Ask Slashdot: Could An AI Conceivably Create Futureproof Product Designs?

dryriver writes: Whether you are into consumer electronics, cars, furniture or other manufactured things, one aspect of them doesn't change -- the physical design or "look" of the product tends to age badly in our perception as newer products are released. When you first buy the product it looks "sexy and new"; 5 years down the road, it just looks kind of "old" or "less sophisticated" compared to the newer, sleeker products. To the question: Could you get an artificial intelligence powered by a neural network to train on hundreds of product designs created over the last 20 years -- possibly by laser-scanning products in 3D or providing 3D CAD files -- and learn with great sophistication how product design or "product looks" evolve as time passes? Could that AI then be coaxed into making fairly educated guesses about how a particular product might look if it were designed in the future, in say 2030? In other words, could a suitably trained AI give a laptop, car, or designer chair to be manufactured in 2020 the "design look of the 2030s" ten years early by extrapolating forward from the training dataset of past product designs?

Comments

Still would get a Yugo

[PKI Champion]

AI still would have given us the Yugo. Sometimes, you just don't produce something good. AI will not change that.

Re: Still would get a Yugo

[Mr. Dollar Ton]

Hello, AI, which foul mouth has trained your instance for today?

Re:Still would get a Yugo

[AHuxley]

AI might feel like giving humans a Melkus like car design.
With advanced 3-cylinder engine so human workers have less engine work to do.
https://en.wikipedia.org/wiki/...

Re: Still would get a Yugo

[Shotgun]

Creating a timeless design that never needs updating is generally not in a businesses' best interest.

You mean like the typical blue suit?

Re:Still would get a Yugo

[RespekMyAthorati]

Or even worse, a Pontiac Aztec.

Re:Still would get a Yugo

[toddestan]

The Aztec's biggest sin really was just being ahead of it's time. It was pretty far out there for 2000, but if it came out today no one would give it a second look among all the other bizarre and ugly designs that are currently being made.

Can a person do it?

[goombah99]

People draw what they thing things will look like in the future all the time.
therefore an AI can too. Doesn't matter where you the the "I".

Of course a dumb AI will extrapolate all products that get smaller with time to a point.

Re:Can a person do it?

[toddestan]

A big part of the lower, wider, and longer thing was not just aesthetics, but because such a thing was finally possible. In the early days, cars had to be built to handle roads that were little more than dirt tracks, because a lot of roads were just dirt tracks. Even into the early 1950's it was still expected that a car would travel significant amounts of miles on dirt and gravel roads. It wasn't until the Interstate system was built that a car could now spend almost all its time on straight, smooth, flat, paved roads. And the result was manufacturers built cars that were designed for cruising all day at highway speed.

Predicting what we have today would mean having to predict the CAFE rules that favor light trucks, which would result in more and more light trucks being sold over cars, and since people generally don't like driving an actual truck, that the end result would be car-like trucks that just truck-like enough to fall under the CAFE rules for light trucks but otherwise are as car-like as possible. Hence the gazillions of cookie-cutter CUVs we seem to be stuck with.

Is this the trend, then?

[DulcetTone]

Do we all just assume AI can do what mankind cannot?

Re:Is this the trend, then?

[HarrySquatter]

dryriver seems to think Slashdot is his personal blog to post the idiocy he comes up when waking and baking.

Re:Is this the trend, then?

[dkman]

And why would business want it?

Business wants to sell you the "new shiny" every year or two.

They no longer want to build things to last. It would be good for the planet if we built things to last, but we can't seem to care enough about that.

Doomed to failure

[alvinrod]

It doesn't matter what designs a person or an AI comes up with, because people will always want something new. Just like every generation goes on to develop new music or new movies even though it's a lot of the same chord progressions or plot lines, it's nevertheless something new. It doesn't matter if you extrapolate further ahead and give us a design ten years before its time. We'll still be tired of it five years down the road regardless of when the countdown starts.

Re: Doomed to failure

[phantomfive]

Indeed, how do you predict fads?

Re:Doomed to failure

[Ormy]

Exactly what I was going to post. How is this not obvious to everyone?

Re:Doomed to failure

[ranton]

It doesn't matter what designs a person or an AI comes up with, because people will always want something new.

In addition to the novelty factor, new products also depend on new scientific advances such as in materials science. Without today's batteries, LCD screens, processors, etc. you could not build a modern smartphone in 1990. The same goes for cars, TVs, refrigerators, etc.

If you look at things with little to no scientific advances, there is less need for new products. The wood furniture in my house does not look that much different than what my parents had, and in many cases were picked up at high end estate sales so they really are decades old. But then again there is still the novelty factor for many/most people, which creates fads in probably every industry.

Re:Doomed to failure

[Matheus]

...but of course AI is also going to lead to the singularity so we won't have to wait for those advances we'll (or at least the AI) will have all of those advances at its disposal now! ;)

anyway.. The root of the article is presuming there is a "perfect" design that all designs are incrementally progressing towards. IF that were the case then sure you could "calculate" what that perfect design was and leapfrog to it BUT as many others have said in many other ways: What people want changes unpredictably, even irrationally, forwards, backwards, left, right, up, down at will and fad. No matter how amazing a thing you produce today, someone will want it different tomorrow.

