The future and robots

Submitted by SJW on 14 January, 2020 - 5:46 Author: Charlie Applebaum

Fuelled by rapid developments in technological innovation hyped in recent years, although mostly developed over the last two decades, many cerebral types suggest we may be at the start of some significant changes in capitalist production. They even gave it a grandiose name: “The Fourth Industrial Revolution”.

Socialists, Marxists, progressives have a history of taking technology and advocating its use for more than just the most efficient exploitation. Perhaps however, the pace of innovation is making this harder. The techy elite, a traditionally well-meaning liberal bunch, and the revolutionary socialist crowd tend not to have massive overlap. I would argue that both could probably learn a thing or two from each other. Current and near term software and hardware have properties which many outside of the tech space struggle with and, we, as conscientious socialists need to understand them to advocate their role in a fairer society. I will outline some key innovations: open source software Roughly, this is software that the users have the freedom to run, copy, distribute, study, change and improve. Thus, “free software” is a matter of liberty, not price.

To understand the concept, you should think of “free” as in “free speech,” not as in “free beer”. The logics and algorithms that control our lives should not be a black box owned and run in secret by wealthy corporations. There is some evidence that this situation is improving. One recent industry survey concluded “Open Source is the foundation now for nearly all applications….. Open Source development has gone from the exception to the rule”. Millions of dollars of software is now available for free on the internet, but we must recognise most of the means of production are still run with non-free software. OSS is also interesting in an economic sense in that as a digital goods it has the unusual property of zero cost for duplication, and just look at the history of napster to see how capitalism struggle with these emerging types of commodities. Whilst OSS growth is good, socialists should be advocating for much more, starting with urging government to open source all of its work. Most socialists are familiar with the idea of a workers’ cooperative or public ownership but fail to see how non-free software continues to be used for exploitation and profit. If really is going to be a widespread replacement of people by machines in workplaces then exploitation and inequality will increase more if we do not get control of the software that will be controlling our lives. “The Cloud” To understand the cloud, you need to understand what came before it.

Previously, when an organisation built a new IT venture it took weeks to get it running. Someone would have to select hardware, wait for delivery, assemble and configure it before a developer could run a new application on it. Then they would set up monitoring, backups, redundancy, and add batteries and generators to keep it running during outages. Contrast that now with one new cloud service, Zeit. Once installed, in a few minutes you can type “now” at a command line and everything — servers, DNS, databases, backups, storage — are all provisioned and published on the internet in 5-10 seconds. Widespread automation in IT jobs, the low hanging fruit for automation, has already happened. Compute, the ability to run some arbitrary software of your choice, is now available to society with about as much friction as getting water out of a tap. Whilst most people cannot do their own plumbing, the end product of running water is ubiquitous and low cost, and Compute is no different.

The bedrock of modern automation is cloud based servers, the cheapest of which is around four dollars a month (AWS). Capitalism has driven costs down to the point where it could be free at the point of delivery if society wished it, at least in those countries with reliable power and connectivity. However we must be cautious. Take market leader Amazon — it’s business model is no longer to become the most successful online retailer. That was achieved years ago. Consider the rate they are investing in logistics, physical stores and even media. They recently purchased Whole Foods in the US, in part for their massive number of high street stores and large logistics network. Amazon want to be the very fabric and infrastructure of capitalism. Underpinning every purchase, payment, fulfilment and delivery, taking a slice at every point, with little democratic oversight and of course, terrible labour conditions.

Modern capitalism has granted us the tools to run all software, all computing, upon this amazing infrastructure for almost nothing. What would socialists do with this digital infrastructure? What might capitalism do next? You only have to look at the drone-like picking staff in Amazon’s warehouses to feel a chill, where routes, pick rates and break times are commanded by digital supervisors running on cheap cloud platforms and hard to automate tasks like shelf picking are done by low-pay workers.

It turns out that managers are far easier to automate than workers, and tasks like picking are still among the most challenging. The first robot, named Unimate, was made in 1954 to move hot car parts for GM. Now reflect on the recent deluge of robot sci-fi — Westworld, Ex Machina, Her, Black Mirror — in this vision of the near future autonomous human like robots will be stronger than us, smarter, bordering on self awareness … and mainly used as human shaped sex toys. Sadly the reality of lifelike robot companions is a long way off. The most advanced human like general purpose robots are basically rubbish, and about as sexy as a Dalek. After 60 years of slower-than-expected progress in robots, recently acceleration can be observed, often funded by military research. However, most of the developments are less to do with the machinery itself — the actuators and motors — and more to do with developments in control software. The technology needed for impressive automation has been advanced for many years — just think of car production lines. However in the production lines of the most profitable company on earth, it still takes over three hundred pairs of hands to make one iPad. Until very recently we lacked the capability to operate the robots with sufficient sophistication without human control, but developments in AI software have started to change this.

For example, take the exciting subject of sorting of mixed recyclables. Once done by armies of humans leaning over conveyor belts, now image recognition cameras and high pressure air hoses can sort recyclables with greater efficiency. Sort quality improves as hard to classify objects are done by human operators and the machine learns over time. These developments suggest impressive but boring single purpose robots moving out of the factories and further into our lives.

And like the flat screen TV, expect the cost to be extortionate at first and plummet as economies of scale kick in. There also seems a tendency of underestimating the old or mundane, and exaggerating the impact of the shiny and new. Take the washing machine — utterly boring — and yet probably the most empowering labour-saving robot humanity has created so far. For every article you read of dystopian robotic futures and widespread unemployment remember the washing machine. Expect automation in the next 20 years to look a lot more like a washing machine than a sexy robotic butler. Driverless cars are probably the next big significantly disruptive robot.

