The problem with randomness

There are many public utilities available, but a newly proposed one looks to be founded in some incredibly interesting science. Scientists all around the world are working on producing public sources of randomness, which could be used for a variety of purposes. These attempts have led to some very thought-provoking questions: how can you create something random when you must use rules to define it, and how can you be certain it actually is random?

To begin with, how do we define randomness? It’s a simple enough concept, but putting it in scientific terms is a bit trickier. It is the level of disorder and unpredictability in a system, with pure randomness being fundamentally unpredictable. Imagine, for example, you were looking at a string of truly random numbers forever: the odds of getting any given number would be exactly the same, and thus each subsequent number would be unpredictable (as opposed to something like throwing two dice, where the outcome is still unpredictable, but probability makes certain outcomes more likely than others).

Scientists all around the world are working on producing public sources of randomness, which could be used for a variety of purposes

The problem with randomness is that most of the ways we can simulate it are not, in fact, random at all. Take tossing a coin – in everyday practice, it’s a random result. However, if you knew the pressure applied to a toss, the turbulence it would encounter in the air, the distance it would cover, etc., you could determine exactly what would happen. We could apply the same criticism to almost any random-number generator, which use complex algorithms to come up with the numbers. They are efficient, but also deterministic – if you knew the rules, you’d be able to reproduce the exact same sequence later on.

So, where do you go from here? Some random-number generators have begun creating sequences based on physical processes, rather than by means of an algorithm. These normally use microscopic phenomena that generate low-level and statistically random (no recognisable pattern) noise – for example thermal noise, or noise generated in the atmosphere. However, these are not without issues either. Some of these generators use radioactivity but, as the radioactivity degrades, so does the randomness of the numbers – this is often difficult to detect in practice too. They’re also fairly inefficient, taking a while to come up with each number, and they suffer from the same issue as the coin toss – nature is governed by laws and patterns, and so it is theoretically possible to use the data independently to come up with the random numbers.

Some random-number generators have begun creating sequences based on physical processes, rather than by means of an algorithm

Perhaps, then, it was time to cast the net further afield, drawing on more and increasingly different inputs to create randomness. The US National Institute of Standards and Technology (NIST) has provided a public randomness beacon, released in July of this year, which draws on a variety of sources to provide random numbers. They began their search in the quantum realm – at its heart, quantum mechanics is a random theory, with only probable instances of things like radioactive decays being possible.

The current NIST beacon has, as part of its function, a quantum-based random-number generator, which measures the arrival time of photons produced by a laser. It combines these results with algorithm-based and noise-based generators, to create random information that should be completely random. Other countries have started work on random beacons, and there is an option to combine the NIST results with random bits from other beacons. As an example, Chile has its own version, which uses (among other things) seismograph data and Twitter feeds fed through a cryptograph. Private companies are also getting in on the action, with one called Cloudflare extracting randomness from lava lamps.

Other countries have started work on random beacons, and there is an option to combine the NIST results with random bits from other beacons

So, this is all well and fun, but what use is this randomness in the real world? Well, it could theoretically be applied to anything that requires decision-making, from choosing a jury to lottery winners. One NIST scientist, Rene Peralta, who runs one of these beacons, had some suggestions: it could be used to ensure random selection of recipients for US visa programmes, eliminating potential bias. She also suggested that it has uses in random testing of voting machines – in a large sample, overseers can’t examine every machine, so only a random few get tested. If it was decided algorithmically, any hackers could just figure out which machines would be tested and target different ones, but these beacons would help pick unpredictable subsets.

Although we speak of randomness in an everyday way, actually producing it is a lot harder than it looks. What’s going to happen with it now? Well, if this is true randomness, expect that to be an entirely unpredictable answer!