Re: Doomed to failure

[dwpro]

I would imagine the same way we predict other shit with AI: gather up data, refine it, come up with novel features, and then see the predictive capacity. My guess is we could probably predict 90% of the fairly routine iterations of things: smaller, rounder, less features, more cowbell. If we ever get to where we can predict something like a notch I'll be shocked though.

No need for an AI

[Rosco P. Coltrane]

Men can create future-proof designs also. In fact, that's exactly how most everything was designed before planned obsolescence was invented in the beginning of the previous century, and consumers started to get brainwashed into wanting the new model of the year of things they already had in perfect working order by automobile manufacturers.

Re:No need for an AI

[Ormy]

This. If future AI is tasked with product design then code will have to be included to force the AI to produce sub-optimal design (i.e. planned obsolescence). Ofcourse all of the AI's code will be proprietary so the public will never have proof unless there is some whistle-blowing.

Re:No

[Darkling-MHCN]

Maybe it could, by analysing present day trends and extrapolating on them. And who's to say AI won't be responsible for changing them?

Statements about technology that include the words "No" and "Never" have often ended seeming silly in hindsight.

What if AI posted slashdot articles?

[Anonymous Coward]

Could AI vape global warming?

We Can’t Have That

[mschwanke97402]

We have to obsolesce current product on regular basis. New product means a whole new sales cycle. If sales were restricted to replacement of worn out or damaged items sales would be a fraction of what they are now. We can’t have that. Investors won’t stand for it.

I'll go with 'no'

[fuzzyfuzzyfungus]

This question seems to hinge on the (almost certainly false) notion that design, and the perception of design, evolves in some trajectory that's independent of its context.

Things that look old do so in no small part because we can compare them with things that are new. Things that look cutting edge do so in part by rejecting design elements that look familiar.

Even if we assumed a scary good, probably better than human level, bot put to the task of inferring what "2020 thinks 2040 will look like" its output would, immediately, be part of 2020-era design, albeit probably a visually distinct flavor; and what 2040 actually look like would include reactions to, away from, against, with nostalgia for, etc. that "2040" design from 20 years ago.

When humans try this we get zeerust. It's not clear why a bot would do better; or that even an arbitrarily talented bot could beat the fact that the future it predicts will automatically become part of the past that the actual future evolves from(recursion is fun and unproblematic, right?).

There's also the problem, outside of some purely decorative objects or ones that aggressively try to defy the constraints of material culture(either trying to look more futuristic than the tech really is, like sci-fi TV props; or are deliberately throwbacks, like SCA longbows and stuff), that things look the way they do in no small part because of the constraints of technology that no amount of industrial designer resistance can get around.

Re: I'll go with 'no'

[phantomfive]

Tail fins on cars will be coming back next year. I have my Cadillac ready for the moment.

Re:I'll go with 'no'

[jBozfDioucNxjF]

This is the best response I've read to this posting so far. The question of whether we could "future proof" visual trends seems intriguing on the face of it, but I feel it breaks down pretty quickly when we start to really think about it. Personally, I don't see a qualitative difference between what I've heard about AI's development of styles (mainly in game play) and what human designers do. I'll just put that on the table — for most design work, we're producing results that AI will be able to produce, just as well. As with much of what AI seems to offer, the main difference in comparing its work against humans' seems to be in terms of the scale of work-time: AI works much faster, so its body of work grows at a much faster rate than a human-driven one, and trends, progress, etc. emerges much more quickly. Considering what little I've read about what AI has done in the world of the game Go, it sounds like, given a set of constraints (in this case the rules of the game), that AI can "shoot off" on its own tangent, picking up where humans left off and leaving them in the dust. I read an interesting quote from a human Go player (paraphrasing): "Playing agains the AI was like playing games from the future." Since the rules of the game are a constant, this seems feasible. However, with something like industrial design (or graphic, fashion, motion, etc.) where the strength of the product hinges in many ways on how well it relates to ideas and situations that occur in cultural contexts, it's not only unlikely that future-proofing would be achievable, it's impossible. Kind of like: Wherever you go, there you are. Design is engaged in a give-and-take with culture, and its product sets the stage for what comes next. So if AI can figure out a new style right now, even if this is based on a deep analysis of past circumstances, that product will itself change the critera for success of future products. What has changed is not the fact of things falling out of fashion, but the rate of development required to keep up, and of course, the rate at which this fashion fallout occurs. So essentially, it could end up being quite the amplifier of obsolesence, rather than its solution. What I like about this question, though, is that it brings up the fact that it's certain that AI will have an *enormous* influence in developing new styles and approaches, not just in technology and producability, but in ideas and meaning as well. This will occur to the point that it may be rather questionable whether it remains common for humans to engage in such endeavours. If we assume that widely available design AI will achieve the same proficiency as a human, but can do it faster and without any of what we might call the "overhead of life" (things like human rights, entertainment, 2/3 of the work day, 2 weeks a year, the retirement years — all devoted to things other than work, etc.), then what point is there to being a human designer? Yet several potential routes to a level playing field do come to mind For example, authenticity. What if AI were to become very adept at translating the zeitgeist in formal fashion, to the point that it can produce intriguing designs at the same level of human-powered systems? And let's also say that the AI is simple, in that there it doesn't embody any type of attempt to emulate the human experience of the cultural criteria to which it responds, but instead simply emulates the results (for example, it can read and form associations but it does not express a preference). Is the work as valid? My guess is that for mass culture, the question won't even come up (the answer is, flatly, "yes"). But for experiences where vintage, provenance, history and knowing are important, the work of such a system may not be intriguing at all. At least not to humans (see preference caveat above). The consequences of this are far-ranging and incomprehensible from our current perspective, and this is only what we can see. When thinking about this topic, I'm often reminded of the "smart phone surprise" that we experienced recent

