What David Graeber and David Wengrow’s “The Dawn of Everything” has in common with sci-fi economics

A few weeks ago, I finished David Graeber and David Wengrow’s The Dawn of Everything. It is not science fiction, nor it is economics. And yet, I propose it has to do with both. The book begins by stating inequality has risen to the top of the agenda for public debate. Scholars, politicians,  business leaders are calling attention to it. Where, they ask, does inequality come from? How can we reduce it to tolerable levels?

The authors then proceed to question these questions. A lot of people believe there is this thing called “inequality”, and it comes from somewhere – it was not always there. This belief is cultural, and we can investigate it. So, Graeber and Wengrow’s research question is not “where does inequality come from?”. It is “where does the question about where inequality comes from come from?”

It is a move typical of anthropology. In fact, Graeber does the exact same thing in the opening chapter of Debt: The First 5,000 Years . He finds himself at a party, and the conversation falls on the 2011 Greek debt crisis. A guest remarks that yes, the Greek people are suffering, but “debts must be paid”! And Graeber wonders: why do people think that? where does this belief come from? And off he goes.

I am a big fan of Debt. I read it twice, back to back, and went back to it several times to re-read inspiring digressions from the main theme. While worth reading, Dawn is not as good as Debt, in my humble opinion. Its arguments are not as thorough, and it tends to treat absence of evidence as evidence of absence. But it does make an important contribution: it replaces the question on the origin of inequality with a better question. That better question is: how did we get stuck?

It works like this. The question about the origins of inequality implies a linear social process. Back in humanity’s hunting-gathering days, the story goes, all men and women were equals. But then, about 10,000 years ago, humanity shifted to agriculture. This created regimes of private property, cities, complex societies, wars for resources and élites to appropriate their surplus. So, we got inequality, and we are stuck with it, because it is the price to pay to have a complex society.

The problem with this story is that is not borne out by what we know. Graeber and Wengrow’s data consist of archeological findings and ethnographies of indigenous societies encountered by European settlers during colonial times. And these, the authors tell us, agree: there is no linear process from equality to inequality. Ancient societies appear to have experimented with many models. Foragers experimented with farming, then let it go. Large urban settlements arose in the absence of agriculture. Farming societies remained egalitarian for centuries. Cities dominated by élites abandoned the construction of pyramids and temples to embark on large-scale social housing projects. There are even documented societies that lived in towns, and farmed, during the winter, and in small bands of hunters gatherers during the summer.

The authors insist that all this happened because the people that made up those societies wanted it to. They were politically sophisticated and reflective. They knew that they could shape their institutions in ways that preserved their freedom and well-being. A part of getting this right was making sure people did not have too much power over one another: in our modern terms, that people would equals.

Ancient societies were not stuck. Its members were free to roam, and were subject to very little cohercion. But entire societies were also free in another sense, that of shaping arrangements that made people’s lives better. Instead, we moderns are very stuck.

And that brings me to the Sci-Fi Economics Lab. When my partners and I dreamed up the Lab, in 2019, we had no idea what Graeber and Wengrow were up to. Yet, like them, we felt stuck. We felt crippled by our inability to imagine living under any system other than late-stage capitalism. To heal, we turned to the imagined futures of science fiction, and to economics as an angle of attack. It was a good choice: we have come some way to freeing our imagination. Today, sci-fi stories set in the world of Witness give us a glimpse of everyday life in post-capitalist systems. Graeber and Wengrow appear to have taken the opposite route, into humanity’s distant past. But we share the same curiosity, and the same conditional optimisms, and I know them for kindred spirits.

What does an art project have to teach us about facial recognition technology?

Reposted from Edgeryders

Paolo Cirio is one of my favorite artists – I saw some of his work at Ars Electronica a few years ago. His latest project (late 2020) is called Capture, and works like this.

  • First, the artist collected 1,000 photos of public protests in France.
  • Next, he zoomed and cropped the faces of police officers in those photos.
  • Third, he printed out these photos and stealthily stuck them up on the walls of Paris.
  • Fourth, he ran facial recognition software on the photos. Hold this in your head, it is important.
  • Fifth, he created a website where people could identify police officers: they were served pictures, and could enter a name.

The project turned out to be… explosive.

  • The police’s trade unions protested, as Capture was putting officers at risk.
  • The Minister of the Interior, Gérald Darmanin, tweeted his disapproval.
  • An exhibition of Capture scheduled to happen at Le Fresnoy, was canceled as the controversy mounted. Note: Le Fresnoy is in a city called Tourcoing, and Darmanin used to be the mayor of Tourcoing, stepping down during the preparation of the exhibition. Le Fresnoy denies receiving any outside pressure.
  • Cirio left France altogether to avoid “retaliation”.

Now, if you are like me, you’ll be thinking: wait, what does it even mean for an artist to “run facial recognition software” on a photo? It’s not like Cirio, or any of us, is sitting on a dataset with identified picture of everyone in France. In the absence of such a dataset, even the best facial recognition software is going to draw a blank, every single time.

