by Vic Grout
Ulrich was dressed as a typical managerial scientist. The only concession he seemed to be prepared to make to his elevated role was to replace the default scientist’s t-shirt with an open collar; the jeans remained the same. He welcomed Bob warmly; and had his enthusiasm returned equally. Once the initial pleasantries were exchanged and security processes complete, Bob moved in on the business in hand.
“So what have you got here?” he opened as they walked together towards one of several large buildings at the centre of the complex. “What’s the problem? What’s not working?”
Ulrich smiled mysteriously. “To take those questions in reverse order, Mr. Weatherill, nothing is not working; we are not sure if we have a problem; but whatever we do have, we are hoping you can tell us what it is!” It was Bob’s turn to grin, his curiosity already raised; he had expected something like this from their remote discussions before he had left London: it appeared he was not to be disappointed.
HGMS-Ion – sometimes just HGMS – was a heavy-ion accelerator plant, which had been open for about thirty years from modest beginnings. It specialised in certain types of controlled atomic collisions, which were difficult to produce in conventional research centres. Scientists from all over Europe – and sometimes the world – came to run specialist experiments in booked sessions with time and availability always at a premium. The centre could support a few dozen such guests at a time and had a growing reputation for results across a range of scientific research fields including biophysics, materials and medicine. Several advances in cancer radiation treatment, for example, had been made there in recent years.
While the experimental research community was generally of the visiting variety, HGMS retained a much larger core of resident support staff. These local engineers and scientists were responsible for maintaining the operational capacity of the centre. In his job as Senior Technical Officer, Ulrich was responsible for coordinating home teams of physicists, chemists, mechanical and electrical engineers, computer scientists and networkers. Instantly, Bob did not envy him in such a role: herding cats sprung to mind.
HGMS was expanding, Ulrich explained with some pride. Over the next eighteen months, the old hundred-metre linear accelerator and twenty-five-metre synchrotron ring were being replaced by a new hundred and sixty-metre accelerator and ninety-metre ring, both of which would increase the centre’s experimental scope considerably. Several other aspects of the site were also being upgraded and improved; some sections were entirely new. Much of the physical work was complete. The engineers and physicists had fulfilled their roles; most of the hardware – from concrete to precision equipment to wiring – was installed and the computer scientists and networkers had moved in … at which point the ‘issue’ had come to light.
“Quite simply,” said Ulrich quietly, followed by a considerable pause, “there is noise on the network. That is it, really; there is noise.” He continued quickly, “We do not know what it is or where it comes from. It does not seem to be doing any damage at the present time but it is there and we do not know why.” We believe we have eliminated all sources of external interference but it is still there. Bob offered nothing at this point beyond a slight raising of the eyebrows so Ulrich continued.
“We are worried that this may become a much bigger problem in the future. You see, the purpose of the network is essentially to carry timing signals around the system so key events across the accelerator can be synchronised and recorded. It needs to be – and it is – incredibly accurate. We are talking nano-second precision here to keep the physical parts of the process in time. Most of the essential network processing itself is chip-based; you will be familiar with Network on a Chip technology?” Bob nodded. “That part is not causing us much concern,” Ulrich said deliberately. “But, of course, the signals themselves still have to be distributed over the physical network to and from where they are needed or produced; and these have to have the same nano-second accuracy – after all, that is the whole point!
“So, naturally, when we have been testing the system, we have been looking at the carrier lines in considerable detail – on a very minute scale – and this is where and when we start to see the noise. We would not have noticed it if we did not need to look so closely.”
“What sort of noise, exactly?” asked Bob. “And why isn’t it affecting the operation of the network?”
“Once again, the second question is easier to answer than the first, Bob,” smiled Ulrich with a trace of embarrassment. By this time, they were inside the central unit building and at the door of the main control room. “The timing signals we are currently using for testing the system are sparse – there are few of them – because the atomic events are currently not real. We are simply generating test events to monitor the behaviour of the system and the network. But the noise itself is discrete, rather than continuous, and also sparse and – at present – they generally do not coincide: one is not interfering with the other. However we suspect that we may have a problem when we try to switch the timing network across to the real system.”
“I would imagine that’s pretty likely,” agreed Bob, thoughtfully. “And the first question? What sort of noise have we got?”
“I’m a quantum physicist, not a computer scientist,” grinned Ulrich. “I think it may be as well for you to look for yourself!” He opened the door and motioned to vast racks of equipment on two sides of the large room. Bob whistled at state-of-the-art kit he only recognised from catalogues. A large display screen was fixed on a third wall. A woman and two men, in their mid-thirties – also casually dressed, sat around a table in the centre but rose together as Bob and Ulrich entered. “As I said, Bob, we are still in a testing phase – not yet live. Please go where you will and do as you choose; and ask for anything you do not see. I put my team and my facilities at your disposal,” he said by way of introduction. An open door in the far corner showed another similar room beyond and a familiar package on a sturdy trolley against the adjacent wall: Hattie was already there.
