Marx’s post-marxist theory of bullshit jobs

Usually, the (shit)posts on this blog are non-rigorous ideas dumps that I consider “food for thoughts” more than anything else, but for once I want to gather a few specific pointers from Marx because I keep searching for them and it’s driving me crazy. That being said I’m still not a Marx scholar and there’s so much I have yet to read.

Labour is the start, not the end

I’m mostly interested in doing a cross-reading of Marx and Baudrillard, focusing on where they agree instead of where they potentially diverge, because I believe they have a lot in common when it comes to the direction in which capitalism is headed.

Marxism is often criticized because of his emphasis on labour value. It is certainly a valid criticism, though it was probably a pertinent analysis during his lifetime. But I believe Marx’s work contains hints of going above and beyond this concept. Incidentally, I was told that he had plans to address and develop this in the third volume of Capital (let me know if you know more!).

I want to look at the crumbs that hint at where he would have been going, especially through the concept of commodity fetishism, which goes in a direction quite opposed to labour value (and somewhat Baudrillardian). What’s interesting to me is that although Marx stipulates that the origin of value derives from labour, he does notice a trend towards empty simulations.

Paperclip Maximization

Marx famously highlights the tendency of capital to continuously grow and demand more and more. Some people (more than I initially realized) have drawn a parallel between the textbook case of a rogue super-AI “paperclip maximizer” and capitalism. The starting point for this article was the vague remembrance of Marx himself talking about this unstoppable cancerous maximization.

The purpose of capitalist production, however, is self-expansion of capital, i.e., appropriation of surplus-labour, production of surplus-value, of profit.

Capital Vol. III Part III Chapter 15. Exposition of the Internal Contradictions of the Law

He even points out to speculation as a core mechanism of this expansion.

The rate of self-expansion of the total capital, or the rate of profit, being the goad of capitalist production (just as self-expansion of capital is its only purpose), its fall checks the formation of new independent capitals (…). It breeds over-production, speculation, crises, and surplus-capital alongside surplus-population.

Capital Vol. III Part III Chapter 15. Exposition of the Internal Contradictions of the Law

That comes into play naturally because a smart decentralized maximizer system will exploit any opportunity it finds to the fullest following a revenue over investment analysis. Speculation is a trade of empty air that can bring revenue for very little (or none) investment. And it’s not bounded in the same way labour is.

The speculative side of Capital

Marx talks a lot about speculation and the trade of empty promises through the notion of fictitious capital, which qualifies floating money without concrete anchor point used to gamble and speculate in stock markets.

The greater portion of banker’s capital is, therefore, purely fictitious and (…) it should not be forgotten that the money-value of the capital represented by this paper in the safes of the banker is itself fictitious

Capital Vol. III Part V Chapter 29. Component Parts of Bank Capital

Gambling in the stock market is not the only way for capitalism to create value out of nothing, though. It also excels at manufacturing demand and creating needs where there was none before. Marx writes about this in the form of imaginary appetites:

The extension of products and needs becomes a contriving and ever-calculating subservience to inhuman, sophisticated, unnatural and imaginary appetites.

Economic and Philosophical Manuscripts of 1844

The fetishism of money

In this world of make-belief, humans then become not the goal, but the means through which this maximizer system operates. It takes a life of its own, and its goals take priority over human ones. Capital becoming its own end is echoed in the concept of fetishism.

Here the products of the human (…) appear as independent figures endowed with a life of their own and standing in a relation to one another and to people. (…) This I call the fetishism which clings to the products of labour as soon as they are produced as commodities and which is therefore inseparable from commodity-production.

The Value-Form, Appendix to the 1st German edition of Capital, Volume 1, 1867

Marx highlights that it is an eminently social manifestation (dare I say “ideology” ?).

This Fetishism of commodities has its origin, as the foregoing analysis has already shown, in the peculiar social character of the labour that produces them.

Capital Volume One SECTION 4 THE FETISHISM OF COMMODITIES AND THE SECRET THEREOF

The specifics of the fodder for this maximization process do not matter, as long as they contribute to the process.

In interest-bearing capital, therefore, this automatic fetish, self-expanding value, money generating money, are brought out in their pure state and in this form it no longer bears the birth-marks of its origin. The social relation is consummated in the relation of a thing, of money, to itself. Instead of the actual transformation of money into capital, we see here only form without content.

