Wild Predictions - Part 1 of ??

I've written before about how easy it is to make up a story about how something works and fool yourself into believing you've discovered something new about the way the world works.  This also applies to predictions about the future, except it's especially deceptive for predictions because of how often they're never measured later on.  If you make a wrong prediction you can just forget you made it move on, but if you make a correct prediction you'll remember that you predicted it and praise yourself for it.  This is especially true if there's no date pegged for the prediction.

It's easy to do, and it's easy to think you're good at prediction when you do it and stumble on a random prediction that comes out right.  Some people take a lesson from this and pre-register their predictions, their confidence that the prediction is true, and then go back and measure that prediction.  I think that's a great idea of limited utility.  And since it's not how most of us think I'm going in a different direction with this post.

I'm going to try a fun long-term experiment.  I know most of my predictions about the future will be wrong.  That's because many important details that will affect the future are hidden to me, so all I can do is extrapolate from trends or from ideas about an idea's potential versus how far it appears from realizing that potential.  This is dangerous, because I may assume something will be true that either never gets realized, or that takes so long to be realized that it's not as transformative as I think it will.  Let me give two examples of personal past predictions I never wrote down, before I get into the next batch of predictions:

Mirasol.  I remember first becoming aware of this technology sometime after Qualcomm acquired it, probably around 2007-8.  The science was sound, and promised so much.  It was faster than e-ink, with a potential frame rate of a few single milliseconds - which meant it could do video.  It was passive, like e-ink, so it didn't drain your battery.  It was full color, relying on wave interference to produce colors, which meant it wouldn't have difficulty producing every color imaginable - far beyond RGB, CYMK, or other color schemes currently in use.  Finally, it was built on MEMS, which are cheap and easy to produce.

At the time (and still to a large extent continuing to this day) MEMS were revolutionizing devices by allowing engineers to design tiny machines on chips that could be manufactured just like normal computer chips.  Mirasol was the elegant MEMS-based solution to e-readers and - eventually - phones, tablets, laptops, and watches.  With display sizes getting bigger and phone components eating up more and more battery, it only made sense that people would turn to a solution that could dramatically reduce power drain and extend battery life.  Alas, it was not to be.  I was totally wrong about Mirasol.

OLEDs.  Wait, why list OLEDs as a thing I got wrong?  When I first started hearing about OLEDs, they were more commonly called LEPs (light-emitting polymers).  I'm not sure when the branding drifted to OLED.  I remember Sony actually selling a proof-of-concept OLED display in their Sony-branded stores.  It was an 11 inch screen, but it was printed onto a piece of glass, and it was beautiful.  It promised a day when you'd have a picture on the the wall that would double as your TV.  OLEDs were thin and could be used in flexible displays - which weren't just theoretical, they displayed working flexible prototypes early on!  Since the light was being produced directly by the pixels themselves, you got true black pixels, and you also got perfect viewing angles.  At the time I thought OLEDs would completely displace LCD displays, just as LCDs had replaced CRTs not too many years before.  There were so many major benefits to OLEDs that it only made sense to switch to them.  Once manufacturing at economies of scale took off, it would be the clear winner.  Indeed, we started to see them make their way into cell phones, and it felt like just a short hop to laptops and TVs.

Instead, LCDs were augmented with arrays of LEDs behind the screens.  These LED-LCD hybrids were shortened to just being called LED TVs.  Companies gradually increased the number of small, bright LEDs they placed behind the displays.  (Indeed, the rise of bright LEDs were the revolution I failed to see coming.)  They also drastically improved the viewing angle problem by re-engineering the display layers.  Now, you could be more than 15 degrees off the perpendicular and still see the image the same as if you were head on.  These changes weren't as good as true OLED displays, but they filled the gap, and displaced LCDs entirely.  Now, you might be tempted to claim that OLEDs just took longer to develop and their time has now come to revolutionize display technology.  But the difference between current LED and OLED is much smaller than OLED and LCD.  We're at more of an iterative step than a magical revolution in new display technology.  And although OLED still looks like the superior display technology, Samsung (who didn't buy into the OLED gamble) is pushing quantum dots instead as a different way to take the next step past LED.

The reason I'm listing this as a 'failed' prediction is because I thought it would change everything.  But it took so long to get to market that engineers working on established technology had enough time to close the gap between the new, promising innovation and the old standard.  Sure, we ended up in the same place, but it wasn't because development followed anything like the path I predicted.