Re:I'll go with 'no'

[RespekMyAthorati]

Hint: use

<br>

to make a newline on /.

AI is a path and not a goal.

[JimSadler]

The path of AI is fascinating. It already is giving us next years products today and that will get stronger and stronger. But there is no ultimate quality to a product. Although difficult better product can always be made. Sometimes game changers are not yet dreamed of. For example graphene and carbon fiber products are in their infancy and 15 years ago would not even be a question in one's mind. For example you might be producing the best canoe ever made but if carbon fiber is in your hands you can cut the weight of your canoe, make it difficult to damage and totally age resistant. But often it is the human will that fails and not the advancing technology. For example a row boat that is quite superior can be made from the same plastic that the common milk jug is made of yet almost no one has chose to do so. A very, very durable row boat that is lite in weight can be made from this type of plastic. Yet it just is not happening.

The state of AI

[elrous0]

Maybe for an encore they can design some videogame AI companions who don't stand out in the open and get themselves shot like fucking idiots.

Re:The state of AI

[drinkypoo]

I'd be happy if Bethesda could just get them not to put their head into my line of fire and stop there.

Re:The state of AI

[sheramil]

Maybe for an encore they can design some videogame AI companions who don't stand out in the open and get themselves shot like fucking idiots.

But you don't want an AI companion in a game who's better than you.

"A good robot learns to read his owner's mind a little, to anticipate little wishes before they become commands. Naturally there's a limit. Too much anticipation scares people just as too much grinning and bowing does. Moderation is the key. Aim to be a smidgen less intelligent than your owner, but a lot more thoughtful. See everything as it affects your owner, and in no other way."

- John Sladek, "Tik-Tok"

Probably not

[SWPadnos]

I'd say it's a hard problem.

The visual design (shape) isn't the only factor for acceptance. It also has to do with features, price, popularity (yes, that's a chicken-and-egg thing), and how contemporary products compete on all those points.

Not only do you have the issue that new things become old (as others have pointed out), but there's also some resistance to things that look too different. Somewhat different is good, very different can be considered weird.

I think there's too much of a fad aspect to be able to predict much. It's a chaotic system.

No

[Anonymous Coward]

And this is a stupid question. Apparently just add "Could an AI..." to anything moronic and it becomes news.

"The AI" is just a bunch of statistical methods

[Mr. Dollar Ton]

that compute averages, which are plugged as coefficients in some equation derived from some theory created by a human brain. "The AI" cannot "create" or "destroy", it can only process the garbage in to produce garbage out. It will not become sentient, and it will not start a war with the humanity, these scenarios are also coming from books created by humans, in which they are a plot device that emphasizes some human trait or other.

Please stop anthropomorphising shit for no reason at all.

Re:"The AI" is just a bunch of statistical methods

[Mr. Dollar Ton]

That thing is a /. editor, so it is probably just a buggy bash script.

Re:"The AI" is just a bunch of statistical methods

[Mr. Dollar Ton]

You make inferences about the whole population from the value of a sample size of one - your own. It is a very, very poor estimator.

The simple and direct honest answer.

[3seas]

No!

No such thing as 'future proof', so no.

[Rick Schumann]

There's no such thing as 'future proof' just like there's no such thing as 'a standard' -- or at the very least a 'standard' only exists until it becomes inconvenient to someone, who them breaks the standard to accomplish what they want, then so much for the 'standard' being a 'standard' anymore.

Also, so-called 'AI' is completely incapable of one vital ingredient: human creativity. Very often human ingenuity and creativity has no rhyme or reason to it, it's seemingly random, and very often astonishing, and that's one of the qualities that defines us. No 'AI' is really capable of this, so far as I've ever seen; all attempts at 'AI art', for instance, are just hideously derivative.