And, well: this is why Cirio is a famous artist, and I am not. The reactions of the representatives of the police has been one of disgust, like they had been contaminated by something filthy: someone had somehow grabbed their photos, and run them through facial recognition software! They had been violated! Rationally, this is nonsense, but emotionally it is very powerful, and therein lies Cirio’s artistic gesture. *Capture* shows us the instinctive loathing and disgust of humans when been intruded upon, when people are looking at you and you were not ready for it. We want to own the image we project.

Thanks to this work, I learned the extent to which facial recognition is already out there in Europe. And, like Cirio says, it’s unevenly distributed: the police can hide as they use violence, but for ordinary citizens the camera is always on. France is one of the European countries where its usage is most widespread. In this 8 minutes video, French Internet lawyers call for an outright ban on facial recognition technology. This is also Cirio’s position – he even engaged the European Commission on this (and the Commission promised to look into it).

It seems that now the Commission gets to walk the talk, because Commission VP Margrethe Vestager is participating in the Us-EU Trade & Technology Council, and the Reclaim Your Face campaign is pushing for a blanket ban to biometric surveillance. We’ll see how that pans out.

Anyway. FWIW I highly recommend Capture. Here is a short presentation video made by the artist:

Photo taken from the artist’s website, unattributed.

Sociopathic innovation: how we are investing most in the most evil technologies (LONG)

TL;dr

Artificial intelligence and the blockchain are the two main technological hypes of the past fifteen years. Both were hailed as technologies with the potential to solve many problems and change the world, for the better. It now looks like their impact is overwhelmingly negative. Though they could be used for the common good, it turns out they are not very good at that. They are better, far better at harming humans than at helping them. They encode dystopian, sociopathic world views; and tend to attract developers, investors and entrepreneurs that share those world views. So,  once deployed, they tend to bring the world closer to them. They are sociopathic tech.

This is disturbing, because mostly everyone fell for them: investors, developers, entrepreneurs, academics, government officials. I call for a re-examination of the achievements of these technologies and the impact they are having on our life and our societies. I would like to see support to innovation systems depend on how new technologies improve the well-being of humans and of the planet, and only on that. In what follows, I review some of the facts as a discussion starter.

Of how Artificial Intelligence excels at everything, except solving problems that matter

I recently had the opportunity to be exposed to the work of Juan Mateos-Garcia, a leading data scientist. Juan and his team had been looking at a large dataset of science papers published on the topic of Artificial Intelligence (AI). Their results look like this:

  1. AI has been undergoing a revolution since about 2012, when deep learning started to systematically outperform established techniques.
  2. Scientific production (papers) is booming. AI is shaping to be a general-purpose technology, like electricity or computing itself.
  3. Industry interest is evident. Many top scientists have been recruited from academia into industry. Venture capitalists have moved to invest in AI startups. Major governments are underwriting large public investments. There are talks of a “AI arms race” between China, the USA and the EU.
  4. AI is dominated by a small number of organisations and geographic clusters. Diversity of its workforce has stagnated.
  5. AI has had no impact on the effort to beat back the COVID-19 pandemic. In fact, all other things being equal, a paper on COVID is more likely to be cited by other papers if it is not about AI.

This final point gave me pause. Something was off. Why would AI not make a valid contribution to fighting the COVID plague? The conditions all seemed to be in place: there was, and still is, plenty of funding for research on COVID. There is a large, inelastic demand for the applications of that research, like vaccines. There is plenty of training data being generated by health care institutions the world over. And, if AI is a general purpose technology, it should apply to any problem, including COVID. The most exciting technology of the moment somehow failed to contribute to solving the most pressing problem of the moment. Why is that?

I can imagine a world where AI is deployed to help in the fight against a pandemic. Wen would use it to engineer a more targeted response to the risks of contagion. Granular risk scores could be associated to individual people and different situations, allowing society to protect the most vulnerable people from the riskiest situations, while leaving low-risk individuals in low-risk contexts free to get on with their lives.

Sounds good, but that world is not the one we live in. In our world, AI-powered, individually customized COVID restrictions would run into non-tractable problems. First, the algos would seize the high correlation between different socio-demographic variables, and decide that poor people, people of color and (in America) trumpists are more prone to the contagion, and should stay at home more than white, affluent liberals. Discriminated groups would react fighting back, challenging the algos as biased, starting litigation and inviting to civil disobedience, as is happening time and time again.

Even if there was no conflict and everybody trusted the algos, it is not clear how we would use effectively the predictions they make for us. First of all, there is the cognitive challenge of understanding the predictions. You could tell someone something like this: “the risk of catching COVID on public transport for someone with your demographic profile went up 20% today, avoid the bus if you can”. But that is unlikely to work, because

  • Most people do not understand risk. For example, they are more scared of terrorist attacks than they are of car crashes, though the latter are far more frequent (hence more dangerous) than the former.
  • AI is Bayesian statistics, and as such it makes prediction not on you, but on somebody who is like you in a quantifiable way. It leaves out everything that makes you unique, putting it in the error term. For example, imagine you are a 45-year old living in the Netherlands who is also an ultra-marathoner. The algo computing your risk factor processes your age and the country you live in, because it has thick enough data in those dimensions. Your ultramarathons stay in the error term, because there are not enough people doing ultramarathons for that activity to be tracked in its own variable. And yet, when looking at the overall resilience of your organism, this is clearly an important information.

