For the last decade, and particularly the last five years, artificial intelligence (AI) technology has undergone an exponential growth in its capacities. AI is dominating high-tech industries, providing the basis of technologies ranging from self-driving cars to facial recognition in large-scale surveillance systems. It has even been used by Deepmind (a Google subsidiary) and Elon Musk’s company OpenAI to master the extremely complex games of Starcraft 2 and DOTA 2, beating the best professional players in the world.
Indeed, one influential AI scientist and former Baidu (the Chinese equivalent to Google) chief scientist, Andrew Ng, argues for good reasons that “artificial intelligence is the new electricity”. It is not surprising, then, that AI has attracted considerable attention from some of the world’s most powerful people. Russian President Vladimir Putin ominously stated last year,
Artificial intelligence is the future, not only for Russia, but for all humankind, […] It comes with colossal opportunities, but also threats that are difficult to predict. Whoever becomes the leader in this sphere will become the ruler of the world.
And this sentiment is shared by the billionaire tech capitalist Elon Musk, certainly not a fan of government regulation, who surprisingly advocates for it in the case of AI,
[There] needs to be a public body that has insight and then oversight to confirm that everyone is developing AI safely. This is extremely important. I think the danger of AI is much greater than the danger of nuclear warheads by a lot and nobody would suggest that we allow anyone to build nuclear warheads if they want. That would be insane…
So, what exactly is so dangerous and powerful about AI technology? Musk is worried that a super-intelligent AI will take over the world. One would expect that someone so integrated into AI tech development would be more mature than to worry about the extremely unlikely scenario (according to experts) of super-smart killer robots, but perhaps he indulges in too much science fiction. More likely, he doesn’t care about the real material impacts that AI development will have on billions of working class people, as we will discuss in this article.
What makes AI technology powerful?
The exponential growth in the power of AI technology is largely due to the theoretical development of deep learning, in conjunction with increased computing power and access to big data.
Deep learning is based on neural network architectures, a large series of nonlinear functions that are capable (in theory) of approximating any real world function by just observing it. In other words, given a huge amount of data, neural networks can learn to uncover the patterns in that data. This includes learning to translate between languages, learning to play the complex ancient board game Go better than the best players in the world, and distinguishing you from your friends’ faces on social media.
This technology was able to advance largely due to the foundational research of two Canadian scientists: Geoffrey Hinton (inventor of the learning algorithm called backpropagation) and Yoshua Bengio (inventor of many different useful neural network architectures). Bengio has obtained more citations in the scientific literature over the past five years than Albert Einstein has achieved over the past 120 years, and Hinton has more than doubled those numbers.
Their theoretical achievements were able to be realized in practice due to advances in computing technology, without which deep learning would be impossible to implement. Related to how Bitcoins are mined with graphics processors (GPUs), deep learning also requires high-capacity GPUs to train AI algorithms over 200 times faster than with a normal processor. The company Nvidia develops this technology, and has a virtual monopoly with 81.2 per cent of the market share for GPUs. This leads to incredibly high prices, and universities must pay them to develop their labs, lest they fall behind in the race to publish papers on AI.
Yet, perhaps most decisive in cultivating the power of deep learning and advanced GPUs is the advent of big data. The internet now allows for huge amounts of data to be collected and extracted with ease by companies and researchers. Everything from the text on Wikipedia, to images of faces on social media, to purchasing history, is all data that is used by companies and researchers to learn complex nonlinear patterns of human behavior. Without data, neural networks cannot learn.
In China especially this has been introduced to an unprecedented degree due to the lack of privacy guarantees for its citizens. According to tech capitalist Kai-Fu Lee, WeChat is now a “universal social media app” that collects and centralizes all data ranging from communication, to individual profiles, to consumer payments with QR codes. All of this data is and will continue to be used to extract and predict more and more complex patterns of human behavior, ranging from market activity to political understanding to friendly interactions.
In summary, the power of AI comes from new algorithms, new computing technology, and huge amounts of easily available data.
Real-world impacts of AI
No expert in AI believes that Skynet will take over humanity anytime soon (or likely ever), despite Elon Musk’s predisposition for sci-fi fantasy. Real, immediate consequences, however, are the typical consequences of what Marxists call the development of the means of production.
Indeed, AI is playing an integral part (along with other important technologies such as quantum computing) in ushering forward a new technological stage in human history, what scientists and economists call the “fourth industrial revolution”. Global consulting firm McKinsey & Company recently released a report detailing that, while the introduction of steam engines offered a 0.3 per cent increase to annual GDP growth between 1850 and 1910, they expect AI to raise annual GDP growth by 0.8-1.4 per cent by 2065. Another massive consulting firm, Accenture, anticipates that AI will offer a $14-trillion boost to global GDP by 2035.
However, these bourgeois predictions should be taken with a large grain of salt. Under capitalism, the development of the means of production is always accompanied by job losses, and AI is no exception. Kai-Fu Lee, who is certainly no Marxist (but much more sober-minded than Elon Musk!) predicts,
As deep learning washes over the global economy, it will indeed wipe out billions of jobs up and down the economic ladder: accountants, assembly line workers, warehouse operators, stock analysts, quality control inspectors, truckers, paralegals, and even radiologists, just to name a few. … Rising in tandem with unemployment will be astronomical wealth in the hands of the new AI tycoons [e.g., Google, Facebook, etc]. … [We] will begin to see greater and greater concentration of these astronomical sums in the hands of a few, all while unemployment lines grow longer.