I don't know about anyone else, but I'm getting more than a little sick of hearing about 'AI' this and 'AI' that, when this half-assed excuse for 'AI' that keeps being trotted out to us like it's something New and Fresh and Innovative is really Old and Busted and really not very good. 'AI', as it currently sits, is 20% fat content ground round (cooked well, to the point of being cardboard), versus the human brains' Filet Mignon (medium rare, of course). They'll both feed you, but honestly which would you rather have? I'll get excited over 'AI' when we can figure out how the little things like 'consciousness' and 'cognition' and 'personality' actually work, so we can create analogues of that in hardware and actually be able to converse in a very real freeform way with a real AI. What we have currently isn't even as smart as an amoeba.

Re:No such thing as 'future proof', so no.

[Visarga]

> Also, so called 'AI' is completely incapable of one vital ingredient: human creativity.

Human creativity belongs to humans by definition. But AI creativity is not limited to human creativity. We have AIs that can imagine faces and scenes, invent new strategies in games and invent proteins to fit a specific function. AI's can even design deep neural nets to the same level of accuracy as the best human researchers, given enough compute. Creativity is just random search filtered through utility.

Re:No such thing as 'future proof', so no.

[Rick Schumann]

Apparently nobody wants to listen to your piss and vinegar rantings any more than I do otherwise you wouldn't be forced to post as an AC, you jackass. Did Slashdot ban you? Is that why you're posting as an AC? Or do you just not want to be accountable for the raw sewage that spews forth from your mouth?

Re:No such thing as 'future proof', so no.

[Rick Schumann]

Creativity is just random search filtered through utility.
Prove it. Rhetorical; you can't.

Re:No such thing as 'future proof', so no.

[TuringTest]

Of course AI has creativity. In recent years we have seen an explosion of neural networks that generate images and videos from learned examples, and which are able to create novel imagery (original realistic faces, streets populated with objects, fill-in gaps in broken images...), with any novel combination of visual styles. If you ask me, that doesn't look much different from what many artists do, even in modern art.

(The most sophisticated networks are even capable of exploring the creation space and produce works of art with large variability by randomly changing their input parameters.)

What AI utterly lacks is discernment. To the algorithm, all its products have identical value; AI is incapable of separating the wheat from the chaff. Even adversary networks are only capable of discarding examples that do not fit a previous established style, much as conservative spectators do for artists.

The unique ability of the artist is being able to tell that one specific instance is worth more than the others.

This is done by finding a meaning for the specific feature of that image which are not present in the others, i.e. being able to tell a story on how that image relates to the general culture, and convincing the spectators that this meaning is valuable. Art is thus not created by creatively generating unique random objects, but by recognizing and nurturing the value they might possess.

I won't be the one to say that AI will never be capable of such discernment, but current techniques are unable to understand culture and assess how the meaning of images relate to their environment; so that final part is indeed missing from its capabilities.

Re:No such thing as 'future proof', so no.

[Rick Schumann]

I don't have to prove any such thing, and the guy I responded to orginally is making claims about how the human mind works that cannot be proven one way or another. Also since you won't even be bothered to use a NAME behind your shitty comment I see no reason to do a damned thing for you except tell you to fuck off, troll.

Re:No such thing as 'future proof', so no.

[Rick Schumann]

Stop anthropomorphizing shitty software programs, you're embarassing yourself.

Re:No such thing as 'future proof', so no.

[TuringTest]

"Object X is able to create a product of type Y automatically" is anthropomorphizing now?

Re:No such thing as 'future proof', so no.

[RespekMyAthorati]

Anymore than you could prove the opposite.
Nobody has a handle on what creativity, imagination or consciousness are.

Re:No such thing as 'future proof', so no.

[RespekMyAthorati]

Stop anthropomorphizing Rich Shumann.
It is obviously a bot programmed to make us complacent about the upcoming robot apocalypse.

Re:No such thing as 'future proof', so no.

[TuringTest]

Yeah, the first part I had already figured out :-)

The future is intrinsically unknowable

[mark-t]

The existence of the halting problem affirms that even in an entirely deterministic system, one can fabricate an outcome that no amount of cognition within that system can predict with accuracy, effectively making it non-deterministic.

So, no... it will not. At best, it may end up being statistically better at it than humans are, but in the end, it will still just be guesswork.

Re:The future is intrinsically unknowable

[RespekMyAthorati]

The existence of the halting problem affirms that even in an entirely deterministic system, one can fabricate
an outcome that no amount of cognition within that system can predict with accuracy, effectively making it non-deterministic.

And how does that apply to a computer, but not to a human?
The halting problem applies to all systems, natural or artificial.

AI != Magic

[eatvegetables]

Really!

Re:AI != Magic

[TeknoHog]

Any sufficiently advanced technology is indistinguishable from magic. Though I guess Clarke was referring to potential alien technology that was developed completely detached from ours.

Perhaps post-singularity "AI" will look completely alien to us, but that points to another problem: Current "AI" is a buzzword that has little to do with the "magic" of the original meaning, much like "hoverboard" and "android". So I would say no to anything promised by "AI", but I wouldn't rule out "things done on computers" in general.