There is one major thing holding back the armies of robotic workers: the nagging red flash of the “charge me” indicator. Until very recently batteries have not had dramatic increases in energy density in decades. Modern lithium ion batteries were invented in the 70s, commercialised in the 90s, and had little major investment until recently. Smartphones, electric cars like Tesla and “green energy” markets have incentivised corporate R&D and gained greater state sponsored research. Papers published in 2016 suggests a 10x improvement in energy density is possible. When this becomes mass market it opens the door for many more types of automation. Battery technology has historically shown slow improvement compared to processing power, but by the time your iPhone can last one full week maybe you will start to find your job will not.

Much like cheese and Brexit, AI comes in two flavours — hard and soft. Hard AI, or artificial general intelligence, is the ability to apply knowledge to solve unseen problems. To be able to fully translate a book you need to fully understand the reasoning of its author. While tools like Google Translate are improving quickly we are a long long way from this type of automation being perfect. In 40 years time however many experts think hard AI will start overtaking humans in most general tasks. Soft AI, or artificial specific intelligence is far more immediately relevant. It is based on statistical learning on large sets of training data to solve complex problems in narrow well defined fields. Think Netflix’s video recommendation engine. So far soft AI has rarely replaced human workers, instead tackling problems at a scale where human labour is not economically viable. If it took you an optimistic 10 seconds to review the viewing habits of a Netflix user and recommend them a new show, it would take you 70 years of nine to five work to complete the full task just once. In fact, in this example we have actually created new jobs as someone needs to build and maintain the recommendation engine rather than the AI replacing human labour.

Here are some examples of soft AI. Google deepmind can already learn and complete the computer games from your childhood with zero human input. As good or better than you in 49/57 games after a few hundred attempts. No human intervention required. News being written by machines. Speech and image recognition took massive leaps in the last year. Subtitles, and categorisation are almost solved. Drones. While the machinery is not new, the brains are. Combined with the ubiquitous fast reliable network and better batteries and the potential for automation are interesting. Self driving cars. They are already here!

Whilst profit is the principal motivation for decision making, a machine is usually chosen to replace a human when the cost savings far outway the typical drop in quality. Beyond pro-chess, it is actually pretty hard to find real world examples where current generation AI is actually better than humans and already be capable of replacing workers. Two interesting cases that have been studied and proven to be significantly better done by a machine are the role of a pharmacist in handing out medicine and spotting complications (but who stacks the machine!), and that of a lip reader where machines already outperform humans on average error rates. Impressive stuff no doubt but not quite the doorstep of mass unemployment that many fourth industrial revolutionists prophesise. More commonplace are examples of technology increasing efficiency of a worker when automation is employed as assistive tooling for existing human labour. Soon doctors will be being advised by digital assistants reviewing case notes and analysing patient biometrics for them. This is far less likely to lead to sudden mass unemployment but does present an issue for workers who are less tech literate as job roles change faster than skills, and employers have little motivation to retrain adults under capitalism.

Trade unions have so far addressed these sorts of issues on a very low level, if at all. We should expect this problem to increase and so the job of training workers by unions needs to be prioritised. Jobs that can be boiled down to repetitive tasks that can be scored and quantified are most at risk, and these tend to be medium skill/middle management type responsibilities — accountancy or paralegals — but significant change is likely to take a decade or more. When we look at the role of technological change from history it is sensible to expect unpleasant disruption for the affected workers even if society generally fares better. Ultimately in many jobs as machine reliability becomes statistically better than the occasional ingenuity of human insight we will see workers being replaced.

Crystal ball gazing is not the job of serious socialists, but let us consider how we might use the spoils of capitalism were profit not the motivation. Say self-driving electric vehicles become widespread. If that was combined with an open source software platform that was publically run in the open and funded and monitored democratically, you could have an efficient public transport system, resembling Uber, that largely runs itself? Usage or payment could even be managed with digital currency until private transportation becomes a nostalgic pursuit and the public demands transportation as a basic right. We already have all the knowledge and hardware to make this happen so it’s not so far fetched as it seems.

The machines are coming There is at least some evidence to suggest the “new jobs” created by the changes may not be enough for full employment under capitalism. Workers in transport, retail roles, cashier, bookkeeping and supervisory work are all easy targets for automation. There are also concerns about a the “hollowing-out” of middle-skilled, middle-wage jobs and “a corresponding rise in employment at both the high and low ends of the skills spectrum”. AI, limitless compute and new battery tech suggest that workplace automation is going to increase.

Whilst I disagree that the scale will be “unprecedented”, not least because the transition costs are likely to be prohibitive and underestimated, there seems to be some merit that the speed of change will be faster than we have seen before. There are some obvious socialist answers here — raising the minimum wage and reducing the working week would help. Bill Gates has suggested a robot tax — but history shows that solutions that will benefit everyone are unlikely to be achieved without significant political will from the majority of people. Socialists need to have more to say on current technological innovation.

Even under a capitalist society the potential for vast improvements in quality of life are huge, and as noisy progressives it is our responsibility to understand them and persuade others of their importance. It is not sufficient to outsource the thinking on technology to the techy liberal elite. These are the tools that socialists will use to liberate the majority of humans from drudgery. We must understand them.

Add new comment

This website uses cookies, you can find out more and set your preferences here.
By continuing to use this website, you agree to our Privacy Policy and Terms & Conditions.