have time for, and serious interest in, is to make a fair and scientific assessment of the capabilities of the curious tools using practices adopted by computer scientists. Incidentally, it can be published even if the curious tools perform miserably; because the spirit of science is about constant improvement and openness to new understanding. But in this case its not new. Its 150 years too late. Its 150 years after the scientific community came to accept how life was designed. In may be unpublishable on that account - for there is nothing new in this approach.

I don’t think you meant what you said, but I’m totally using that line. “Yeah, I was generating concepts, and then, well, I just fagged out, whatever”

I’m interested in the posts on this thread so far. Having researched a while into this area doing my university thesis, I happened to stumble upon tools, such as generative design based on evolutionary algorithms, in many industrial engineering research papers of late. Other similar tools include shape morphing and shape grammer. For a while, I was sucked into the potential of these tools. Having a bunch of forms that are automatically generated and from which you can pick from and develop into potential products was incredibly appealing.

However, after researching deeper into this, I eventually came to a similar conclusion as sanjyoo9. Basically lets break down what we as designers are always trying to do - add value by creating meaning in products we design. What do we mean by meaning? Meaning is created on many levels - the visceral, the behavioural and the reflective (as termed by Norman). If we closely examine generative design tools, we find that these do indeed help us design better, but only in the visceral level of emotional design. If we drilled down even further, I would say that it is specifically aided in the visual aspect of design.

At the end of the day, what I would like to say is this. I don’t think generative design should be dismissed as a tool, but do see it as a tool that aids the type of product we are designing. Generative design would probably be an excellent tool if we were to design say a mature product, where the behavioural aspect of it is not going to be redesigned and the reflective aspect is probably non-existent. An example, a computer mouse. Funtionally, we’re all used to just putting our hands over it and we aren’t going to be philosophisizing over the mouse. So, satisfying the visceral aspect is suffiicient to succeed in the market I guess.

I agree with you, and I don’t think anyone here is discounting generative design as a tool. But, is “reskinning” a mature product without many fuzzy factors (such as a mouse) the sustainable future of design? I would say it is not.

(BTW, the mouse is reaching the end of it’s life, though it may never go away completely. Touch screens and track pads and other things are replacing it. I use a pen/cintiq at a desk, and a trackpad on the go, even in 3D apps )

SK, you’re argument for GD would go a long way if you could answer Yo’s question. The fact that you have avoided it leaves us all to believe you can’t answer it and thus it really kills your argument. From what I have read, designers are not against GD they are against using it exclusively as an answer to design. It seems to me it’s only one step in the entire process of design.

Then again maybe you can’t answer the question because you work for Skynet.

I agree with you, and I don’t think anyone here is discounting generative design as a tool. But, is “reskinning” a mature product without many fuzzy factors (such as a mouse) the sustainable future of design? I would say it is not.

I think we may simply conclude that generative design is not a universal design tool for all product categories then.

It is true that reskinning doesnt really seem much, but consider the fact that many subsequent product development within a product line are essentially reskinning processes coupled with minor technical upgrades, no? These are the products that generate the most profit for companies. The breakthrough products, that are gonna ‘change the world’ are one in a million?

I would like to clarify that I’m not a adamant proponent of GD. I have enormous respect for the work of you as designers. It’s just that I’ve been through this entire argument doing my thesis and this was what I wanted to highlight.

Uh, periods of human history that embraced mental deliberation tended to be a transition away from medieval methods, often spanning periods of centuries before reformation occurred, so I don’t think medieval is a good definition as much as “accepted” or “traditional” methods might be because ID is a relatively new process of design compared to something like blacksmithing or roof thatching, ID is hardly a perfected, established process, if it was we would have much more credibility that other established professions enjoy.

Calling it medieval is insulting because it suggests that, as a whole, Industrial Designers are resistant to new tools, processes, or ways of designing. Generalizing, criticizing, and mocking people is hardly a way to inspire them to experiment and change, no matter how empirical your data or how right you may actually be. This may be why your entrepreneurial foray into generative design tools was not a huge commercial success.

Ok I do not intend to mock nor insult, though I find many insulting nature’s design process, which was not understood in medieval days. The monkey + typewriter story is an example. I have deep respect for both nature and science and I would appreciate some respect for those.

Let’s replace medieval design with manual design.

I’m not sure how my Typewriting Monkey Theorem would have offended anyone, I used that as an example in the context of handing a generative design tool to a non-designer, something that some people hear fear. I was also making a point about intent and it’s influence in using any tool. Earlier in the discussion some were talking about design tools in the hands of non-designers and I agree; garbage in, garbage out. Just like a typewriter in the hands of a monkey, but maybe that species will be able to write soliloquy’s someday, maybe not, evolution will tell.

SK Wrote
The answerer is simple. You don’t even try. The complexity of it is beyond the reach of current computational process. Generative processes is a dumb assed process but an immensely powerful one. It can create great results, as nature does; provided the initial models are built sensibly (with some design intent) and selection is made intelligently. So, finally its about choice. It’s designing by selection.

