Monday, November 18, 2019

Pathfinders: Socialism – there’s an app for that (2019)

The Pathfinders Column from the November 2019 issue of the Socialist Standard

Funny how, if you do a column for long enough, you can meet yourself right back where you started. When this column began, in January 2005, the very first article asked if the popular computer simulation game, Sim City, could ever be used to create a realistic model of a global socialist society in operation. As there wasn’t a computer big enough to do this at the time, we suggested distributed processing using a global network of home PCs crunching data in the breaks when their users weren’t at the keyboard. In theory this might have worked, but in any case we had no suggestion at the time for how to make the model sophisticated enough. Simply consider one average human being, and the range of possible actions open to them in any given situation, and the variables quickly become enormous. Multiply those by the population of the world, and the task was beyond incomputable. We threw the question out there anyway, knowing we were asking for the moon on a stick.

Well, Moon, it’s time to meet Stick, because things have changed. If 2005 doesn’t seem that long ago, remember that the Sim City article appeared four months before the first ever YouTube upload, ‘Me at the Zoo’, by Jawed Karim. Facebook was just a year old and only 5-7 percent of people in western countries used social networking sites. In the same month Microsoft released its XP Professional operating system. Reddit was launched in June. Twitter, Tumblr, Instagram and Snapchat did not yet exist.

Since then, raw computing power has increased by orders of magnitude. The advent of big data, collected through trading sites and social networks, has created a new science of mass behavioural analysis. Artificial intelligence, given clear rules and parameters, can now out-think any human on the planet. We are starting to get thinking and planning tools that are unimaginably faster, and involving data sets that are vastly bigger than anything conceivable even in 2005.

A recent article in New Scientist shows just how far things have come, with a new generation of simulated models which are able to plot predictions at the crowd or mass level on the basis of individual behaviour. Multi-agent artificial intelligence (MAAI) allows ‘predictions to be made with extraordinary accuracy by testing them in highly detailed simulations that amount to entire artificial societies’ (5 October). This may sound far-fetched, but it’s being done now. ‘MAAIs are already being used to build digital societies that simulate real ones with uncanny accuracy’.

Instead of primitive top-down social models, MAAI uses agent-based modelling, in which individual ‘agents’ are ‘programmed to interact with one another and their virtual environment and change their behaviour accordingly’. One early non-AI model was developed to predict the spread of the Ebola outbreak in 2014, using known parameters ranging from demographics to disease pathology to cultural factors such as burial rites. At the same time researchers developed ‘what if’ interventions to see what might impede the disease spread. Without interventions, the model predicted 1.4 million infections. In the event, smart interventions suggested by the model kept the figure down to 28,000. This doesn’t prove the model was responsible per se, or that unknown factors didn’t play a role, but it is nevertheless powerful evidence in favour.

Even so, the model had to be kept simple, with a very limited range of individual behavioural options, because that was all the available computing technology could cope with. Instead of near-zombies, what researchers really wanted were ‘intelligent agents’ able to emulate the behaviour of thinking and acting for themselves.

MAAI is delivering just this. ‘One of the things that has changed is an acceptance that you really can model humans,’ says one researcher. ‘Our agents are cognitively complex. They are simulated people with genders, ages and personalities. […] They’re social in the way humans are. They learn from each other, react to each other and to the environment as a whole.’

You might be impressed if they could do this at the scale of a village. In fact the technology can already model a city as big as London, and the plan is to scale it up to a population the size of the US, then China, and ultimately the world. Just as so-called Industry 4.0 is introducing the digital twin, whereby a whole factory can be managed and monitored via its virtual equivalent using a vast array of sensors attached to every moving part, so it should soon be possible to ‘build an artificial society, try things out and see what works’.

Today’s researchers are understandably thinking about models which address questions internal to capitalism. But the possibilities for socialists are as dazzling as they are obvious. What if you could model a global, democratic, non-market society of common ownership? What might it look like? Could there be different but workable versions? What forms of direct or representative democracy would be most feasible and at what scales? Which forms of science and culture might bloom and which might wither on the vine? What might we lose, and what gain, without the cruel driving force of money? People comprehend what they can see with their own eyes. We could potentially bring the concept of socialism to life the way a 3D chart brings a table of data to life, and in ways we haven’t even thought of yet.

But even if we could do all this, would it necessarily convince anyone? However sophisticated the model becomes, it is still just a digital model. As such, it doesn’t prove anything in the real world. But socialists, like scientists, know that they can never expect to get absolute, concrete proof of anything. The most we can do is amass such a weight of evidence that people are gravitationally inclined towards it. MAAI will help us do this. Make no mistake, other political and commercial groups will use it, for their own manipulative purposes. We can use it too, but with no nefarious objective and with our code and parameters open to scrutiny. If our model reveals ways in which socialism might go wrong, or break down, say in circumstances of large-scale harvest failures, we would certainly want to know in advance. But if we can demonstrate that socialism works as a stable system, without the kind of wild fluctuations you get with market societies and the kind of inequality, wars and environmental damage that the market also produces, then it will be a lot harder for people to dismiss out of hand. Maybe we could get it on a phone app. Somebody down the pub says: ‘Nah, it would never work, mate, not in a million years’. You say ‘Really, you think so? Then have a look at this…’
Paddy Shannon 

No comments: