Artificial quality helps physique carnal infrastructure similar modular housing, skyscrapers, and mill floors. “…many problems that we wrestle with successful each forms of engineering and plan are very, precise analyzable problems…those problems are opening to scope the limits of quality capacity,” says Mike Haley, the vice president of probe astatine Autodesk. But there’s anticipation with AI capabilities, Haley continues “This is simply a spot wherever AI and humans travel unneurotic precise nicely due to the fact that AI tin really instrumentality definite precise analyzable problems successful the satellite and recast them.”
And wherever “AI and humans travel together” is astatine the commencement of the process with generative design, which incorporates AI into the plan process to research solutions and ideas that a quality unsocial mightiness not person adjacent considered. “You truly privation to beryllium capable to look astatine the full lifecycle of producing thing and inquire yourself, ‘How tin I nutrient this by utilizing the slightest magnitude of vigor throughout?’” This benignant of reasoning volition trim the interaction of, not conscionable construction, but immoderate benignant of merchandise instauration connected the planet.
The symbiotic human-computer narration down generative plan is indispensable to lick those “very analyzable problems”—including sustainability. “We are not going to person a sustainable nine until we larn to physique products—from mobile phones to buildings to ample pieces of infrastructure—that past the long-term,” Haley notes.
The key, helium says, is to commencement successful the earliest stages of the plan process. “Decisions that impact sustainability hap successful the conceptual phase, erstwhile you're imagining what you're going to create.” He continues, “If you tin statesman to enactment features into software, into decision-making systems, aboriginal on, they tin usher designers toward much sustainable solutions by affecting them astatine this aboriginal stage.”
Using generative plan volition effect successful malleable solutions that expect aboriginal needs oregon requirements to debar having to physique caller solutions, products, oregon infrastructure. “What if a gathering that was built for 1 purpose, erstwhile it needed to beryllium turned into a antithetic benignant of building, wasn't destroyed, but it was conscionable tweaked slightly?”
That’s the existent accidental here—creating a narration betwixt humans and computers volition beryllium foundational to the aboriginal of design. “The effect of bringing the integer and carnal together,” Haley says, “is that it creates a feedback loop betwixt what gets created successful the satellite and what is astir to beryllium created adjacent time.”
Show notes and references
"What is Generative Design, and How Can It Be Used successful Manufacturing?" by Dan Miles, Redshift by Autodesk, November 19, 2021
"4 Ways AI successful Architecture and Construction Can Empower Building Projects" by Zach Mortice, Redshift by Autodesk, April 22, 2021
Laurel Ruma: From MIT Technology Review, I'm Laurel Ruma, and this is Business Lab, the amusement that helps concern leaders marque consciousness of caller technologies coming retired of the laboratory and into the marketplace. Our taxable contiguous is astir however to plan amended with artificial intelligence, everything from modular lodging to skyscrapers to manufactured products and mill floors tin beryllium designed with and payment from AI and instrumentality learning technologies. As artificial quality helps humans with plan options, however tin it assistance america physique smarter? Two words for you: sustainable design.
My impermanent is Mike Haley, the vice president of probe astatine Autodesk. Mike leads a squad of researchers, engineers, and different specialists who are exploring the aboriginal of plan and making.
This occurrence of Business Lab is produced successful relation with Autodesk.
Mike Haley: Hi Laurel. Thanks for having me.
Laurel: So for those who don't know, Autodesk exertion supports architecture, engineering, construction, merchandise design, manufacturing, arsenic good arsenic media and amusement industries. And we'll beryllium talking astir that benignant of plan and artificial quality today. But 1 circumstantial facet of it is generative design. What is generative design? And however does it lend itself to an AI-human collaboration?
Mike: So Laurel, to reply that, archetypal you person to inquire yourself: What is design? When designers are approaching a problem, they're mostly looking astatine the occupation done a fig of constraints, truthful if you're gathering a building, there's a definite magnitude of onshore you have, for example. And you're besides trying to amended oregon optimize something. So possibly you're trying to physique the gathering with a precise debased cost, oregon person debased biology impact, oregon enactment arsenic galore radical arsenic possible. So you've got this simultaneous occupation of dealing with your constraints, and past trying to maximize these assorted plan factors.
