Tuesday, 28 January 2025

AI as an Apparatus of Capture

You must have a heart of stone not to laugh. Hundreds of billions of dollars wiped off the value of leading US tech firms on Monday's markets as their expensive plagiarism machines, the misnamed language models trading under the name 'Artificial Intelligence', have been out done and out-performed at a fraction of the cost by the Chinese alternative, DeepSeek. Having announced untold levels of investments in new chips, new data centres, and new power stations, this vast supportive infrastructure designed to assure the dominance of Silicon Valley firms for the next century has seemingly evaporated as venture capital flees elsewhere. Egg on the face of far right oligarchs, and humiliation for Donald Trump's efforts at hitching his backward-looking project with the wave of the future.

America, the self-proclaimed home of buccaneering entrepreneurship and freedom stopped exporting advanced chips to China and showered the techbros with staggering sums. The ludicrous resources at their disposal brute-forced the pace of innovation. Think of all the money that has gone into producing applications less accurate than internet search engines, creates its own "facts" through "hallucinations", and has trouble drawing hands. A hell-for-leather effort that wanted to cement an unassailable technological lead. And that's been shown up by a network of Chinese coders using open source software, no state backing, and little but their skill and ingenuity. They have produced a better model with nowhere near the same demands on chips and energy. The tech oligarchs have got to be hoping the PhD-level language models, which according to the much-publicised rumour mill is weeks away, is more than a case of their getting high on their own supply. If not, some are looking at significant dips in their fortunes.

Yet none of this changes the fundamentals about so-called AI. It is a technology designed not to speed things up and make life easier. Like the vast majority of innovations in production, it's designed to deskill and disempower labour, forestalling the looming threat of a long-term shift in the balance of capital and labour in labour's favour. As discussed many times here, capital accumulation, particularly in the most advanced industrial countries, has become increasingly dependent on immaterial labour: the knowledge, skills, personalities, and social aptitudes. This is because capital itself is directly dependent on the production of information and relationships to sell a service of some description. However, the main force of production that counts is the brain, or rather the person of the worker. And it's inherently leaky. Despite the stringencies of copyright, industrial secrets, non-disclosure agreements, whatever competencies someone builds working with one employer remains with the worker, and can be the basis for future work elsewhere - be it for other employers (competitors), their own business, or outside the economy entirely. Capital ponces off the commons and whatever is produced "leaks" back into and builds the commons, leaving capital revealed as a parasitic social relationship. Whereas previous waves of technological development have increasingly done away with the masses of people production required and, more importantly, the knowledge of production that gave labour some leverage over the capital that employed them; the creativity, the ability to produce new information and maintain relationships appeared impervious, except around the edges where repetitive tasks were concerned.

AI changes this. The tech bro dream are machines that can code the software for other machines, which to a degree is possible now. They want to be able to generate their own images and text without having to bother with human artists and writers. Or the production of films, television, and music that can do without crews, actors, and musicians. In corporate and public sector bureaucracies, it's the automation of clerical tasks and, in some cases, front-facing work involving customers and clients that promises the most, freeing up managers for "strategy". Who themselves will become increasingly replaced by "thinking" machines. Even everyday communication skills, such as how to craft an email, is now something any old AI chatbot will happily do for you. What only a few short years ago was temporarily "captured" to generate surplus value is in the process of becoming absorbed into itself as 21st century fixed capital. AI might seem convenient, but it's first and foremost there at capital's convenience.

Opposition to so-called AI is not reactionary, or anti-technology. It is a healthy response to a power grab that will result in a privately-owned monopoly over creativity. This means, ultimately, to put hundreds of millions out of work globally and secure capital against its dependence on the intellectuality and sociality of human beings. These bots are "trained" by "reading" the sum total of our species' cultural output without any recompense whatsoever. What is presently ours becomes theirs, with the possibility of our common heritage being reduced to regurgitated AI slop that becomes a cultural staple, and one owned by the firms who end up winning the AI race. The loss of, the possibility of the privatisation of social competencies is real. Therefore, the rapid rise of DeepSeek and the humbling of the USA's richest companies, while funny, does not change the dynamics of the situation at all. Because this is capitalism, AI is more than a toy and a liberator of free time: it's the latest, and possibly the most complete means of capturing and imprisoning the soul.

11 comments:

Darren said...

What kind of materialist witters on about the soul? “AI” is large scale industry to service work’s manufacture. It turns the computer from a tool to a machine and its purpose is the real subsumption of service industry labour. If it can be made to work decently (and Deep Seek seems like a step forward) it is just another advance in the means of production to be expropriated from the expropriators. To the extent that it renders much of the drudgery of routine service industry work unnecessary, it is welcome.