Capital Vol. III Part V Chapter 24. Externalization of the Relations of Capital in the Form of Interest-Bearing Capital

Form without content… Doesn’t it sound Baudrillardian? The culmination of this empty and ambivalent form, its total incarnation, is the ultimate commodity, the universal equivalent, the meta-fetish, value incarnate: money.

On the other hand, interest-bearing capital is the perfect fetish.  It is capital in its finished form—as such representing the unity of the production process and the circulation process

Theories of Surplus Value, Marx 1861-3 Addenda. Revenue and its Sources. Vulgar Political Economy

The relations of capital assume their most externalised and most fetish-like form in interest-bearing capital.

Capital Vol. III Part V Chapter 24. Externalization of the Relations of Capital in the Form of Interest-Bearing Capital

Capital, even fictitious, becomes the core of the ideology, the ultimate signifier, the ontological and teleological foundation of a society.

Petit objet de consommation

In fact, in “Pour une critique de l’économie politique du signe” (that I’ve read to prepare this essay), Baudrillard already makes the link between consumer object of economics and objet petit a of desire of psychoanalysis.

Seule la psychanalyse est sortie de ce cercle vicieux, en rattachant le fétichisme à une structure perverse, laquelle serait peut-être au fond de tout désir.

“Pour une critique de l’économie politique du signe”

He gets pretty upset at Marx because, to him, there is no such thing as “non-fetish”. Only the form and appearances ever mattered to begin with.

Il apparaît alors que le « fétichisme de la marchandise » s’interprète, non plus selon la dramaturgie paléo-marxiste, comme (…) une force qui reviendrait hanter l’individu, coupé du produit de son travail, (…) mais bien comme la fascination (ambivalente) d’une forme (logique de la marchandise ou système de la valeur d’échange)

“Pour une critique de l’économie politique du signe”

His main point of contention is that the concept of “use value” is already filled with implicit capitalist ideology, that there is no such thing as a need that is not socially mediated. To him, in a capitalist system, all apetites are more or less imaginary.

Marx dit en substance : « La production ne produit pas seulement des biens, elle produit aussi des hommes pour les consommer, et les besoins correspondants. » Proposition détournée le plus souvent dans le sens simpliste de la « manipulation des besoins » et de la dénonciation des « besoins artificiels ». Il faut voir que ce que produit le système de la marchandise dans sa forme générale, c’est le concept même de besoin constitutif de la structure même de l’individu

“Pour une critique de l’économie politique du signe”

But he himself acknowledges that this is not in opposition to Marx’s worldview, but rather its logical conclusion pushed even further.

C’est ici que joue l’idéalisme marxiste, c’est ici qu’il faut être plus logique que Marx lui-même, dans son propre sens, plus radical : la valeur d’usage, l’utilité elle-même, tout comme l’équivalence abstraite des marchandises, est un rapport social fétichisé.

“Pour une critique de l’économie politique du signe”

My gut feeling is that Baudrillard is just a few steps ahead of Marx on a not-so-dissimilar path. Maybe the path’s starting point as labour value was indeed always an illusion, and maybe not, but I do feel that the two thinkers come together about the destination.

Very late stage capitalism

So where did Marx see this path going? What did he think the outcome of these trends was going to be? He famously wrongly (as of yet) predicted the unavoidable uprising of the proletariat and the end of capitalism under its own contradictions (thwarted by the hard work of the CIA, probably xD).

But in a way, despite his attachment to the notion of labour, even he foresaw the famous development illustrated by Baudrillard, the move away from concrete and meaningful towards speculative and simulative, which ends up dominating everything:

But it is evident that with the development of the productive power of labour, and thus of production on a large scale: 1) the markets expand and become more distant from the place of production; 2) credits must, therefore, be prolonged; 3) the speculative element must thus more and more dominate the transactions. 

Capital Vol. III Part V Chapter 30. Money-Capital and Real Capital. I.

It’s almost foretelling a future of pure imaginary speculation completely decorrelated from labour. That is the same conclusion reached by Baudrillard who observes that there is no more value in things.

Il n’y a plus de scène de la marchandise : il n’y en a plus que la forme obscène et vide. Et la publicité est l’illustration de cette forme saturée et vide.

“Simulacres et simulation”

In a future (present?) where meaning is dead, only appearances and their (speculative) simulacra matter anymore.