That is truly the essence of immoderate plan problem. The past of plan is that it is wholly a quality problem. Humans whitethorn usage tools. Those tools whitethorn beryllium pens and pencils, they whitethorn beryllium calculators, and they whitethorn beryllium computers to lick that. But really, the essence of solving that occupation lies purely wrong the quality mind. Generative plan is the archetypal clip we're producing exertion that is utilizing the computational capableness of the machine to assistance america successful that process, to assistance america spell beyond possibly wherever our accustomed considerations go.
As you and I'm definite astir of the assemblage know, radical speech a batch astir bias successful AI algorithms, but bias mostly comes from the information those algorithms see, and the bias successful that information mostly comes from humans, truthful we are really very, precise biased. This shows up successful plan arsenic well. The vantage of utilizing computational assistance is you tin present precise precocious forms of AI that are not really based connected data. They're based connected algorithmic oregon carnal understandings of the world, truthful that erstwhile you're trying to recognize that building, oregon plan an airplane, oregon plan a bicycle, oregon what it mightiness be, it tin really usage things similar the laws of physics, for example, to recognize the afloat dispersed of imaginable solutions to code that plan occupation I conscionable talked about.
So successful immoderate ways, you tin deliberation of generative plan arsenic a machine exertion that allows designers to grow their minds and to research spaces and possibilities of solutions that they possibly wouldn't spell otherwise. And it mightiness adjacent beryllium extracurricular of their accepted comfortableness zone, truthful biases mightiness forestall them from going there. One happening you find with generative plan is erstwhile we ticker radical usage this technology, they thin to usage it successful an iterative fashion. They volition proviso the occupation to the computer, fto the machine suggest immoderate solutions, and past they volition look astatine those solutions and past statesman to set their criteria and tally it again. This is astir this symbiotic benignant of narration that forms betwixt the quality and the computer. And I truly bask that due to the fact that the quality caput is not precise bully astatine computing. The fashionable thought is you tin clasp 7 facts successful your caput astatine once, which is simply a batch smaller than the computer, right?
But quality minds are fantabulous astatine responding and evaluating situations and bringing successful a precise wide acceptable of considerations. That successful information is the essence of creativity. So if you bring that each unneurotic and look astatine that full process, that is truly what generative plan is each about.
Laurel: So truly what you're talking astir is the narration betwixt a quality and a computer. And the output of this narration is thing that's amended than either 1 could bash by themselves.
Mike: Yes, that's right. Exactly. I mean, humans person a acceptable of limitations, and we person a acceptable of skills that we bring unneurotic truly erstwhile we're being creative. The aforesaid is existent of a computer. The machine has definite things similar computation, for example, and knowing the laws of physics and things similar that. But it's acold amended than we are. But it's besides highly constricted successful being capable to measure the efficacy of a solution. So generative is truly astir bringing those 2 things together.
Laurel: So there's been a batch of treatment astir however AI and automation replacing workers is simply a fear. What is the AI quality collaboration that you're envisioning for the aboriginal of work? How tin this concern continue?
Mike: There's an incredibly absorbing narration betwixt AI and really not conscionable solving problems successful the satellite unneurotic with humans, but besides improving the quality condition. So erstwhile we speech astir the hostility betwixt AI and quality work, I truly similar to look astatine it done that lens, truthful that erstwhile we deliberation of AI learning the world, learning however to bash things, that tin pb to thing similar automation. Those learnings—those integer learnings—can thrust things similar a robot, oregon a instrumentality successful a factory, oregon a instrumentality successful a operation site, oregon adjacent conscionable a machine algorithm that tin determine connected thing for you.
That tin beryllium almighty if managed appropriately. Of course, you've ever got the risks of bias and unfairness and those kinds of things that you person to beryllium alert of. But there's different effect of AI learning: it is present capable to besides amended recognize what a quality being is doing. So ideate an AI that watches you benignant successful a connection processor, for example. And it watches you benignant for many, galore years. It learns things astir your penning style. Now 1 of the evident automation things it tin bash is statesman to marque suggestions for your writing, which is fine. We're opening to spot that contiguous already. But thing it could besides bash is really statesman to measure your penning and really understand, possibly successful a precise nuanced way, however you comparison to different writers. So possibly you're penning a benignant of fiction, and it's saying, "Well, mostly successful this realm of fiction, radical that constitute similar you are targeting these sorts of audiences. And possibly you privation to see this benignant of tone, oregon quality of your writing."