Okay, with that out of the way, I want to make some more wild predictions that have a great story behind them, but which will also likely fail some day.  Some of them may succeed - and you can say you saw them here first! - but many will almost certainly fail.  I'm not putting a confidence percentage on these predictions, because that's not how humans normally tell these kinds of stories about the future.  My hope is that in a few years I'll be able to point to this page to help make the point that just because you can tell a convincing story about something that doesn't make it certain, or even likely, to happen.  Plus, this kind of exercise is fun, so I'm hoping to do a few more prognostications that I can apply the label "Things I'm Probably Wrong About" to.

MRAM.  This technology is really cool, not because it's a high capacity form of RAM, but because it promises to bridge the gap between RAM and system memory.  For those who aren't as tech savvy one of the biggest reasons all computers are slow is because they work on a hybrid system of storing information: slower system memory paired with fast RAM.  System memory is your hard drive.  It's where all your photos and files are stored, but it's also where every program and the operating system are stored.  However, even fast SSDs are still way too slow to run your computer off of.  So a long time ago computer architects designed a system where the computer would work directly from a faster system of memory - RAM - and only save things down to the slower system of memory when it mattered.  When a program 'loads', much of what it's doing is bringing all the data stored on the system memory into the RAM.  It's just copying data from one place to another, and that takes time.

Why can't you just keep all your files in RAM?  Because RAM relies on electromagnetic charges to store each piece of information.  If you lose power for a second the electromagnet fails, and you lose everything stored in the RAM.  That means if all your photos, files, and your operating system were in RAM they're gone the second your toddler unplugs the computer.  MRAM relies on an interesting kind of digital component that was somewhat recently discovered, called a memristor.  In other words, instead of storing information through charge, which requires a constant application of electricity to maintain it, MRAM stores information through resistance, which only uses electricity when it's being written or read from.  That's it.  That's all that's needed to change computers entirely.  Instead of having two systems of memory, constantly copying things back and forth, you'd have one system of memory and no need to copy in order to run things.

If we imagine the possibilities, we can see files that never have to load, programs that never have to be 'killed' and that don't take any time at all to launch to or switch between.  There's no lag switching between tasks, screens, etc. because everything that's on your phone/laptop/device is always loaded.  The only loading screen you'd ever see would be downloading content, which upcoming 5G speeds will hopefully make obsolete as well.  MRAM has the added benefit that you can fit it really densely on a chip because they can produce it in 3D, not just on the 2D surfaces chips of yesteryear (mostly today as well) were built on.  Imagine a tiny thumb drive capable of storing ten times what your current hard drive can (or more).

Drone delivery and driverless cars.  Lots of people talk about these two technologies displacing workers.  The most visible are truckers and taxi drivers.  But I want to focus on other things we can do once we have drones driving around the streets and flying around our neighborhoods.  It's hard to predict the ways people will be creative once a technology makes new forms of interactions available.  For example, who would have guessed how eBay and Amazon would revolutionize commerce?  Later companies like Uber changed things again by combining new technologies with the internet.  So what can you do that's new once we figure out drone delivery?

Let's get the easy stuff out of the way first: Amazon can deliver fresh groceries.  Just about any company can deliver things to your door quickly and cheaply.  That further erodes the value of most big stores, since it would be quicker to buy online than to put on pants and drive to the store, and many businesses will either need to change their business model or go out of business.  But drone delivery allows more than just getting it to you faster.  For example, let's say you want to order a certain blouse, but you're not sure it fits.  Why not try it on, then return it and try on a different size until you find what you like?  Instant and easy returns suddenly become much easier.

All this is the easy stuff.  Let's get a little farther out.

Right now services like Grub Hub are trying to deliver food to your door so you don't have to go to the restaurant.  What happens when you can just order a burger and have it flown to your house?  It's quicker.  Okay, now what happens when you combine the marketplace idea (like eBay) with drone delivery of food?  Disregarding the food safety implications (which would need to be worked out) I could make some extra stroganoff for dinner and instead of putting it in the fridge and hoping I get to it later this week, you could buy it from me.  Thus, this could dramatically change the stratification of food production and delivery.

You'll still have the current models:

  • Supermarkets (maybe fewer physical markets than normal),
  • In-home prep for in-home consumption, 
  • Casual dining (like most burger joints, you order and serve yourself)
  • Formal dining (a waiter takes your order and brings you food)
Add to that new types of restaurants:
  • Internet-based grocery markets
  • In-home prep for external consumption 
    • Surplus meals sold after they're made
    • Small dedicated meals only made for sale
  • Delivery-only restaurants (commercial cooking, but with no seating area or drive-thru)
That last category could produce multiple tiers of food type.  Now, instead of making boxed dinners like Hamburger Helper, you could make fresh dinners and deliver them around meal time direct from the 'factory'.  As remote-based restaurants take advantage of economies of scale, they'll be able to bring their prices down closer to the cost of today's supermarkets (which have to hold product on the shelf in a store).