The AI world order will combine winner-take-all economics [i.e., capitalism!] with an unprecedented concentration of wealth in the hands of a few companies in China and the United States. This, [Lee believes], is the real underlying threat posed by artificial intelligence: tremendous social disorder and political collapse stemming from widespread unemployment and gaping inequality.AI Superpowers, 2018, pp. 53-58
In other words, under capitalism, AI means the rich get richer while the workers enjoy increased exploitation, unemployment, and social crisis. As well, contrary to what many predict, wide-scale introduction of AI will not lead to unheard of levels of growth. In an epoch of capitalist crisis like the one we live in today, the capitalists are incapable of using the amazing technological advances to deliver economic growth. The fact is that there are already historically low rates of what is known as “capacity utilization,” which means that the capitalists are already unable to use the productive capacity at their disposal.
This is because world capitalism is saddled with a crisis of over production. The introduction of AI would only exacerbate this by making millions of jobs obsolete thus reducing the consumer market. This is the fundamental barrier to growth today. If the capitalists cannot sell their products, they will not invest and the economy will enter crisis. Only under a democratically planned socialist economy could we harness the amazing potential of AI for the benefit of all.
Impacts of AI today
We don’t need to wait for the future to see how AI will impact us—it is already playing a disruptive role in the lives of workers. We experience unprecedented levels of privacy violation: from systematic surveillance via cameras with precise facial recognition technology, to implicit bias and racism in algorithms used for purposes such as credit assessments and the criminal justice system. Indeed, an algorithm may assign a longer prison sentence to a black person because the data suggests “they are more likely to reoffend.” In addition, every person who uses social media is well aware that everything they do is invasively tracked by advertising companies to a creepy degree; this is because all of your data (your likes, dislikes, buying history, etc) is held and monopolized by private companies like Google and Facebook.
AI also motivates large-scale exploitation in new forms. Recalling that these AI algorithms require huge amounts of human-labelled data to work well, it’s not surprising that the extremely mind-numbing work of so-called “data farming” has become a new, prominent source of income for private companies. AI startups in Finland have adopted the creative business strategy of hiring prison labourers to do their data farming. Meanwhile, Amazon has created “Amazon Mechanical Turk” as a platform for extreme exploitation of humans to do data farming for companies and universities, where the workers make a median wage of $2/hour from the safety of their home computer. In China, “data factories” are becoming ubiquitous, and their workers make between 10-20 yuan ($1.47-2.94) per hour.
In the meantime, the CEOs of big tech companies are purchasing their own private islands, and their companies are recording record-breaking 20, 30, and 40 per cent profit margins.
A Luddite or socialist solution?
A first instinct by many workers may be to simply get rid of AI altogether, to smash the computers and prevent our jobs from being taken away. At best, this can only serve to momentarily delay the advent of a new technology. In the 19th century, the rage of Luddite workers was characterized by destruction of factory equipment in protest of their deplorable working conditions, but this did nothing to stop the industrial revolution.
A genuine workers’ movement would not call for the eradication of this technology, but for the seizure of the means of production by the working class. If we democratically control this technology, then it could be used for the collective benefit of society as a whole. For example, as AI improves productivity, we could simply decrease the work week according to the productivity increase, not increase unemployment! But, you cannot control what you do not own. Currently, just a handful of billionaires and corporations reap most of the benefits of new technology, while millions become unemployed and hungry.
As a development of the means of production, the technological powers and opportunities offered by AI could be used and harnessed for large-scale social good. Rather than determining the best algorithm to advertise selfie sticks, AI could be used to facilitate large-scale economic planning to meet the needs of the majority. Indeed, central planning is already done within big capitalist monopolies such as Walmart and Amazon (which have larger GDPs than many small countries combined), a phenomenon that was predicted and observed already by Vladimir Lenin in 1916. These corporations recognize that central planning is the most efficient and profitable method, but in these instances the wealth generated is privately accumulated by the wealthy executives and stockholders, whereas economic planning based on social ownership would benefit the vast majority.
With the October revolution of 1917, the Soviet Union was created and began the task of building a socialist society. Economic planning was introduced which allowed the soviet union to grow by leaps and bounds unseen before. This was prior to the dawn of computers and cellphones, and yet in a relatively short period of time, the USSR went from an illiterate feudal society to the second most powerful global superpower of the 20th century following WWII. This was of course in spite of the bureaucratic distortions introduced by the Stalinists who crushed the workers democracy and usurped political power.
Nowadays, with the massive computing power, algorithmic development and data available, the task of economic planning is incomparably easier than in the USSR more than 100 years ago. Economic production, management and distribution could be systematically introduced into society with the help of these new technologies. This would be accompanied with large-scale democratic decision-making, already available (in part) due to the mobile phones sitting in nearly every person’s pocket, each of which is more advanced than the entire computational power the Soviet Union ever had access to. Such economic planning would be able to free humanity from the absurd destructive anarchy of the free market, allowing society to finally control production for human need, rather than to be enslaved by production for private profit.