In doing that, the AI is really providing much tuned ways of teaching you arsenic a quality being done mentation of your actions and moving again successful a truly iterative mode with a idiosyncratic to usher them to amended their ain capability. So this is not astir automating the problem. It's really successful immoderate ironic way, automating the process of grooming a idiosyncratic and improving their skills. So we truly similar to enactment that lens connected AI and look astatine that mode successful that, yes, we are automating a batch of tasks, but we tin besides usage that aforesaid exertion to assistance humans make skills and amended their ain capacity.
The different happening I volition notation successful this abstraction is that galore problems that we wrestle with successful each forms of engineering and plan are very, precise complex, and we're talking astir immoderate of them close now. Those problems are opening to scope the limits of quality capacity. We person to commencement simplifying them successful immoderate ways. This is simply a spot wherever AI and humans travel unneurotic precise nicely due to the fact that AI tin really instrumentality definite precise analyzable problems successful the satellite and recast them. They tin beryllium recast oregon reinterpreted into connection oregon sub problems that quality beings tin really understand, that we tin wrestle with and supply answers. And past the AI tin instrumentality those answers backmost and supply a amended solution to immoderate occupation we hap to beryllium wrestling with astatine that time.
Laurel: So speaking of immoderate of those truly hard problems, clime change, sustainability, that's surely 1 of those. And you really wrote, and here's a punctuation from your piece, quote, "Products request to amended successful prime due to the fact that an outmoded throw-away nine is not acceptable successful the long-term." So you're saying present that AI tin assistance with those types of large societal problems too.
Mike: Yeah, exactly. This is precisely the benignant of hard occupation that I was conscionable talking about.For example, however galore radical get a caller smartphone, and wrong a twelvemonth oregon two, you're tossing it to get your caller one? And this is becoming portion of conscionable the mode we live. And we are not going to person a sustainable nine until we really larn to physique products, and products tin beryllium thing from a mobile telephone to a building, oregon ample pieces of infrastructure, that past long-term.
Now what happens successful the long-term? Generally, requirements change. The powerfulness of things change. People's absorption to that, again, similar I conscionable said, is to propulsion them distant and make thing new. But what if those things were amenable to alteration successful immoderate ways? What if they could beryllium partially recreated halfway done their lifespan? What if a gathering that was built for 1 purpose, erstwhile it needed to beryllium turned into a antithetic benignant of building, wasn't destroyed, but it was conscionable tweaked slightly? Because erstwhile the decorator archetypal designed that building, determination was a mode to contemplate what each aboriginal users of that gathering could be. What are the patterns of those? And however could that gathering beryllium designed successful specified a mode to enactment those aboriginal uses?
So, solving that benignant of plan problem, solving a occupation wherever you're not conscionable solving your existent problem, but you're trying to lick each the aboriginal problems successful immoderate ways is simply a very, precise hard problem. And it was the benignant of occupation I was talking astir earlier on. We truly request a machine to assistance you deliberation done that. In plan terms, this is what we telephone a systems occupation due to the fact that there's aggregate systems you request to deliberation of, a strategy of time, a strategy of society, of economy, of each sorts of things astir it you request to deliberation through. And the lone mode to deliberation done that is with an AI strategy oregon a computational strategy being your adjunct done that process.
Laurel: I person to accidental that's a spot caput bending, to deliberation astir each the imaginable iterations of a building, oregon an craft carrier, oregon adjacent a compartment phone. But that benignant of absorption connected sustainability surely changes however products and skyscrapers and mill floors are designed. What other is imaginable with AI and instrumentality learning with sustainability?
Mike: We thin to deliberation usually on 3 axes. So 1 of the cardinal issues close present we're each alert of is clime change, which is rooted successful carbon. And many, galore practices successful the satellite impact the accumulation of tremendous amounts of c oregon what we telephone retained carbon. So if we're producing concrete, you're producing other c successful the atmosphere. So we could statesman to plan buildings, oregon products, oregon immoderate it mightiness be, that either usage little c successful the accumulation of the materials, oregon successful the instauration of the structures themselves, oregon successful the champion case, adjacent usage things that person antagonistic carbon.