One of the biggest places you gain benefit from in this model is that the inventory is constantly moving through the supply chain - instead of waiting on a shelf, then waiting again to be made.  McDonald's knows how much they sell, because they produce it at such volume that they can predict how many hamburgers they'll sell in the city of Seattle in the month of August within a reasonable day-to-day margin.  Once they get down to a specific store and a specific hour it can be more difficult to predict with precision.  How are they supposed to know that Chinese tour bus is coming through?  But the more they're able to centralize their operations to one large production facility, the smaller those outliers get as they're swallowed up in the noise of thousands of transactions an hour.

So if the price to 'eat out', or more accurately to order prepared food, gets close enough to the price to prepare food what happens?  At some point we won't just eat out a lot.  Many of us will switch to eating prepared food exclusively.  We won't need to take the time to prepare our own food.  Like cord cutters who don't order cable, the next generation may opt to forgo having a kitchen.  Maybe they'll keep a refrigerator for leftovers, or because it's quicker to store cold drinks there.  A minority of people might really like cooking and keep that as a hobby.  But since it's a hobby they can easily make money at, they'll probably want to have a larger kitchen.  After all, if I'm going to make chicken cordon bleu, I could just as easily toss ten in the oven as two.  Now, instead of dinner costing me money, I'm making money off of dinner.

(The French translation of 'cordon-bleu' is 'blue ribbon'.  As the blue ribbon is usually given to the first place in a contest, you could say that 'chicken cordon bleu' is just a fancy way of saying "winner winner chicken dinner".)

Okay, have we plumbed the depths of drone delivery enough?  Not even close, but I'll stop after pointing out two more things:

The next generation of GPS is expected to have much higher precision than the current generation.  Why is that important?  Well, let's say you want a drone to deliver your burger, not to your home but to your hand as you sit at a park bench.  Or better yet, it would be nice if the drone could deliver a funnel cake to you while you sit atop the Ferris wheel or walk through a crowd.  If you have your phone on you, the drone knows where you are, but if you have high-precision GPS and the drone has some basic image-recognition software (that's a hand, put the package in the hand) the drone could pick you out in a crowd and deliver your order to you wherever you happen to be.

And this extends to more than just food.  Got a splinter and need some tweezers?  No problem.  Forget the sunscreen?  Got it covered.  Want to play beach volleyball, but there's no equipment handy?  It's on the way.

This brings up one final possible change to how we interact with the world around us that drone delivery could make.  What if, instead of owning a football you play with every once in a while, you just rent it?  Since you can rent it and have it literally in your hand wherever you happen to be within ten minutes or so (depending on where the warehouse is located, and the logistics of delivery - packaging has to change if we're delivering it to people in their hands) the benefit of owning certain types of things diminishes substantially.  Why store all that stuff - why have to look for all of it when you need it - when you could just buy what you need when you need it, or just rent it?  Maybe you do want to keep some things that you use often.  Other things are of sentimental value, so you'll keep those as well, but what's the point in having it all on-hand all the time?  Lots of people store things in safety deposit boxes because a bank can store it more securely than they can.  If you didn't have to go to a bank, you just had to wait ten minutes for your stuff, wouldn't it make more sense to store most of what you own off-site?  That way it's cataloged and never gets lost or stolen.  It's always available, easy to find out there in the 'physical cloud'.

For the small stuff, a little flying drone delivers anything you want and carries it away again when you're done with it.  For large or heavy items - including people and laundry - a driverless delivery vehicle comes and drops it off.  A laundry service that operates at scale makes your inefficient in-home system less of a value proposition, and there's no reason to own your own driverless car when you can just hire out the vehicle you want to use when you want to use it.

Thus, the home of the future has fewer storage closets, no laundry room, no garage, and possibly no kitchen.  Moving across country is a matter of packing up a small amount of frequently-used items and telling the rest of your stuff to move to the local storage depot in your new city/country, which happens automatically.

Designer enzymes - quantum computing.  One more and then I'm done (for now).  This one may be more esoteric to anyone who hasn't taken a certain amount of biochemistry and organic chemistry.  In fact, it's nigh invisible to most of us, even though it's at a foundation layer for much of the world we live in.