Re:AI != Magic

[fish_in_the_c]

To our ancestors traveling at 70 mph would have seemed like magic.
However, AI != Omniscience and never will.
The predictive power of your model will always be limited by the amount of data you have , which will be limited by your ability to predict it. For any given topic AI might be able to make 'better' predictions by creating a better model, but even then they would be 'within a range of possibilities' not with absolute certainty.

Re:No

[drinkypoo]

This is a variation of the question about god and big rocks.

You can't future proof as long as there are new discoveries to be made. You can't predict scientific discovery, by definition. QED, you can't future proof.

You could extend product lifespans by forcing manufacturers to offer a transferable warranty which lasted a reasonable period. Way more than 90 days, that's for damned sure.

What is the motivation of this question?

[shanen]

Or perhaps the deeper question is why the Slashdot editors thought anyone would be interested. Let me strain my brain a bit...

(1) Future-proof software. It's called upgrades.

(2) Future-proof goods and services. It's called people, as in people don't actually change that much.

(3) Future-proof fashions. Only if advertising is outlawed.

Question over-answered.

AI can create stylish design, by not doing that

[raymorris]

Almost all of the comments on this article say "no".
So I'll play Devil's advocate and say, in particular way.

Picture a house from the 1970s. Laughable style, right?

Yet Mount Vernon is stylish almost 300 years after it was built, and always has been. The Supreme Court building is impressive architecture almost two THOUSAND years after the design was first built. ( https://en.m.wikipedia.org/wik... )

In other areas of fashion and style, Levi's 501 jeans look as good today as they did 150 years ago. The two-button dark navy suit is timeless, and the Burberry trench coat.

A 20-year old Ford F-150 doesn't look outdated at all. Even a 30-year old model doesn't look particularly out of place today, it simply looks like a truck.

AI could very easily figure "these designs have been stylish for 200 years, they just might remain stylish for another 10 or 20 years".

PhantomFive asked "Indeed, how do you predict fads?" Perhaps you choose *style* over *fads*. We've all seen people ask questions like:

How do I use wget to read and reply to my daily email?

The answer, of course, is that there are plenty of very good ways to read, sort, and reply to email. Wget isn't one of them.

"How can you predict fads" [in order to look stylish in 10 years], is not the question, perhaps. Perhaps, in order to look stylish in 10 years, pay no attention to trends. Look stylish by being stylish, not by monkeying a tad, current or predicted.

Having said all of that, it does occur to me that several of today's 14-year old pop stars will be 19 year old pop stars in five years, when their fans are 19 and spending money at the mall. The best / most popular will still be popular in 10 years, when their fans, now in the mid-twenties, will have developed had a lot of their style tastes that they'll carry as a base for their aesthetic for decades after. Looking at the styles popular with 12-15 year olds MIGHT give some good hints about the styles those kids will still like ten years from now, after removing certain things like any particular bold color that is currently popular with kids. Particular bold colors come and go, of course.

Re:AI can create stylish design, by not doing that

[serviscope_minor]

Picture a house from the 1970s. Laughable style, right?

No one's disputing that old stuff can look good or be useful. Second when new houses are build they're not built to look like houses from 1970 even though we're perfectly capable of doing that.

Yet Mount Vernon is stylish almost 300 years after it was built, and always has been.

Indeed some things are timeless. Timeless enough to replicate.

For example the mid century modern style of furniture is still in production today. You have to sometimes look quite hard to tell the difference between a new piece and something made in he 1950s.

No one's busy replicating 1990s style though.

A 20-year old Ford F-150 doesn't look outdated at all

I think you've settled on a trend here though with the houses and cars: they're not changed very often. Houses just last and cars now seem to last a lot longer than they used to (blinker fluid now lasts 10,000 miles!). That's going to slow down the pace of change. People want something that looks new, not strange. So a new thing can only look a little bit beyond what's out there. Combine that with a slow turnover and you get slowing change.

AI could very easily figure "these designs have been stylish for 200 years, they just might remain stylish for another 10 or 20 years".

Sure.

But for all the Levis 501s out there, there are a thousand moon boots. I think the idea is not to spot existing levis 501s, but to make the next ones.

Getting to that though it just seems like another AI fantasy thread.

Unless we use deep learning in the cloud via a REST service running on Node.JS. Then it'll be solved.

Re:AI can create stylish design, by not doing that

[JaredOfEuropa]

How much of can be attributed to reputation rather than tasteful design? For something to become a design icon, it does need to be designed well, but beyond that? According to that article, the design of the Parthenon was most unusual (perhaps even unique) in Roman times, but became an icon of Roman design as one of the best preserved buildings from that time. Mount Vernon is a well designed, well proportioned house of a type that has always been seen as stylish, because it is the kind of house a rich man with taste would build, it's rather unassuming... but you need to be wealthy to afford it. Such houses (in a similar style but a lot smaller) have become very popular to show wealth without being ostentatious. And there's no doubt that 501s are good jeans, but isn't the name a big part of that? Do people even recognize them when they try them on (without looking at the label)?