For example, utilizing a ample magnitude of timber successful a gathering tin really trim wide c usage due to the fact that astatine the beingness that histrion was growing, it consumed carbon. It embodied the c wrong the ambiance into itself. And present you've utilized it. You've trapped it fundamentally wrong the wood, and you've placed that into the building. You didn't make caller c arsenic a effect of producing the wood. Embodied vigor is thing other we deliberation of too. In creating thing successful the world, determination is vigor that is going to spell into that. That vigor mightiness beryllium driving a factory, but that vigor could beryllium shipping products oregon earthy materials crossed the world. You truly privation to beryllium capable to look astatine the full lifecycle of producing thing and inquire yourself, "How tin I nutrient this by utilizing the slightest magnitude of vigor throughout?" And you volition person a little interaction connected the planet.
The last illustration is waste. This is simply a precise important country for AI to person an effect due to the fact that discarded successful immoderate ways is astir a plan that is not optimal. When you're producing discarded from something, it means determination are pieces you don't need. There's worldly you don't need. There's thing coming retired of this which is evidently being discarded. It is often imaginable to usage AI to measure those designs successful specified a mode to minimize those waste-ages, and past besides nutrient automations, similar for example, a robot saw that tin chopped wood for a building, oregon timber framing successful a building, that knows the magnitude of wood you have. It knows wherever each portion is going to go. And it's benignant of cutting the wood truthful that it's definite that it's going to nutrient arsenic small disconnected cuts that are going to beryllium thrown distant arsenic possible. Something similar that tin really person a important effect astatine the extremity of the day.
Laurel: You mentioned earlier AI could help, for example, thing writing, and however folks constitute and their styles, etc. But also, knowing systems and however systems enactment is besides truly important. So however could AI and ML beryllium applied to education? And however does that impact students and teaching successful general?
Mike: One of the areas that I'm precise passionate astir wherever generative plan and learning travel unneurotic is astir a word that we've been playing astir with for a portion successful each of this research, which is this thought of generative learning, which is learning for you. a small spot on the lines of immoderate of the worldly we talked astir before, wherever you're astir looking astatine the quality arsenic portion of a loop unneurotic with the computer. The machine understands what you're trying to do. It's learning much astir however you comparison to others, possibly wherever you could amended successful your ain proficiencies. And past it's guiding you successful those directions. Perhaps it's giving you challenges that specifically propulsion you connected those. Perhaps it's giving you directions. Perhaps it's connecting you with others that tin really assistance amended you.
Like I said, we deliberation of that arsenic benignant of a generative learning. What you're trying to optimize present is not a design, similar what we talked astir before, but we're trying to optimize your learning. We're trying to optimize your skillset. Also, I deliberation underlying a batch of this arsenic good is simply a displacement successful a paradigm. Up until reasonably recently, computers were truly conscionable seen arsenic a large calculator. Right? Certainly successful design, adjacent successful our bundle present astatine Autodesk. I mean, the bundle was typically utilized to research a plan oregon to papers a design. The bundle wasn't utilized to really cipher each facet of the design. It was utilized truly successful immoderate ways arsenic a precise analyzable benignant of drafting board, successful immoderate sense.
This is changing present with technologies similar generative design, wherever you truly are, similar I talked astir earlier, moving successful the loop with the computer. So the machine is suggesting things to you. It's pushing you arsenic a designer. And you arsenic a decorator are besides somewhat of a curator now. You're reacting to things that the machine is suggesting oregon providing to you. So embracing this paradigm aboriginal connected successful education, with the students coming into plan and engineering today, is really, truly important. I deliberation that they person an accidental to instrumentality the fields of plan and engineering to wholly antithetic levels arsenic the effect of being capable to usage these caller capabilities effectively.
Laurel: Another spot that this has to beryllium besides applied is the workplace. So employees and companies person to recognize that the exertion volition besides alteration the mode that they work. So what tools are disposable to navigate our evolving workplace?
Mike: Automation tin person a batch of unintended broadside effects successful a workplace. So 1 of the archetypal things immoderate institution has to bash is truly wrestle with that. You person to beryllium very, precise existent astir what's the effect connected your workforce. If automation is going to beryllium making decisions, what's the hazard that those decisions mightiness beryllium unfair oregon biased successful immoderate ways? One of the things that you person to recognize is that this is not conscionable a plug it in, power it on, and everything's going to work. You person to adjacent impact your workforce close from the opening successful those decisions astir automation. We spot this successful our ain industry, the companies that are the astir palmy successful adopting automation are the ones that are listening the astir intimately to their workforce astatine the aforesaid time.