Quickly, Bob was properly introduced to several others in the HGMS senior network team and given a hurried tour of the facilities, meeting further key personnel along the way, including the Network Manager, Karla Heintze. He visited the ion source area, the accelerator line, the storage ring and finally the delivery centre: a huge warehouse, in effect, where massive, somewhat ad-hoc, caves of concrete denoted the end of the line for any given ion beam – and where the results of individual experiments were recorded. They finally returned to the network control centre to discuss the problem. Bob asked Karla and her team as many questions as seemed sensible at the start but, beyond the facility’s basic operation, they could offer little insight into the problem itself. He therefore soon started up Hattie and began to take readings; initially at the network centre, then around the facility. As a slightly clearer picture began to emerge, other questions followed and slowly he started to see what was happening. At least, that is, Bob could gradually see the noise itself – the effect; the cause was another matter entirely.
He had been briefed very well. What he was seeing was both discrete and slight. It came in small bursts rather than being continuous background noise and it was generally – although not always – at a somewhat lower voltage than the signals the network was designed to carry. Although Hattie picked it up fairly easily, the HGMS folk had done well to notice it at all. Occasionally, of course, Hattie picked up a legitimate data frame, which made calibration easier – fortunate because neither of them were familiar with working at this ultra-precision scale.
Despite Ulrich’s (and, it appeared, Karla’s) assertion that the noise was not currently interfering with real signals, Bob was fairly confident that, if he waited long enough, he would see this happening: it seemed a statistical certainty. Sure enough, after a while – and with Hattie’s accuracy ramped up close to her limit, they witnessed a frame collision and monitored the subsequent retransmission of the data from source. That certainly would present timing problems w
hen the system was operational, Bob reasoned, but, other than that, the system seemed to be behaving normally. The only issue was what the noise was and what was causing it.
And it seemed to be everywhere. Hattie sniffed the network from the central matrix to the extremities of the test system and the spikes were network-wide. Occasionally, they saw something more than simple low-voltage spikes: the odd signal was either slightly higher voltage or longer in duration and, very occasionally, even seemed to have something like a square-wave binary data form – although generally no more than the equivalent of a few bits. That was really weird; but it might just be coincidence at this stage.
Bob adjusted Hattie again; this time to look for and record the longer periods of noise. He set her to run for half-an-hour while he joined the others for coffee. When he returned and looked at the oscilloscope log, he was able to focus on the ‘bigger’ noise, which made up less than a hundred-thousandth of the total signals Hattie had processed. On the whole, the results confirmed what he had thought before: among the random electrical spikes, were a very small number of longer, larger noise signals that looked a little like fragmented binary digits. One, in particular, stood out: comprising what looked like (even if it was unlikely that it was) about thirty bits – several bytes – of data. Either way, it was pretty obviously a hardware fault; but where was it coming from and how was it managing to propagate across the entire network? Normally, bogus data from a malfunctioning transceiver would not get past the next device. Were all the devices faulty? Unlikely, he thought.
He left Hattie running – scanning links across the whole network – and went to ask some questions. But he did not get many answers; at least not any that helped much. ‘When had they first noticed the problem?’ No-one was quite sure; it had been mentioned two weeks or so before but only documented the previous week. ‘Had it started suddenly or gradually? Had the pattern changed at all?’ No-one could say. ‘What about background interference? Radiation perhaps?’ No, they had measured and tested for all that: there was nothing to be found away from the network itself. ‘Could it be a bad batch of equipment? Was all the kit from the same supplier?’ No: there were at least three different suppliers for different levels – regulations required that. Moreover, it had been ordered and delivered at different times: a production problem was impossible.
For the first time in a while on a real case, Bob was beaten. He had no idea what the problem might be. The network was properly configured and it was hardly credible that all the equipment was failing simultaneously across the whole network – after all, to all intents and purposes, it was working. It was just noisy – dirty almost. What else could it be? He did not know. He stayed late, until the end of the afternoon became the early evening, but eventually switched Hattie off and confessed to Karla, then Ulrich, that he could not help. Of course, there would be no bill; even so, he would look at the data he had logged in more detail when he had time later. It was still possible that he might yet get some insight, but he might not. What hurt his pride even more was, in addition to the admission of defeat, he could not help but think that Ulrich seemed less than surprised. There was almost a resigned – perhaps a knowing – look from him as they parted. The car took him back to the airport in Frankfurt, where he checked into his hotel. Early the following morning, he flew to Luxembourg.
*
However, Bob barely stopped thinking about the Darmstadt Dirty Network, as he started to call it, through most stages of the journey and in the quiet of his hotel room. Hattie had gone off with the overnight couriers; she would not arrive before him this time as the EC had explicitly said that they wanted to talk to him – not run any diagnostics. She would be there the following afternoon if it turned out she was needed after all. However, Bob had copied the complete log from the final four hours at HGMS onto a memory pen and was able to scan through it on his tablet. For want of anything better to do, he started on it in the departure lounge.