Back in my undergraduate organic chemistry class, we started out the unit for amines with a demonstration chemical structure written on the board: methamphetamine.  The professor began the lecture and was describing some of the properties of amines (nitrogen-containing organic molecules) when one of the students called out "synthesis" from the back of the class as a joke.  Everyone laughed, but the professor seriously considered the request.  "No," she replied after a brief hesitation, "I'm not going to waste class time on that.  It's too easy.  You can work it out yourselves."

What she meant was that it was too easy to develop a process of going from a few precursor molecules, adding them together in the right conditions, and producing methamphetamine as the end product.  That doesn't mean it's actually easy to make it.  A large part of organic chemistry class was learning how to create a desired chemical from a few starting chemicals.  It was a fun kind of puzzle, if you like that sort of thing (between synthesis and spectroscopy o-chem was fun for me, sorry to those of you who struggled with it).  The more complex the chemical, the more difficult the puzzle, because it would take multiple steps to build up the final chemical you were aiming at.  For each step you had to outline any new chemicals you needed to add (i.e. boron tetrahydride as a hydride source) as well as any additional conditions that would need to be present to make the reaction take place, like pressure and heat.

This exercise is something chemists have to deal with all the time, though.  They have to place caustic chemicals under intense heat and pressure in multi-step reactions in order to get new and interesting chemicals out the other end.  There are a few problems with this, though.  Maybe you want one product, which your process does make, but the reaction creates other similar products out of the same reaction.  Sometimes chemical reactions can make more than one thing.  This is like giving a kid three Legos, telling them to combine them 'however they want' and collecting the results.  The Legos have to combine them based on certain rules (no peg to peg), but within those rules the kids are free to combine them however they want.  If you give the kid something simple, like two single-peg pieces, there's only one real way for them to combine them.  But the more complex the piece, the more variations are possible.  This same dynamic happens in chemical synthesis.  There are ways around the problem, but they also require their own extra reaction steps.

The end result is that either you work to create a final product that's very valuable (you're willing to perform multiple costly, dangerous reactions to get it; then additional time and expense separating out the right product) or you just try to find something in nature that's already making something similar and use that instead.  Sometimes you use simple reactions on large batches to create cheap products, but the more complex the synthesis the more difficult and expensive the end result.

This is really too bad, because there are all sorts of fun things we can do once we start designing our own molecules from scratch.  There's a whole sub-field of chemistry that focuses on something called meta-materials, which are large molecules designed to produce certain specific outcomes. For example, some enterprising scientists designed a class of meta-materials that could redirect light so that it exits the material in the same direction it enters the material.  In other words, it's a 'cloaking' material.  There are some restrictions involved, but this is just the tip of an iceberg of possibilities we're only beginning to explore.  These could be the future's equivalent of MEMS/NEMS.  Of course, too many of the molecules we can imagine would be far too difficult and expensive to produce - especially at scale.  Every once in a while some new type of fertilizer or something becomes available because a chemist somewhere figured out how to make the reaction economical to execute at scale.

Okay, now look at this same situation from a biologist's perspective.  Nature creates many of the same molecules we already use, and does it more effectively than we can in a factory.  Indeed, it's almost always more efficient to grow a whole plant that creates one of its many thousands of molecules for some ancillary purpose, kill the plant, and extract the one chemical we want than it is for us to just make a large batch of that one molecule.  Why is that?  How can plants make complex molecules at room temperature under normal pressures that we can only make after a dozen steps, some of which are at hundreds of degrees and at dangerous pressures?

They take advantage of a magical kind of catalyst called enzymes.  A catalyst helps a reaction proceed with lower activation energy (that's what we're using the temperature and pressure for) than would be required without it.  Enzymes are special kinds of catalysts, though.  They make sure only one reaction happens.  Instead of making many different chemicals to separate out, only the one chemical you wanted is produced.  They can sometimes do complex operations, too.  For example, there's an enzyme that helps DNA when it gets too twisted by grabbing one strand, cutting the DNA, and unwinding the cut strand around the other strand so it can be reattached after getting untwisted.  That's a complicated interaction happening at ridiculously small scales.  And it's all done by these tiny, complicated molecular machines called enzymes.

If enzymes are just large molecules, though, don't they take a lot of these same chemical reactions to produce?  Not really, because they're just proteins.  We know how to produce proteins like this in bulk.  It's mostly just a matter of figuring out a sequence that will produce the shape and properties we want.  Then we can just create the DNA code for it and start producing the enzyme.  There's a lot of complexity hidden under the surface here, but trust me when I tell you that it's a solved problem.