So, you are right that an AI might very will pick out the defining elements of those design classics. But I think it will then simply have come up with "good design" rather than "lasting style". Because while good design is almost a prerequisite for something to become a design classic, it is not enough. Luck plays a big part as well, there are plenty of examples of great designs that still were forgotten at some point (and are sometimes rediscovered).

Re:AI can create stylish design, by not doing that

[phantomfive]

Create the styles, don't follow them.

But can an AI predict the weather a year from now?

[Visarga]

Fashion is a chaotic process like weather. If we can't predict the weather more than 10 days ahead, why would we be able to predict fashion?

No, because

[CptLoRes]

Design has long since stopped being about functionality, and is now more about the latest trend. And trends aren't logical, they are driven by marketing to increase sales.

Can AI solve the halting problem? no

[locater16]

The question is literally "Can AI solve the halting problem?"

This is because, wouldn't any hypothetical "futureproof" product then be surpassed by the AI working more to create an even more futureproof product? At what point do you stop, do you ever stop?

Thus, the question becomes "can AI solve the halting problem" and the answer is, no you dumbass, the halting problem is undecidable and is the very math problem that helped create the modern notion of the computer in the first place. The problem that answered "no you can't solve this no turing machine can."

Re:Can AI solve the halting problem? no

[RespekMyAthorati]

The question is literally "Can AI solve the halting problem?"

Nonsense.
The halting problem concerns finding a general solution to the question "will this program ever halt?" that always works.
All the proposed fashion-prediction AI has to do is come up with statistical
fashion forecasts that are better than most human marketing managers.

a neural network is not Intelligent

[GodWasAnAlien]

a neural network is intelligent the way that a calculator is intelligent or a sorting algorithm is intelligent.

When AI comes, our perception of products is not relevant, as they will make stuff for themselves.

Hopefully the AI will be our friends.

The current mis-use of the AI to mean something other than passing the Turing test
just assumes that AI will never come, and we will just have neural networks coming up with efficient solutions forever.

AI can only come from an self adaptive system, not the neural network in box special purpose systems we have now.
A milion monkeys typing on keyboards, even if they know how to create neural network software, will not create "AI".

Separation of concerns, please

[diakula]

Separate function from outlook. Make the outlook like a skin that you can change. Make economy around skins that appease. Engineering underneath just works, for 100 years :D - I wish...

no way

[samantha]

Obviously not. Most likely in 10, 20 years we will not even have the same products. Think how many products were replaced by the smart phone. If in 10, 20 years we have working neural lace then smartphones themselves would likely no longer be relevant. Also think how much cell phones changed from funky bricks to today's smartphone. No way looking at designs up to the funky bricks could anything, no matter how smart make a future proof brick that was relevant.

AI is not magic. There is no magic.

It's a recurring trend

[tofus]

1950s: Microelectronics will solve everything!
1970s: Computers will solve everything!
1990s: The Internet will solve everything!
2010s: Big Data will solve everything!
2020s: Artificial Intelligence will solve everything!
2030s: Quantum Computing will solve everything!

My prediction for the 2040s: There will still be plenty shit left unsolved.

Re:No

[Anne Thwacks]

Nothing AI can do, but this is a recent phenomenon, mostly created by American car manufacturers.

To overcome the very limited life of the product, they created a new product each year and a fashionista mentality that you change your car like you change your clothes.

In Europe, cars were being made that last 30-100 years. However, American cars, devoid of quality control, were much cheaper, and as the market expanded rapidly, most buyers, even in Europe, were people who had never owned a car before, and had no idea of the difference in quality between a Packhard or Bentley and a Ford.

The rest is <strike> a total fuckup </strike> history.

Re:Facepalm

[Anonymous Coward]

You nailed it.
Stuff looks old because it has been around, not because it was designed a certain year.

Sufficiently good AI may be able to design something that looks good today, like a star designer. It will still age. Some things looks good even when old, but usually because it is a valuable item. Grandfathers gold watch, or a perfectly kept veteran car.

Perhaps an AI can become so advanced, it accurately predicts the design style 20 years into the future. So if you make such an item and keep it secret, it will then seem modern when you unveil it in 20 years. But if you mass produce and sell this thing, it will be old 20 years from now. By using the predicted future design, you turned it into "todays design" and therefore changed what the future would evolve to. (And if you re-ran the fantastic AI predictor, informing it that you commercialized the future design - it would indeed confirm that the future styling will now be different.)

This is the problem with all future predictions; if you act on them, you change the future. Then you need a new predicition that takes into account just how you acted on the predicition. But if you act on that . . . The plot of many a scifi.

Re:Shoes

[Anne Thwacks]

Can a woman buy just one pair of functional shoes? No.

That is why God created "a woman's right to shoes".

What is the question?

[TJHook3r]

AI could conceivably come up with an optimal design for a product but the author seems to want something else - a subjective 'beauty'? It's unlikely that a computer would design a commuting car that is not aerodynamic for example, unless a human adjusted some weightings somewhere... wonder what that slider would say? 'Pimp Factor'?

