Page updated: 7 January 2021
Ever-faster and better computers can improve SAR operations. But it's not just about collecting more data...
The pace of technological change is accelerating exponentially. However, it's not all driven by smaller transistors any more.
Until now, much of the acceleration in computing power was driven by the ability of those making microprocessors to halve computing chip size and cost every 18 to 24 months. Known as Moore’s law[i], this phenomenon has led to the incredibly fast growth we have seen in computing power without proportionate increases in expense and has led to laptops and pocket-size gadgets with enormous processing ability at fairly low prices.
In recent years, some commentators have expressed concern that the accelerating pace of technological improvement could not last, given that the size of computer chips could not shrink indefinitely. Yet, while transistors cannot keep shrinking indefinitely, we are still seeing major advances in the speed and capability of computers. McKinsey estimates that the current pace of change is 300 times the scale, and with roughly 3,000 times the impact, of the industrial revolution.
How is this continued increase in computing power possible?
Moore’s law itself does appear to be coming to an end. However, recent advances in quantum computing and 3D molecular computing are being predicted by some theorists as being likely to continue the trend of improved computational capacity over time.[ii]
The reason for the continued rapid pace of technological evolution is that other technologies are increasingly being used to squeeze more performance out of the same basic materials. Further speed and performance improvements are increasingly now being achieved, for example, through:
Of these options, quantum computing was originally considered the most unlikely to succeed. However, in September 2019, Google announced that it had successfully performed the first complex computation only possible on a quantum computer. This signalled a major step forward in computing power. Quantum computers can perform calculations hundreds of thousands of times faster than classical computers. While quantum computers are not about to become widely available any time soon, their creation means that certain previously unsolvable statistical problems may now be able to be cracked.
[i] Dorrier, Jason. “Will the End of Moore's Law Halt Computing's Exponential Rise?” Singularity Hub, SingularityHub, 7 July 2017, singularityhub.com/2016/03/08/will-the-end-of-moores-law-halt-computings-exponential-rise/
[ii] “Moore's Law.” Wikipedia, Wikimedia Foundation, 3 Oct. 2017, en.wikipedia.org/wiki/Moore%27s_law.
Much faster computers like quantum computers offer the potential to help solve the so-called ‘travelling salesperson problem’. Also, better ability to process data rapidly can make it easier to find people faster.
Faster computers can help find the quickest route to a destination
The ‘travelling salesperson problem’ involves the trying to find the most optimal route to multiple locations most quickly.
Consider, for example, a situation whereby Coastguard, LandSAR and/or Surf Life Savers are all required in multiple locations simultaneously. These so-called Mass Rescue Operations (MROs) are (currently) “low-probability, high consequence events that require a response to provide immediate assistance to a large number of people in distress (Greenberg, Testing New Zealand’s Readiness for a Mass Rescue Operation 2018).
What is the best way in these situations to deploy resources so as to save as many people, as quickly as possible?
It turns out that the possible options for how to coordinate activities quickly become impossibly large for classical computers to find the best route.
Quantum computers, by contrast, will be able to definitively answer statistically-challenging questions like how best to deploy multiple resources to multiple locations in quick succession very soon.
And they can also help find a needle in a haystack (or a person in an ocean)
Improved ability to process data is important for all businesses. However, it is particularly relevant for the SAR sector because agencies need to sort through vast amounts of data to rapidly pinpoint people in distress (for example, a stricken boat in the middle of the Pacific Ocean).
The ever-improving ability of computers to see and process large volumes of data, along with better satellite and GPS capabilities, will keep enabling improved performance in future.
More immediately, the SAR sector has recently significantly improved its ability to understand those in need of search and rescue, and how to better serve them, by introducing SARdonyx.
This new integrated information system replaces existing single-agency data stores with a multiple-agency solution which brings together information on all SAR operations.
This will no doubt improve both the ability to take a strategic perspective across all partner agencies and enhance cooperation and decision-making across the whole sector.
Artificial Intelligence (AI) may be able to super-charge the ability of the Search and Rescue (SAR) sector to find people more quickly.
Virtually every business on the planet is now struggling to make sense of the sheer quantity of raw data available. Data itself is not much help unless it is turned into insights that can improve decision-making. This is why AI might be worth exploring for the search and rescue sector.
Given the rapid date of improvement in computing power, there may be major enhancements to the SAR sector operationals which could be enabled, starting from the new capability offered by SARdonyx.
In particular, there may be opportunities to partner with organisations like Google or IBM, both of whom are heavily invested in Artificial Intelligence (AI) technologies.
