The Other 5G

…in which the author realises the question “what is 5G?” is the most pointless one of all.

I have something to confess.  I’m a middle-aged guy.  I’m not proud of it, and I try not to talk about it too much, but I can’t escape it.  I graduated before there were mobile phones.  I saw my first IBM PC when I was a year into my PhD.  In the meantime, though, I’ve witnessed the birth of cellular, and seen it stumble and grow through the national analogue standards, into the pan-European digital standard that became GSM, marvelled at the introduction of 3G (what will we ever do with all that data?) and watched its true globalisation with 4G/LTE.

And now, for richer, for poorer, in sickness and in health, 'til death us do part, we’re all talking about 5G.

We’re all talking about 5G, but how much listening’s going on?  There’s a huge gulf between the skeptics and the enthusiasts, as if each group is in its own cone of silence. It’s clear that there’s a huge science fiction element to many of the applications proposed for 5G, with many desirable unsolved (or even unsolvable) communications problems being stirred indiscriminately into the pot, each undermining the credibility of 5G a little more.  The marketing videos showing driverless vehicles navigating unmarked road junctions manage to be scary and funny at the same time.

When you look below the surface, though, there’s a lot of hard-core engineering going on at the radio layer, as the world realises that spectrum below 6GHz will not be enough for the longer term demand. There’s a groundswell of activity going on to find engineering solutions to millimetre-wave radios all the way up to 100GHz.  The chasm between the use-case fantasy and the engineering reality will close eventually, and a few people will have to eat their words, just as those who couldn’t see what all the fuss was about with 3G had to eat theirs.

But in discussing 5G recently, I daydreamed back to the time when the IBM-PC was new, and another Fifth Generation was on the horizon.  Then, as now, much of the energy was coming from the East (and I don’t mean Old Street roundabout).

Back in 1981, most of the activity in the new and exciting personal and mini-computer market was in the US, with Prime, DEC, DG, Osborne, Apple and others, though it was IBM entering the market that seemed to make the world sit up and take notice. At the other end of the scale, much of the simulation work being done in UK academia was being done on US built supercomputers built by Cray, Amdahl and others.  But all of the economic news seemed to be coming from Japan.  Through the seventies, Japan had cemented its transformation into a world industrial superpower. I remember my PhD supervisor remarking on the rise in publications in our field being made simultaneously in Japanese and English, and me commenting with genuine concern that we will need to really worry when they stop bothering to publish in English.

So when Japan’s Ministry of International Trade and Industry (MITI) announced in 1982 its Fifth Generation Computer Systems (FGCS) project, I remember a real frisson running through my cohort of new PhD students, including me.  Was the hi-tech race over, before we’d even entered it?  At a time when real computers were mainframes, and small computers were simply about word processing, desktop calculating or industrial control, this seemed a genuinely possible outcome.  If the 5th Generation initiative succeeded, all computer technology would now originate in Japan.  Wafer Scale Integration (another “inevitable” trend of the time) would become the norm, with whole operating systems (not to mention applications) burnt into the silicon at manufacture.  While WSI wasn’t really inherent to FGCS, the two together seemed destined to rule out future innovation, and specifically, any future innovation in hi-tech in Europe.  The cost of a new design of computer and of new software would just be too high.  We would simply have to accept what the Japanese 5G initiative delivered, on platforms that WSI manufactured.  And that would be it, for ever and ever.  It would be the end of engineering.

But just as Lord Kelvin predicted the end of physics a hundred years earlier, thoughts of the end of engineering were somewhat premature.  While the Japanese 5th Generation Computing project did some interesting work, the reality was that its innovations were made redundant by technology shifts elsewhere.  Plain old von Neumann architecture computers just got faster and faster, and the software to run on them got better and better, at a rate that made the impact of the FGCS initiative less and less relevant.  WSI also became a technological dead-end.  Operating systems evolved to provide hugely flexible platforms on which innovative software could continue to be written and updated.  The partitioning of systems has rationalised to the point that no-one thinks that embedding a Fortran compiler into the silicon is even remotely sensible.  Don’t laugh – I remember actually discussing this.

So are there lessons from that 5G for this 5G?

There are some important differences.  That 5G was up against silicon process and programming language limitations that were ultimately blown through by developments in other spheres.  This 5G is up against a data crunch, largely driven from laws-of-physics limitations in how much data we can get through the available spectrum.  If we accept that the limits are physical, then the drive to open up new spectrum, and use the spectrum we have more flexibly and efficiently (key tenets underlying 5G), then the solution drivers behind 5G will remain compelling.  But if the data crunch is a fiction, invented by the industry to keep itself relevant, then we could find ourselves in similar place.  One similarity is that we are extrapolating a long way into the future.  The MITI vision in 1982 was a ten year one.  The 5G vision today is on a similar timescale.  Millimetre wave spectrum is scheduled to be re-allocated to mobile broadband at WRC-19, and only following that will national spectrum auctions be organised, networks built out and begin operating.  Ahead of that, we will continue to deploy LTE/4G in the sub-6GHz spectrum we have, using small cell concepts well-established in 3G technology.  Just as MITI found during the eighties, the future of computing lay not in huge mainframes, but in surprisingly powerful laptops and handhelds, we may find out in the coming decade that the data growth on which millimetre wave technology depends for a market arrives more slowly or not at all, and that existing 3G and 4G technology does very nicely thankyou.  Through the other end of the telescope, the Internet of Things, which needs extreme power efficiency rather than extreme data throughput, may find itself perfectly happily served by enhancements to narrowband GSM, or low power WiFi 802.11ah, or other emerging proprietary standards.  Other applications proposed by the 5G community, such as Ultra-Reliable Networks (for traffic management and other safety critical applications) may find limited market acceptance, unworkable business cases, unacceptable safety margins, or may, in the end, just not work.

Even so, the mobile industry has a pretty good record of making long term bets.  In one of the panel sessions at MWC this year, François Rancy pointed out that the spectrum allocations made at WARC-92 are still in use over twenty years later.  If the ten year bet on 5G is made on similar premises, we’re in good shape. 

Having said all that, this 5G is a field ripe for disruption, just as it was thirty-odd years ago.  The use cases and applications that cram 5G conference proceedings today feel like the visions of the prisoners in Plato’s cave. We see shadows on the wall of an ideal that we can never approach and we pretend they’re real. But we are not prisoners. We can venture outside the cave and see the world as it is, not by its shadow.  5G today may turn out to be as irrelevant as 5G in 1982, but we’re not going to change that by projecting more and more fantasies into the shadows on the wall. It’s time to press the technology in both directions – both towards higher capacity and more spectrum for mobile broadband, and towards higher efficiency and longer range for machine-type communications, and see what it can actually do. We’ve dreamed as much as it’s safe to dream without confusing sleep with wakefulness.  When we see what the technology can deliver, the applications to use it will spring from the ground like plants after desert rain. Then we can build businesses around them, and evolve the businesses we already have to exploit them.

And if history tells us one thing, it’s that our imaginations are pretty feeble in predicting the future.  We can make fairly linear predictions about spectrum availability and radio performance improvements, but what people do with that performance will continue to surprise us.  What truly distinguishes 5G from its forebears won't be clear until it’s upon us.  So let’s stop predicting.  Let's stop asking, what is 5G?  Let’s just make it.