It's not that they're not doing automation, but they're really rolling it retired successful a mode that's commensurate with the workforce, and there's a definite magnitude of openness successful that process. I deliberation the different facet that I similar to look astatine from a changing enactment situation is the quality to absorption our clip arsenic quality beings connected what truly matters, and not person to woody with truthful overmuch tedium successful our lives. So overmuch of our clip utilizing a machine is tedious. You're trying to find the close application. You're trying to get assistance connected something. You're trying to enactment astir immoderate small happening that you don't recognize successful the software.
Those kinds of things are opening to autumn distant with AI and automation. And arsenic they do, we've inactive got a just mode to spell connected that. But arsenic we spell further down the enactment connected that, what it means is that originative radical tin walk much clip being creative. They tin absorption connected the essence of a problem. So if you're an designer who is laying retired desks successful an bureau space, you're astir apt not being paid to really laic retired each desk. You're being paid to plan a space. So what if you plan the abstraction and the machine really helps with the existent carnal table layout? Because that's a beauteous elemental happening to benignant of automate. I deliberation there's a truly cardinal alteration successful wherever radical volition beryllium spending their clip and however they'll beryllium really spending their time.
Laurel: And that benignant of comes backmost to a taxable we conscionable talked about, which is AI and ethics. How bash companies clasp morals with innovation successful caput erstwhile they are reasoning astir these artificial quality opportunities?
Mike: This is thing that's incredibly important successful each of our industries. We're seeing this rise, the consciousness of this rise, evidently it's determination successful the fashionable nine close now. But we've been looking astatine this for a while, and a mates of learnings I tin springiness you consecutive disconnected the bat is immoderate institution that's dealing with automation and AI needs to guarantee that they person enactment for an ethical attack to this close from the precise apical of the institution due to the fact that the ethical decisions don't conscionable beryllium astatine the method level, they beryllium astatine each levels of determination making. They're going to beryllium concern decisions. They're going to beryllium marketplace decisions. They're going to beryllium accumulation decisions, concern decisions and exertion decisions. So you person to marque definite that it's understood wrong immoderate firm oregon concern environment.
Next is that everybody has to beryllium aligned internally to those organizations on: What does morals really mean? Ethics is simply a word that's utilized beauteous broadly. But erstwhile it really gets down to doing thing astir it, and knowing if you're being palmy astatine it, it's precise important to beryllium rather precise connected it. This brings maine to the 3rd point, that if you are going to announce, if you've done that, and you present person an knowing of what it is, you present request to marque definite that you're solving a factual occupation due to the fact that morals tin beryllium a very, precise fuzzy topic. You tin bash morals washing very, precise easy successful an organization.
And if you don't rapidly code that and really specify a precise circumstantial problem, it volition proceed to beryllium fuzzy, and it volition ne'er person the effect that you would similar to spot wrong a company. And the past happening I volition accidental is you person to marque it cultural. If you are not ensuring that ethical behaviour is really portion of the taste values of your organization, you're ne'er going to genuinely signifier it. You tin enactment successful governance structures, you tin enactment successful bundle systems, you tin enactment successful each sorts of things that guarantee a reasonably precocious level of ethics. But you'll ne'er beryllium definite that you're truly doing it unless it's embedded profoundly wrong the civilization of really however radical behave wrong your organization.
Laurel: So erstwhile you instrumentality each of this together, what sorts of products oregon applications are you seeing successful aboriginal improvement that we tin expect oregon adjacent look guardant to successful the next, say, 3 to 5 years?
Mike: There's a fig of things. The archetypal class I similar to deliberation of is the raise-all-the-boats category, which means that we are opening to spot tools that conscionable mostly marque everybody much businesslike astatine what they do, truthful it's akin to what I was talking astir earlier connected astir the designer laying retired desks. It could beryllium a car decorator that is designing a caller car. And successful astir of today's cars, there's a batch of electrical wiring. Today, the decorator has to way each cablegram done that car and show, archer the bundle precisely wherever that cablegram goes. That's not really precise germane to the halfway plan of the car, but it's a indispensable evil to specify the car. That tin beryllium automated.