In recent years there has been an explosion of interest and investment in one particularly useful element of AI, called machine learning.
Machine learning happens when a machine, over a period of time, gets better at a task it wasn’t explicitly told how to do. It is now possible to program computers to learn through trial and error, the same way a human does, but with the added advantage of being able to process vast quantities of data (e.g. spotting anomalies in large volumes of satellite images).
If you can measure it, you can improve it. But computers that learn can do both much faster.
There have been plenty of recent demonstrations of how powerful machine learning is enabling computers to eclipse human capabilities in certain situations. For example, machine learning computers have already been able to beat world chess champion Garry Kasparov, ‘Go’ strategy champion, Lee Sedol, and the best players of ‘Jeopardy’ in the world.
Using the information collected and integrated in SARdonyx, as well as the wealth of case studies and rescue footage data collected by SAR agencies, it may be possible to bring a machine-learning approach to search and rescue.
Such an approach could help identify new and improved ways of both preventing harm and conducting search and rescue operations most efficiently and effectively in future.
Of course, when considering the rapid pace of technological improvement, it is important to acknowledge that across the NZ Search and Rescue Region, there are vastly different economies and capability to take up technology.
Samoa and Tonga, for example, do not have the same new technology purchasing power as New Zealand, just as New Zealand does not have the same adoption potential as Australia. In this context, we can expect the adoption of technology across the Search and Rescue region to be somewhat variable in the future.
However, just as Australia or New Zealand occasionally purchases a rescue vessel for smaller nations like Niue or Tonga now, smaller nations across the region may have expectations or aspirations to access new technology with the support of larger players like New Zealand.
Artificial intelligence, big data and powerful new algorithms offer significant opportunities to improve SAR operations and management by enabling fresh insights.
Critical to every SAR operation is intelligence. As described in the section on the SAR process above, the NZ Police and RCCNZ put a lot of work into gathering and interpreting information to inform a successful SAR operation.
As one NZSAR Secretariat staff member put it
“if a SAR goes wrong, it’s usually because we end up looking in the wrong spot, are sent with the wrong asset or not enough assets. It’s all about the decisions made by the coordinating authority.”
In the past few years, the cost of storing data has plummeted while the ease with which it can be captured has increased exponentially.
However, the storage of data itself is useless. In order to become useful to assist decision-making, data needs to be turned into information, and information into knowledge. Currently, only around 0.5% of the world’s data is actually analysed[i]. The NZSAR sector has, itself, identified the potential for more of its own data to be analysed and used to predict and prevent the need for SAR.
This is already being done in certain parts of the NZSAR sector (e.g. the Mountain Safety Council and Maritime NZ already take a data analytics-led approach to much of their prevention activity).
Computer assistance is no longer just restricted to data analysis. In particular, in the last five years, significant advances were made in the fields of computer vision, speech recognition, and language understanding.
All of these advances were made with the assistance of narrow artificial intelligence (AI). In the future, we can expect computers to get exponentially smarter at understanding, reasoning, seeing and learning. In other words, they will get more useful for the real world.
In particular, the improving ability of computers to capture, store and analyse vast amounts of data cheaply can considerably improve the quality of decision-making. This is partly because computers, with their ability to store information perfectly and rapidly analyse vast amount of it, have access to insights that humans alone do not.
However, it takes active effort on the part of organisations to actually capture the relevant data, store it, analyse it, then inform decision-making with it.
As AI becomes much more able to interpret data, it may be that certain elements of the SAR intelligence process could be streamlined and fast-tracked further by AI systems. In particular, AI systems are increasingly proficient in areas relevant to SAR, including perception analysis, decision-making, mapping, path planning and victim detection.[ii]
For now, the NZSAR sector appears to be struggling to keep up with the rapidly increasingly available technological advances. For example, LandSAR have commented:
“We lag considerably behind in how we capture, store and use SAR data. Even new projects…reflect old style thinking and are unlikely to get us to where we could/should be. The problem lies not so much in the technology, but in the human willingness to engage with it, to its full potential.
[It is]…limited by the availability and inclination of operators to collect and enter data (i.e. If we were to collect all the salient data, the job would be too big and people would avoid doing it). Even once the data is available, there is a widely varying degree of ability to apply meaningful insight to it. Using the data to discern useful patterns (such as key success / effectiveness factors) requires subjective judgements that can be lacking.”[iii]
Another area identified by the NZSAR sector as potentially useful for further exploration and investment is capturing lessons from operations and sharing them (e.g. to support training efforts).