Phil said...

What sort of materialist can't recognise the non-spiritual uses of the word in everyday parlance?

Braingrass said...

Machines can't consume and therefore ultimately destroy capitalism from within. If there are no workers to buy stuff, how do you make a profit?

Anonymous said...

Pretty sure that a certain materialist talked of vampires, hobgoblins and spectres.

Anonymous said...

What I can't understand is why there is no (or very little) demand from workers and/or their representatives (in which I include, precariously, the Labour Government) for AI to be used to shorten the working week. All the media stories are about how many thousands of jobs will be replaced, why no push back arguing to reduce the working week.

In the seventies, when robots were first being introduced into car factories, I went on strike for a shorter working week. We were successful and forced the Ford Motor Company to share the benefits of automation with its employees. Why aren't the Union's leading such demands today?

Anonymous said...

If AI does result in mass unemployment, that’s not going to promote growth. Yet Labour are keen to promote it. Something doesn’t add up.

Darren said...

Fair enough. Apologies all round for being testy. But I think the whole AI discussion tends to leap to hyperbole and the rhetoric around this latest form of automation needs to be brought down to earth. As the next comment (from the same?) Anon suggests and as I tried to point out after my out of order snark, there are threats and opportunities.

Kamo said...

I work in an area of Technology which has been exploring AI applications for a few years, so I've been keeping an eye on its development out of self-interest as much as self-interest.

A few thoughts:

Firstly, take claims around DeepSeek with a pinch of salt, it's likely that it had more assistance than claimed. I think AI had become a bit of an asset bubble, DeepSeek has burst it. Some stress about Bond villain-esque tech bros, capitalists stress over fact that majority of gains from innovation fall to consumers not capitalists, the salience of Zuckerburg, Musk, Bezos et al obscures the utility found in their products by hundreds of millions.

Secondly, AI has useful applications, but is better at certain tasks than others; millenarian panic is unnecessary. LLM are just very clever prediction engines based on pattern recognition and probability, the more formulaic and predictable something is, the better they are, but they also generate (sometimes superficially plausible) shite which can be dangerous. So you still need people to verify and validate outputs. If they are trained on gargbage they produce garbage; DeepSeek doesn't like subjects the Chinese Communist Party doesn't like.

Thirdly, new technology always creates frictions, the electric lightbulb brought huge global benefit but was terrible for gas lamp lithters. In the case of AI, it's a threat to lower value, predictable, formulaic tasks, and it's the least creative 'creatives' who are at risk. Churnalists and opinionists who churn out stock output. The 'grievance porn' genre of The Gruan could easily be replaced by AI without anyone noticing, they follow standard themes, with formulaic and predictable patterns, lack depth, originality and critical thinking. Plenty of political opinion could go the same way, the less genuine insight or intellectual rigour the easier it is. On the other hand a more talented writer could use AI to form competing narrative outlines and then adapt and redact them using their tacit knowledge and insight, making the process deeper, richer and more efficient. It could supercharge talent whilst drowning mediocrity.

Fourthly, and finally, none of this matters really matters for UK policy. Even if compute costs can significantly be reduced the UK still has the highest industrial energy costs in the world, so relatively we're just not competitive. And when you consider that the payroll tax increases announced in the last budget will hit smaller orginsations hardest, the odds are against innovative challengers developing out of the UK

Zoltan Jorovic said...

AI accelerates the self-destruction of consumerism. Unless they can somehow get the bots to pay each other for services, how will the bulk of humanity pay for whatever the bots are 'creating'? The current virtual world is funded largely by advertising and monetising people's data. But if people are largely rendered redundant (apart from nail bars, barbers, delivery drivers & baristas - until the robots do that too) and so deprived of the means to pay for anything, where is the 'market'?

Anonymous said...

Their hopes may well be technologically in vain. Big tech assumes the only form of growth is exponential, but AI's capabilities seem to grow logarithmically. In other words: diminishing returns as they throw more data and more compute power at it. And therefore generative AI appears to have largely peaked:
https://m.youtube.com/watch?v=dDUC-LqVrPU

Anonymous said...

The very nature of growth in AI is inherently haphazard. Fits and starts punctuated by "eureka" moments which catch almost everyone off guard. If you understand the nature of the tech, then you understand how it can't be otherwise. But unless the whole shebang hasn't yet attained critical mass, and collapses into another "winter" (probably the terminal one, other things being what they are), then the eureka moments are going to keep coming.

If you want to see the real "point" of today's AI, then follow the scientific news. The sciences, especially biological, are being literally remade by it; the new frontiers opening up simply couldn't be reached any other way. The extent to which this depends on the "Big Tech" AI edifice is debatable, but it isn't zero.