On the whole, the afternoon’s records were much more of the same: huge numbers of meaningless spikes, what looked like a few bits’ worth of broken data (suggested by something like a square-wave form) and, very occasionally, a longer sequence of (what might be) binary code (or it might just be random patterns) – but nothing really that he had not seen from the shorter half-hour session earlier.
One such block, however, caught his eye. Perhaps a lifetime of peering at network traffic had given him some unusual insight but one of the longer blocks had a vaguely familiar shape. The individual bits (if that was really what they were) actually looked to be organised into eight-bit bytes; he could see this from their tendency to be packed (filled up) with default ones and zeros. Furthermore, some of the early bytes even had a recognisable shape, something vaguely like a protocol flag. Could this be a real network frame?
His first thought was that Hattie may have somehow picked up a valid frame amidst the noise she was set to look for. However, he dismissed this idea quickly for two reasons. Firstly, the data was too short – it was an incomplete or aborted frame at best. Secondly, the flag and structure he thought he recognised for the would-be frame was for a protocol he knew would not be running on a network of that kind. It looked like an ATM cell but the HGMS network was running a high-speed extension of a simple EtherCAT protocol. Either way, Hattie would not have detected a legitimate frame like that. He was either wrong or it was pure coincidence or … perhaps the noise was playing ‘Shakespearean monkeys’?
The plane was landing so he gave it no more thought.
*
Aisha Davies was at home, reading from her tablet.
“A key element of creativity in humans involves abstraction. I don’t know how abstraction occurs, but we know that there is a very low energetic differential between the ‘fire’ and ‘non-fire’ states of neurons to save energy, and we also know this leads to massive randomness of firing. A likely implication is that to extract a useful signal from this sea of noise, there could be some sort of ‘regularity spotter’. This would be a neural primitive, probably at the level of description immediately above biophysics: perhaps regular firing is an energetically desirable state because it’s simpler to maintain?
“Whatever happens at the primitive level, it may have been conserved and perhaps scaled-up with variation, perhaps to high level abstraction in humans. But there are two potential problems before we can generate and manipulate abstractions. First, some emotion or affect associated with abstractions (or potential abstractions) in memory may have to be diminished or we would remain forever trapped in a deterministic world. Second, any proposed activity has to sit within the overall global imperative of all neural systems, that of using an appropriate structural-functional substrate. So I believe that the global imperative of the brain is not minimisation of surprise, error or uncertainty, but minimisation of assessments of lack of control. As regards a possible structural–functional substrate, it seems to me that an individual neuron could act as a comparator, evaluating what it expects to see against what it really sees and signalling any mismatch. In fact, this role may lie behind predictive coding formulations of hierarchical inference in the brain. This is surely rich with implications?
“I argue that all nervous systems are basically collections of comparators at all levels of description, from individual neuron-comparators up to assemblies of neuron-comparators that form ever larger comparators. In this model, comparators at all levels of description can have variable goal states and actual states and, crucially, can learn. All neuron comparators, individually or in assemblies, are seeking to minimise assessments of lack of control. Thus, I suggest that all nervous systems only really ever process control, not information, though for convenience we can treat some abstractions of control as something we call ‘information’. This commonality explains how nature can ‘bolt on’ a wide variety of sensors and actuators across the range of species – indeed, it is difficult to see how these faculties could do anything other than grow incrementally out of a
pre-existing (distributed) control–based context – and it resolves, at least conceptually, the related problem of how different processes in an individual brain can talk to each other without an overly burdensome ‘translator’.
Aisha was in conversation with an Italian colleague – or possibly ‘acquaintance’: she was unsure as they had never met physically – on the origins of creative thinking and problem-solving in humans. She had been considering some of the more abstract aspects of intelligence since writing the ‘Singularity’ piece with Andy. Somewhat unusually for Aisha, she was largely in agreement with Professor Paulo Di Iorio (her current correspondent) on where creativity came from; they were merely having a protracted discussion on how it might be measured or simulated – and it really was not getting anywhere. She typed a terse response.
“OK, that makes sense as a theory; but what advantage do we get from a model like that? How would we apply it?”
Chapter 7: Connections
The arrivals area at Luxembourg Airport was fairly quiet. Emerging from the security area, Bob was met by a suited man carrying a tablet displaying his name. He was accompanied by a woman, dressed similarly smartly. The man introduced himself as Carl, ‘his driver’, but looked and behaved as unlike a driver as anyone Bob could imagine. The woman was never introduced, either by ‘Carl’ or herself, and never said a word throughout the journey. They walked to what seemed to be a private car park. Carl held the doors open for Bob and the woman to step into the back of a large black car, with heavily darkened windows.
As they drove away, signs of RFS were becoming commonplace: every few minutes suggested an example. There was a general air of dysfunctional chaos everywhere. Automatic systems, such as lights, controls and sensors – which had always been taken for granted, were misbehaving from time-to-time. Confusion was rife, leading, in the main, to good-humoured amusement, and occasionally frustration; but accidents were also happening. Dealing with the problems was being made harder by a gradual loss of confidence in the technology being used to do it. But still, the majority of issues were short-lived: often dangerous – but transient.