Il n’y a plus d’espoir pour le sens. Et sans doute est-ce bien ainsi : le sens est mortel. Mais ce sur quoi il
a imposé son règne éphémère, ce qu’il a pensé liquider pour imposer le règne des Lumières, les apparences, elles, sont immortelles, invulnérables au nihilisme même du sens ou du non-sens.
C’est là où commence la séduction.

“Simulacres et simulation”

The price of information

I want to mention briefly here an interesting perspective that I got while reading Kurzweil: the trend of production cost. As things become easier and easier to produce, they get cheaper. Eventually, in an age where 3D printers manipulate molecules, matter becomes interchangeable. The value of a product will not be so much in the actual particules composing it but rather in the information that describes how to make it.

This is perhaps the strongest defense of the move of capitalism towards a service-based society and an information-centric economy. But is it really what’s been going on?

Bullshit jobs

There’s a case to be made that our society has moved further and further away from concrete value into imaginary speculation. More and more human time is dedicated to busy-work, empty talks of hot-air and other bullshit jobs, to use the term consecrated by David Graeber.

Baudrillard concurs with this vision whose best portrayal remains for many the movie Office Space.

Il en est de même du travail. L’étincelle de la production, la violence de ses enjeux n’existent plus. Tout le monde produit encore, et de plus en plus, mais subtilement le travail est devenu autre chose : un besoin
(comme l’envisageait idéalement Marx mais pas du tout dans le même sens), l’objet d’une « demande » sociale, comme le loisir, auquel il s’équivaut dans le dispatching général de la vie. (..) le scénario de travail
est là pour cacher que le réel de travail, le réel de production, a disparu.

“Simulacres et simulation”

Bullshit jobs and busy work proliferate as a distraction, to counterbalance and overcompensate the disappearance of meaning. And more hot air begets more hot air.

The IT crowd

One domain where this is especially clear is information technology. Although some of the work in technologies is undoubtedly useful, there is a clear trend of “keeping oneself busy” by constantly reiterating on the design of your favorite products for no reason at all.

This article is already pretty long, so I’ll keep examples to a minimum. I’m sure you have encountered plenty already in your life: features get added, removed and added again, apps split and merge, brands go through endless cycles of redesigns and rebranding, everyone makes their own version of the same thing… Every bug fix introduces two new bugs… Games are now services… When’s the last time you’ve seen a software that was actually finished?

Some of it is doubtless legitimate security arm’s race, but that doesn’t account for everything. Truly, we could not have dreamt of a better vessel than software and apps for busywork. I bet 99% of the code written in 2013 was already replaced 5 years after.

This resonates with mankind’s natural taste for nostalgia and familiarity, most notably in the videogame industry. Scores of companies make the exact same product hundreds of times. Cycles of remakes are shortening, companies are releasing the same game over and over again… Do we really need yet another Final Fantasy 7 game, or Marvel movie? How does it even feel to be working on these? What percentage of jobs are really essential? I wish a fraction of this energy went into making products standard and backward compatible instead 😦

A great reset

And yet, people cling to work as the main core of their identity, as if they had any meaning. Capitalism does, after all, need the anguish of unemployment to thrive. This anguish has now turned into psychosis, with everyone talking about the great reset, a fairytale in which migrants steal people’s precious jobs and ways of life. It’s killing me that ironically, in a weird distortion of this ludicrous fiction, AIs are actually coming to replace people’s jobs and ways of life.

But are most jobs even worth saving to begin with? There’s no shortage of calls for a new society, where humans would be freed from work and would stop defining themselves through their jobs. As for me, I keep looking at the incredible cost (human effort, political risks, environmental…) of keeping this intricate system of bullshit and illusions alive, and I can’t help but wonder if this energy might be better spent…

But I also know that meaning is extremely important to human life and political polls show clearly how adverse to change humanity is. So as much as I would like to overthrow capitalism and found our own meaning making system, I think that might not be entirely realistic. What might be, though, is to embrace bulllshit work, and give one to everyone, so that they can feel accomplished while AI takes over what actually matters. That’s probably where we’re headed, TBH. But then, please, let’s maybe leave a way out of this system for the people who don’t want to be defined by work.

Making a self-aware twitter AI with GPT2

The story so far

It was more than a year ago that I had my playing with gpt2 phase, resulting in a short story co-written with the AI and this little blog http://yo252yo-a.tumblr.com/ which I kinda stopped updating after a while.