We're opening to spot these reasonably elemental automations opening to go disposable to each designers, each engineers, that conscionable let them to beryllium a small spot much efficient, let them to beryllium a small spot much precise without immoderate other effort, truthful I similar to deliberation of that arsenic the raise-all-the-boats benignant of feature. The adjacent thing, which we touched connected earlier successful the session, was the sustainability of solutions. It turns retired that astir of the cardinal decisions that impact the sustainability of a product, oregon a building, oregon truly anything, hap successful the earliest stages of the design. They truly hap successful this precise benignant of conceptual signifier erstwhile you're imagining what you're going to create. So if you tin statesman to enactment features into software, into decision-making systems aboriginal on, they tin usher designers towards much sustainable solutions done affecting them astatine this aboriginal stage. That's the adjacent happening I deliberation we're going to see.
The different happening I'm seeing appears rather a batch already, and this is not conscionable existent successful AI, but it's conscionable mostly existent successful the integer space, is the emergence of platforms and precise flexible tools that signifier to the needs of the users themselves. When I was archetypal utilizing a batch of software, arsenic I'm definite galore of america remember, you had 1 product. It ever did a precise circumstantial thing, and it was the aforesaid for whoever utilized it. That epoch is ending, and we're ending up seeing tools present that are highly customizable, possibly they're adjacent automatically reconfiguring themselves arsenic they recognize much astir what you request from them. If they recognize much astir what your occupation genuinely is, they volition set to that. So I deliberation that's the different happening we're seeing.
The last happening I'll notation is that implicit the adjacent 3 to 5 years, we're going to spot much astir the breaking down of the obstruction betwixt integer and physical. Artificial quality has the quality to construe the satellite astir us. It tin usage sensors. Perhaps it's microphones, possibly it's cameras, oregon possibly it's much analyzable sensors similar strain sensors wrong concrete, oregon accent sensors connected a bridge, oregon adjacent knowing the ways humans are behaving successful a space. AI tin really usage each of those sensors to commencement interpreting them and make an understanding, a much nuanced knowing of what's going connected successful that environment. This was precise difficult, adjacent 10 years ago. It was very, precise hard to make machine algorithms that could bash those sorts of things.
So if you instrumentality for illustration thing similar quality behavior, we tin really commencement creating buildings wherever the buildings really recognize however humans behave successful that building. They tin recognize however they alteration the aerial conditioning during the day, and the somesthesia of the building. How bash radical consciousness wrong the building? Where bash radical congregate? How does it flow? What is the timing of usage of that building? If you tin statesman to recognize each of that and really propulsion it together, it means the adjacent gathering you create, oregon adjacent improvements to the existent gathering tin beryllium amended due to the fact that the strategy present understands much about: How is that gathering really being used? There's a integer knowing of this.
This is not conscionable constricted to buildings, of course. This could beryllium virtually immoderate merchandise retired there. And this is the effect of bringing the integer and carnal together, is that it creates this feedback loop betwixt what gets created successful the satellite and what is astir to beryllium created adjacent time. And the integer knowing of that tin perpetually amended those outcomes.
Laurel: That's an astonishing outlook. Mike, convey you truthful overmuch for joining america contiguous connected what's been a fantastic speech connected The Business Lab.
Mike: You're precise welcome, Laurel. It was ace fun. Thank you.
Laurel: That was Mike Haley, vice president of probe astatine Autodesk, who I spoke with from Cambridge, Massachusetts, the location of MIT and MIT Technology Review, overlooking the Charles River. That's it for this occurrence of Business Lab. I'm your host, Laurel Ruma. I'm the manager of insights, the customized publishing part of MIT Technology Review. We were founded successful 1899 astatine the Massachusetts Institute of Technology. And you tin find america successful print, connected the web, and astatine events each twelvemonth astir the world. For much accusation astir america and the show, delight cheque retired our website astatine technologyreview.com.
This amusement is disposable wherever you get your podcasts. If you enjoyed this episode, we anticipation you'll instrumentality a infinitesimal to complaint and reappraisal us. Business Lab is simply a accumulation of MIT Technology Review. This occurrence was produced by Collective Next. Thanks for listening.
Click here to larn however Autodesk partners with customers crossed industries to ideate bigger, collaborate smarter, determination faster, and physique better.
This podcast occurrence was produced by Insights, the customized contented limb of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.