Future = Past + Present

[JohnJohnsonOkah]

Yesterday gave birth to Today and Today will give birth to Tomorrow, not next-tomorrow. So, if AI creates a seemingly next-tomorrow, that next-tomorrow is automatically seen as tomorrow and this leads leaves next-tomorrow still blank. NERDS WILL UNDERSTAND this simple logic. Therefore the answer is 'It's Impossible'.

THEY DON'T EXIST

[argStyopa]

How about you actually create an AI, and then get back to me?

"AI" as touted today everywhere doesn't exist! Say it again together: THERE IS NO ARTIFICIAL INTELLIGENCE. Complex if/then loops with dynamically corrected weighting is not the same as actual creative intelligence.

Asking if an AI could do it is semantically equivalent to asking "Could Unicorns do it?" because both of them are today imaginary things.

Wat utter nonsense

[gweihir]

"AI", what we have of it (which does really not deserve the name) cannot do better than smart humans when judgements are required. In fact, it does much worse. Stop with the quasi-religious belief in "AI"already. The only thing automation (which is all we have) can do is make standardized decisions that are significantly worse than those of a smart human, but do so much faster and much cheaper. For many things, this is enough. But it does in no way allow a better quality of the decisions or new insights.

I Must Already Be Future-Proofed

[careysub]

Reading the summary about how designs look dated after a few years, to me it is like reading about a theological argument within a religion I never heard of. The claims made are completely alien to me, have no reflection in my own experience. I don't think products a few years old look dated. I don't think newer products look sleeker or more sophisticated. Often I think they look stupid and annoying. Can AI make that stop?

AI lacks INTEGRITY

[ElitistWhiner]

Design with integrity is the intangible underlying success

The chain is eternal

[Orgasmatron]

Fashion integrates the past. The 2020 design incorporates the 2019 design, the 2018 design, the 2017 design, etc...

If we get the 2030` design in 2020, then we will get the 2040` design in 2021, etc. By the time 2030 gets here, no one would ever actually design the 2030` design because the 2029 design, the 2028 design, etc were changed. The actual 2030 design will bear no resemblance to the old 2030` design.

(Those ticks should represent "prime", which I don't think I can do properly here on slashdot.)

Re:Talk about future proof

[crunchygranola]

Even more remarkable about the B-52: the program to build it started in 1947 and it first flew in 1952. It is expected to remain in service until at least 2050 at which time the design will be more than a century old and the youngest airframes would be 88 years old, with others being over 90.

Betteridge's law of headlines

[luis_a_espinal]

Could An AI Conceivably Create Futureproof Product Designs?

My CS is rusty, but I'm sure this is homomorphic to the problem of creating a program that can tell if another program is going to halt with the right value (ergo, mathematifucking impossible.)

But beyond that, in the general sense, headlines starting with a question must be answered with a categorical "no." Damn writers need to do a bit better with their homework.

maybe AI can decide the perfect hem length as well

[mikeabbott420]

Fashion is about change for changes sake. It's not about finding an optimal solution, it's about showing you have the latest thing. It's less about functionality and more about novelty.

Basically, 'Yes', but an uninteresting 'Yes'

[scorp1us]

So NN AIs, just reduce predictive error over the data set. It doesn't create anything novel, it just learns to mimic. It can't invent Van Gogh's style, but it can mimic it and apply it to something else.

So if you feed in all the things from the past and feed them the modern version in, you'll get the meta narratives:
1. More use of plastic.
2. Physically smaller, lighter, less material used.
3. More fragility as designs get tuned to minimize in-warranty repairs and maximize out-of-warranty issues.
4. Increased ease of manufacture & higher yield rates

These are things you already expect. Only that last one, "ease", is the one where magic happens. Since future manufacturing techniques limit what we produce today. It's not really predictable though, so it's where the "magic" happens. But that does follow the same meta narratives.

AI Can Improve Software Design

[northerner]

One area where AI can shine is software development. Human intuition and guidance is needed to design a really useful application. However, the thousands of things that need to be just right under the hood for the application to be bug free, secure, and bulletproof could be effectively managed by an AI. Fantastic applications that have been around for decades still have faults and vulnerabilities unearthed because the humans can't rigorously think through all the interactions and idiosyncrasies. An AI can meticulously go through the code in a way that would take a human-only team a long long time. How great would (insert favorite OS here) be if the security and usability features were consistently baked in at the deepest levels?

In those "predictions of the future" ...

[CaptainDork]

... depictions, the tall buildings, flying cars, and "swooshing" doors are all stupid-looking.

No.

[fish_in_the_c]

No amount of intelligence can solve an unsolvable problem.
No amount of data will allow you to predict the future.
The best model in the world breaks down quickly because of the laws of mathematics that are laid out in Chaos theory.
      "Small variances that cannot be accounted for make huge changes in actual results over time."