But I was bound to come back to it some day! It all started when I decided to open a twitter account for my podcast. I very naturally made a little script to schedule all my tweets (from Google Spreadsheet ^^) so that I could enqueue tweets, obviously. I also went back in time to the archive of my facebook/tumblr/whatever posts to see what could fit this new account since I posted so much enlightening things over the years xD

Once this was in place, it was like my twitter account was managed by a nice little bot (who was simply posting things from a queue, but still). As its parent, I obviously wanted to see it grow: how cool would it be if it could learn and evolve by itself? Could it ever be self-aware (lol)? After all, it already had access to twitter, and it had a collection of my tweets to learn from.

Setup

So I dusted off my colab repository of GPT2, since GPT3, despite all the hype, remains pretty inaccessible. Most notably, I had to make it work with an old version of tensorflow (the recent versions broke it), and I also made it read and write directly to Google Spreadsheet /o/ In the end, I only had to run the code in the colab to fetch the data, train on it, and post it directly in the queue to be twitted. Pretty sweet setup.

The problem is that GPT2 produces mostly crap. And I didn’t know what temperature or training set would be ideal for my purposes. It was time to experiment!

Validation

I ran several training sets on several temperatures. For each, I personally annotated 200 results. I dont think the result will be super significant, but it’s better than nothing.

The success criteria was: is this tweetable (i.e. relatively grammatically correct, at least a bit interesting/surprising, and of course different from the training set). The good samples will be posted on our twitter with the hashtag #shitmygpt2says.

Training sets

The basic training set was the queue of all our tweets for the podcast twitter account, including the archive of all my past tumblr/facebook posts that I sanitized for the occasion (a lot of work xD).

But like my previous attempts, I thought it was a bit sad to limit myself to things produced by me when I had the perfect chance to merge my brain with the people I admire. Furthermore, I kinda wanted to make my twitter AI standalone and able to “learn” as time passes, even though GPT really isn’t the best framework for that ^^

I ended up making a twitter list of people I admire, and used their recent tweets in my dataset. The idea was to make my model aware of “recent events”, recent words, etc…

Yet, I wanted to keep a feeling that the writing style was distinctly mine. It is accounted for in the success criteria, and the core of this experimentation was “how should I mix the training set to keep awareness of the recent world but still control the style of the output?”.

Sequential vs merging

In my previous attempts, I mostly used a “merging” approach feeding everything to the learning phase. The alternative is to feed two corpora in succession during the learning phase.

From what I observed, it seems that GPT2 absorbs the style of whatever it was fed last, even if it is for very few training epochs. For instance, when I fed it corpus A for 1.5k epochs and then corpus B for 100 epochs, it produced results that looked like corpus B, even though it exhibited some signs of having learned A every now and then (pretty rarely though, that’s why I kept so many epochs in the first phase of training).

I kinda think of it with a cooking metaphor, when I first marinate the model in corpus A and then lightly sear it with corpus B.

Here are the experimental results that loosely validate this:

We notice here btw that the merging strategy is pretty poor because consistency of the training set is pretty important with GPT2. The first three lines did not exhibit a strong difference, making me believe that 1k epochs is enough for GPT2 to “forget” about the initial corpus, which is how I ended up with the 1.5k/100 mix which gave me the best outcomes.

Best parameters

Here is the total result of my experiments. GPT2 produces around 93% of crap, which makes sanitizing a pretty tough job ^^ It appears that this could drop to 80% or below by using correctly the “marinade/searing” technique and keeping the training set uniform.

As it is widely known, temperature below 0.8 is pretty bad, but I find myself often pushing above 1, though it seemed to do pretty poorly with my best data sets. I’ll keep using different temperatures as they produce different types of results that I enjoy in their own way. But I’ll probably stop using text corpora as a base (past writing, night vale scripts, etc…) because they don’t seem to bring anything to the table (and could even be detrimental, better stick to tweets).

So we’re pretty far from a self-aware AI that learns from its mistakes, but seeing that I’ll always retrain it on recent tweets, and that it will be trained on my own tweets that include the proposals it made and I kept, I hope that as time passes it’ll still learn to be a bit better (it already started annotating posts with the #shitmygpt2says hashtag itself).

In the future, I’ll run this every now and then in its best configuration, and keep posting on twitter with the hashtag #shitmygpt2says. Stay tuned if you’re interested!