In your above reply, what stands out are this words,…selection is made intelligently. So, finally its about choice. It’s designing by selection. This words suit gamblers, non creative people and to those grey market manufacturers(Duplicate product manufacturers). Industrial Design is a serious field and so is the work of designer’s, where design or designing is not on the basis of intelligent selection, choice and design by selection. Creativity, imagination and vision is what designer’s ingenuity, compared to any other field or people.

SK Wrote
I certainly seem more interest 5 years latter. I guess, man have heard about it now. But the ship has left the port for IDers.

This is what alienated most of the people including myself. Industrial Design and IDers role is very different which people need to look much closely.
Yes, I have mentioned before and many others have also said, Generative design(Curious tool) will be useful to designer’s in every aspect and vice versa, but it will be only designer’s ingenuity that can find meaning in those hundreds of design generated by generative design. Ship will never leave port for IDers or maybe people have become too curious

I’ve been following this for a while and really surprised what I’m hearing. To me, it’s pretty simple.

1. I wouldn’t call “Generative Design”, Design. Engineering, perhaps. What it is in fact, is just the solving of a Linear Dynamical System of an n x m matrix, where the columns are n, and the rows are m, each column relating to an input and the rows relating to an outcome. I’ve been listening to a Stanford lecture on the subject recently, and while I’m no expert, this is exactly what it sound like to me. This method is very common in engineering, where you know the variables (ie. figuring out a launch trajectory for a rocket where you can control several factors - thrust, angle, weight, trim, etc.).

Design is about solving those problems where you don’t have all the variables or can’t quantify them.

1. As such, in real-world design, I can’t see much use for this method. Very often, as has been mentioned, there are not only intangible factors (prettiness, style, modernism, etc.) but also the plain fact that even with the tangibles, not all solutions are optimal as in engineering (though in engineering, there is a way to weight the value/cost of each input and calculate more desirable outcome, for example if weight has a cost X in fuel, or size has a cost in materials).

2. In most real-world design, even if you were doing something basic, such as the garbage can or toy figure previously mentioned, by the sounds of it, even inputing these things is a slow, complex process. If as the previous posted mentioned, it takes two weeks to input such variables into a program, to me it sounds like a huge waste.

This is of course not to mention that for a good design, there are likely 1000s of individual variables that should be taken into account. A good designer can do so, placing different emphasis on different variables through experience and judgement.

1. Not to mention a good designer will not be creating all the useless variations that someone needs to weed through to find something decent. A real challenge would be having the generative method create 20 random designs and put those against 20 designs from a designer. I can pretty much guarantee that if scored (by a designer or consumer) that the designer’s solutions would taken at a sum far outweigh the random.

It would take you 2 weeks to program and create CAD variations? I can sketch 1000 concepts before you even start generating one.

1. Bottom line, apart from perhaps some random patterns and such, I see no value in this method. Still, this is not really design, and even being compared to CAD is not fair. CAD is a tool, generative methods are a process. Big difference and obviously since the OP is not a designer, something I can see how it could be confused. Same way so many students think CAD is a process, rather than a tool and make pretty renderings of bad designs.

R

Selection is not gambling. It is a critical part of the manual design process. Designers always come out with multiple design possibilities before they narrow down to one. This happens thorough the design process at every level of detail and the final outcome is very much depended on these selections. Don’t see the gambling element here.

As far as companies are concerned, they pay designers to come up with designs that people are likely to buy. Their own existence depends on it. How the designers come up with the solution, they do not really care. They would certainly welcome the opportunity to select possibly through consumer testing designs that are likely to succeed, without shooting in the dark according to the hunches of the designer - which is exceedingly risky.

This stuff that you describe happens more than 10,000 times every second in your PC and has no connection to curious tools.

Curious tools are primarily for intangible complex issues that are beyond the capability of computational processes. There is plenty of other software to deal with the rest. No software that I know of is written to give human being a chance to calculate because it can do some calculations too. If a human being is involved, its primarily because that involvement is necessary.

Did not say it takes weeks. Depends very much on the complexity. Running or the program takes less than 10 minutes. But then you need a properly structured CAD model. So this model building time is related to the CAD model building time. What you need to create is nothing more than a CAD model, but structured in way that it can be parametrically varied. I see that you design shoes. Once basic model is built, parts of it (that are common to all shoes) can be re-used many times allowing you to create many different designs. In a way this method forces efficiency on you as you need to build variable models instead of static models. So, the overall time that you spent on design will significantly reduce specially in cases where the product type remains the same.

It would be best to verify this. Would you like to participate in the planned “Manual method vs Curios Tools” contest ?

Else, you can try using Curious tools yourself and make your own conclusion. I would be happy to help you try it out, I am sure that you will be able to create some interesting shoe designs.

SK, do you have any complete concept to market case studies using GD for product design?