The world is full of architectural follies, some truly inspired and some just silly. Some are small and circumspect, while some are built on a grand scale. One at least, at Songshan Lake, Dongguan, on the outskirts of Shenzhen in China is, in area, the size of a small town. This is where Huawei is now building its new R&D campus and to be fair, whether it proves to be a `folly’ or not will be down to what comes out of it over the coming years. But while the insides of the buildings may be made up of offices, laboratories, test and development suites, the outsides are anything but.
Company founder and CEO, Ren Zhengfei, is a fan of western architecture, and consequently challenged the architects to 'do Europe' on the exteriors of the buildings. Only a few of them have so far been built, but there are already touches of Heidelberg Castle, Burgundy, Verona, and the Palace at Versailles to be seen. The plan has the UK represented by `Windermere’, which conjures up thoughts of steam-powered pleasure vessels and slot-machines.
But that is another matter. What matters more to the company will be the challenges set for the estimated 25,000 scientists and engineers who will end up working there. Some of those roles were outlined by senior executives at the Analyst Summit, and while some were what might be expected from a comms equipment company on the verge of the next big market development, others did see the company eying up one or two of the most serious technology challenges that anyone could tilt their cap at.
William Xu, Director of the Board and President of the Institute of Strategic Research of Huawei, announced that Huawei is now hanging all its R&D effort on what he called `Innovation 2.0’, where the target is much more 'R' than `D’ – it is all about vision-driven theoretical breakthroughs and inventions. They are looking to drive innovations in theories and basic technologies. To this end, the Institute of Strategic Research will invest in three areas: basic scientific research and its associated cultivation of emerging talent, technological research, particularly in the area of developing diversified industry applications (an area where the company expects to be working closely with the academic community to speed up the process of translating university research into real-world applications), and lastly technological innovation through its many joint research programs.
Its bleeding edge will look at…
Its prime targets are research programs to find ways around three of the major scientific/technological road blocks in the way of communications and IT development out into the future – what Xu defines as the `memory wall’; the bug-bear of communications, the Shannon Limit; and the impending end to the effects of Moore’s Law. Most people will know of Moore’s Law, named after its proponent, Intel co-founder, Gordon Moore, who observed that the number of transistors it was possible to diffuse into an integrated circuit of a given size roughly doubled every two years.
There was also the additional observation that, because of the way the production process could be refined once up and running, improvements in the production yield meant that unit costs also halved over the same time period. Coupled with performance improvements as a by-product of manufacturing refinements – mainly in the form of 'shrinking' the chip design – this has given a continuous, many-fold improvement in capabilities at roughly static costs for the last 50 years.
But now the laws of physics are coming into play. The chip makers are now down at the level of single atom manipulation, and the concern is that the laws will prevent them going any smaller. David Wang, Huawei’s Executive Director of the Board, Chairman of its Investment Review Board, and President of ICT Strategy & Marketing, says that the alternative being targeted is optical computing, where photons of light are used instead of electric current. The idea behind this has been around for a good few years but so far it has not proved to be a workable solution. But if its many technical issues can be resolved, it could provide the next big step forward in computing, not least because it might be possible to have one device – one 'transistor' – operating on more than one process at the same time. This is because light streams of different colours could be processed together without interfering with each other.
The same basic technology might also be useful in bypassing the Shannon Limit. This is the maximum rate that data can pass through a channel with a given bandwidth and a given noise level. Exceeding this maximum channel capacity is likely to lead to at best limited data loss. Ways of overcoming this limit are now emerging that use multiple colours and multiple input/multiple output (MIMO) neural networks that show the potential to deliver 600 Gbit/sec per wavelength, and up to 120 wavelengths per fibre.
The ‘memory wall’ is the need to keep pace with the rate at which humanity is creating data, for current technologies are hard-pressed at the moment, and increasingly failing to keep up. The option Huawei is keen to start research on is the use of DNA as a storage medium. Its storage capabilities are evident in all of us – a single gram of the stuff is said to be capable of storing a Zettabyte of data. The downside so far has been in finding a way to read that data back. University researchers are now reporting that bacterial nano networks are showing real promise at delivering that half of the equation. William Xu suggested the techniques based on gene splicing and gene sequencing can respectively provide the `read’ and `write’ functions that will be needed.
But what about now?
Back in the here and now, Huawei Chairman Ken Hu took the audience through some of the current crop of 5G-related developments the company is bringing forward. He said the company is seeing two main trends, one of which it has called Zero Search, and the other SuperSight. The latter is what one might think it means, the use of AI and analytics to see as yet undiscovered information as well as more granular details of what is currently known, and therefore understand it and its issues more clearly.
The former seems a tad contradictory however. Zero Search means not having to do any searching, because the AI systems will know us so well it will be able to second guess our needs. For the terminally lazy this may sound heaven-sent, but its alter-ego could go as far as moulding users to accept whatever is pushed at them as being appropriate to their current needs - mere recipients of what the system sends - and compliantly content with that situation to boot. It has to be said that, looking from the outside, this does seem to map onto many peoples’ perception of China, as well as some of its avowed recent public order and management policies.
The statistics of 5G in terms of performance have been widely reported. Perhaps more significant is the direction of travel Hu expects to see. Using 2025 as the target year, he expects to see 6.5 million base stations installed, with 2.8 million users, and coverage for 58% of the world's population:
5G is a connectivity platform, not a communications pipeline, so it provides the chance to redefine the devices that are used to exploit it.
While being a passing remark, it is also carries a tremendous challenge, especially for those looking to new devices that can couple interoperability of the speed and bandwidth of 5G with new ways of achieving business advantage and results. He expects it to at least trigger the arrival of new VR/AR application and devices this year, and expects to see 9% off all businesses and 340 consumer users by 2025. AI should be in widespread use by 97% of all businesses by that time. The company has also recently introduced ModelArts, an AI applications development toolkit developed specifically for its own moves into delivering cloud services. According to Hu, there are already over 200 projects developed using the tools.
Is this scary stuff? Maybe. Would it be scary if it was about a US, European or even UK company? The fact that this is Huawei and China still has this strong, latent history as 'the other' will probably make it scary for a while yet. But it is conducting research that it would be foolish as a business to ignore, and being reasonably open about it all as well. We ignore it at our own peril, for the results are very likely to be damn useful