So why aren't we using enzymes more often?  Well, it's really hard to design the enzyme.  It has to interact with the molecule we want to work with, change its shape, and force a new reaction in exactly the way we want.  That's a hard theoretical problem, but it's an even harder computational problem.  All this usually happens in water with various salts present, and that means you have to model all this behavior in a computationally complex way.  So, in part we just don't have the computer power to figure all this out.  Once you figure out the right shape/sequence, predicting how a protein will fold is easy to define mathematically, but takes a long time to compute.  This is what Folding at Home is trying to do when it uses distributed computing to cobble together a massive virtual supercomputer.  It's using all that power to work out the computationally complex problem of protein folding, which is at the heart of how enzymes work.

One major application of quantum computers that's already beginning to become useful is for this prediction of protein folding.  These molecular-level interactions have to be known with high precision if we're ever going to design our own enzymes from scratch.  Using traditional computers to achieve that precision takes way too long.  Maybe if we're lucky and Moore's Law continues for a few more decades we'll get there on traditional computer architectures.  More likely, as quantum computing starts coming into its own in the next few years we'll use it for this kind of dynamic molecular modeling.  That's the missing tool we need in order to design our own enzymes.  And when we can design our own enzymes, we can create chemicals and materials that would seem like magic to anyone alive today.  At that point, it would not be wrong to characterize us as truly having 'command of the elements'.

I could go on to talk about the massive wave of innovation that would allow, but I've taken too long just describing the process that will lead us there.  I know it's complex, but the potential for revolution is high.  It would allow us to tune the properties of literally any chemical structure (within the limits of the periodic table and the constraints of basic chemistry).  So any chemical (solid, liquid, or gas) you can think of could be a target for improvement and innovation.  Here's a short list:

  • Rocket fuel
  • Paints and pigments
  • Glues and epoxies
  • Synthetic plastics, rubbers, polymers, etc.
  • Pharmaceuticals
  • NEMS
  • Metamaterials (superconductors?, ultra capacitors?)
  • Solar panels, batteries

This could be bigger than all the other innovations on this list combined.  Indeed, the next age after the 'internet age' will likely be the 'designer enzyme age'.

We need a better name for that.  Maybe call it the 'Desyme Age'.

Only pay attention to the man behind the curtain

That's all I'm going to speculate on for now.  Reading back over these, it feels like we're at the cusp of some really amazing innovations in the near future.  Some of these problems are really close to being solved, at which point life will change dramatically.

But the point of this post is to help you calibrate away from the expectation that a prediction is true because 'he tells a good story'.  It's very tempting to think you know something new because you read a really compelling prediction.  But trends aren't destiny, and sometimes we overlook important details.  For example there are a few different types of non-volatile memory out there, besides MRAM.  I think MRAM currently has the most promise, but maybe if all you watch for is MRAM you'll miss the non-volatile memory revolution because it happened elsewhere.

Or we could see gradual improvements, instead of one huge technological breakthrough, as was the case with LCD to OLED.  Since I first made this prediction privately a few years ago, we've seen loading and boot times go down significantly.  The more those times go down, the less important something like MRAM could become.  Maybe when it's introduced at scale it's only for minor power-saving purposes.  Or maybe before MRAM can take off we get cheaper high-capacity RAM.  That would allow your computer to load the whole hard drive into RAM, thereby making the hard drive just a 'backup' in case you lose power temporarily.  Or we could see speeds for current hard drive technologies increase enough that they're on par with RAM.  This is functionally no different from developing 'non-volatile RAM', of course.  But it's a gap that has been shrinking since SSDs were first introduced to replace HDDs.

It's not important how we get from here to there.  The point is that just because an idea sounds amazing doesn't mean we'll implement it at scale before gradual improvements to the current system make that innovation moot.

Or maybe what we thought was a simple problem close to being solved turns out to be a Gordian's knot of complexity that never gets unraveled.  This could easily be the case with drone delivery, driverless cars, or designer enzymes.

So why bother making predictions?  Well, for one because it's fun to speculate on what's theoretically possible "once we get the engineering right".  (I know my engineer friends cringe at that.)  For another, it's in imagining the future that we build the path to create it.  If it's possible to do some of the things outlined above, it will be exciting to make them happen.


Popular posts from this blog

Reverse Engineering Life

Cancer update: precision oncology comes into its own