It is the same as predicting the weather. The only way to have enough data to predict the weather reliably for more then a few weeks out is to know the position , speed and temperature of all the atmospheric gases and their influencer's ( aka every object on the earth) from the current time until the time you want to predict. In other words omniscience.

If you want to understand look up hurricane modeling. Is there room for AI to help improve some of these predictions? I don't see why not, but you are limited in your predictions by the account and accuracy of your data. Which is in tern limited by your ability to gather and process that data.

Fundamental flaw with this approach.

[Invisible Now]

The assumption that future designs are a learnable linear extrapolation from previous designs is questionable. As designs have occasional generational, non linear discontinuities. And fads and fashion details like color are discontinuous. Think avocado bathroom fixtures.

so, you make it and warehouse it for 10 years?

[RhettLivingston]

If it looks right for 10 years from now, it isn't going to look right now. The product would fail. Many products fail because they are ahead of their time.

Re: AI can create stylish design, by not doing tha

[JaredOfEuropa]

Pantheon! Damn spell checker.

Probably not, and here's why

[jarrowwx]

On the surface, this sounds like a reasonable task to give to an AI. But here's the thing to remember: AI is not magic. It is a nifty trick for finding a mathematical formula for something. If there is no formula underlying it, an AI can't learn it. Hence, no one can use AI to predict the lottery numbers.

So, is there a mathematical formula for popular whim? Well, there is a formula for the way the brain works, so in theory, it should be possible. However, when you look at what goes into that formula, the odds of pulling off what you are proposing drops dramatically.

What is popular rides a knife's edge between what is familiar and what is novel. If it is too familiar, it is boring. If it is too novel, it is scary. It has to find the balance in order to be popular. But there's the rub: What is familiar is constantly changing. That's also why "what is old is new again" happens: because the key elements have not been seen in so long that they start to become novel again.

What is being proposed is a little bit like a temporal paradox: The very act of creating a popular product changes the consciousness of the population, rendering their predictions unable to stand the test of time simply because they were seen sooner rather than later.

That's not to say that there isn't value in using machine learning to build models of good design and aesthetics. There most certainly is.

But you probably won't design a product that will still be in vogue 10 years from now, because if the product is too different from today, it won't be popular, and if it is popular, it will change the trajectory of what is popular in 10 years. You can't win. Not in that particular game, anyway.

You CAN, however, use it to identify the elements of good style, and combine it with a generative model to create designs. And there is a way of probing a model to create novelty, which can be used to explore the solution space. That could be a very profitable way to generate an endless supply of "fresh, desirable designs."

Re:Facepalm

[Shotgun]

I find that Microsoft's latest desktop looks a helluva lot like DesqView.

Progress

[Shotgun]

I doubt this will work, but it may actually advance the field of AI in an interesting way.

One way to describe an intelligent person is to consider if they can look at a situation and predict the outcome. But, there are multiple ways to come to that conclusion. A person can look at socialism and determine that the end result is totalitarianism. Was the determination arrived at by looking at the history of socialist attempts. Or, did the person construct a logical model of human interactions and see the logical conclusion?

It is the second method that is more robust and would be described as intelligent or wise, vs the former which would be considered "educated"(?)/"trained"(?). There is no indication that it does, but I would think that the AI would have to be trained on the merging of trends instead of dictionary data, which would be a great advancement in all the AI I've ever seen.

Nobody can do it

[Roger W Moore]

Nobody, person or AI, can do it. Let's, for the sake of argument, say that someone does make an AI capable of accurately predicting future styles. They then use this in 2020 to release a design which is from 2030. The mere act of doing this will then make others copy this new style so the effect is that the 2030 designs will be introduced in the early 2020's. Hence by 2030 things will be different to what was predicted back in 2020 because by releasing the 2030 designs earlier you have changed how styles will evolve.

In addition to this, there are non-evolutionary changes that cannot be predicted because there is no pattern to extrapolate. For example, it is very unlikely that anyone extrapolating designs of mobile phones prior to 2007 would have accurately predicted what a phone today would have looked like because in 2007 the iPhone came out and everyone's designs changed.

Re: But can an AI predict the weather a year from

[stepho-wrs]

You just described the perfect clothes for Europe and the top half of N.America but certainly not for Australia, South East Asia, Africa or the Middle East.

Re:Don't say 'an AI'

[Jesus H Rolle]

From the U of Helsinki: Because AI is a discipline, you shouldn't say "an AI", just like we don't say "a biology". This point should also be quite clear when you try saying something like "we need more artificial intelligences." That just sounds wrong, doesn't it? (It does to us.)

Why do you quote the University of Helsinki when they're dead wrong? Of course AI is a discipline, and the noun phrase is uncountable when referring to said discipline; however, it also serves as a countable noun meaning "something man-made which has intelligence". When we discuss whether Alexa, Google, etc. are expert systems or artificial intelligences, we're quibbling over definitions, but the grammar itself is fine. Finns should stick to Finnish, and being attractive, and programming AI - hands off my English!