Given this situation, I suspect most people would end up following their own belief system rather than the algo’s recommendations. People who fancy themselves strong and resilient might say “yes, this gizmo is predicting high risk, but it is not talking about me, I am healthier and stronger than most!”. Or, vice versa, “yes, a low risk is predicted for outdoor mingling, but with my history of allergies I still don’t feel safe”. This is de facto happening right now with how people process scientific findings about COVID-19. Some people prefer to trust their own immune systems over the pharma-punditry complex. Others made COVID restrictions into some kind of weird religion, following them “above and beyond” even when science is calling for their relaxations. Even if a good AI-powered risk prediction system were in place , many humans are way too irrational to take full advantage of it. They prefer simple rules, applicable to all: “1.5 meters”, “wash your hands” and such. The promise of AI, providing personalized recommendation to each and every one of us, clashes with the human need for stability and security.

In conclusion, AI had no grip on COVID, and is unlikely to have any grip on any similar high-stakes problem. So, what is AI good for? We can start with the applications already being developed:

With the exception of machine translation, these applications are all detrimental to human well-being, for world-eating values of “detrimental”. We are seeing yet another example of Kranzberg’s First Law in action: AI is not good, nor is it evil, nor is it neutral. It could be used for good, though I am unconvinced it would work very well: but it is when you use it for evil, dehumanizing purposes that it really shines.

That such a potentially toxic technology is attracting so much attention, public funding and private investment is a spectacular societal and policy failure. And that brings me to the blockchain.

Of the blockchain and its discontents

The blockchain, as by now everyone had to learn, is the name of a family of protocols that allow data storage not in a single repository, but in many. Using cryptography, the different computers who adhere to the same protocol validate each other’s copy of the database. This prevents a “rogue” participant from altering the records, as the alteration would only be present in a single computer and not be validated by the others. This system was first proposed to solve a problem called double spending when no trusted, centralized authority is present.

That was 2008. In these 13 years, blockchain solutions have been proposed for many, many problems. To my knowledge, none worked, or at least none worked any better than competing solutions that used a more conventional database architecture. This makes sense, because blockchains are self-contained systems. They use cryptography to certify that in-database operations, but they cannot certify anything that exists outside the database. Any system based on a blockchain relies on external sources of information, known as “oracles”. For example, if you were to build an identity system based on the blockchain, you would have to start by associating your name, date of birth etc. to a long string of digits. Once stored on the blockchain, the association is preserved, but some external “oracle” has to certify it before it gets stored. In the absence of a credible external certification, the system could work technically, but it would produce no impact. I could create my own identity system, but no one would use it, because I am not trustable enough when I issue a digital ID to your name. There are entities with the trustability to start such a system, for example major governments. But, because they are trustable, they do not need the blockchain at all. I have lost count of technologists who told me:

Any technology which is not an (alleged) currency and which incorporates blockchain anyway would always work better without it. (source)

But the blockchain is not just another clever technical solution in search of a problem to solve. I argue it is a major source of problems in itself. Consider this:

  • The distribution of Bitcoins is extremely unequal, with a Gini coefficient estimated at 0.95 in 2018 (theoretical maximum: 1; Lesotho, the most unequal country on the planet for which we have data: 0.65). In fact, inequality seems to be a feature of blockchains, not just of Bitcoin – for example, it is estimated that the bulk of the monetary value conjured by Ethereum-based non-fungible tokens (NFTs) is appropriated by “already big-name artists and designers”.
  • Blockchains use a lot of power. Every update anywhere in the system needs to be validated by network consensus, which includes a lot of computers exchanging data. Bitcoin alone consumes about 150 Terawatt per hour, more than Argentina. Providing computer power to the Bitcoin network is rewarded in Bitcoins, through a process known as “mining”: this provides the incentive to underwrite all this computation. In  bid to make what they see as easy money, Bitcoin miners have resorted to malware that infects people’s computers and gets them to compute SHA-256 , incorporated into the builds of open source software projects; resurrected mothballed power stations that burn super-dirty waste coal; installed mining operations in Iranian mosques (which get electricity for free) and engaged in plain stealing. Their carbon footprint is enormous: one Bitcoin transaction generates the same amount of CO2 as 706,605 swipes of a Visa credit card. Some blockchains have less computationally expensive systems of verifications, but they are still more energy- and CO2-intensive than traditional databases.
  • Fraud – especially to the detriment of less experienced investors – is rampant in crypto.
  • Crypto has provided a monetization channel for ransomware attacks. Ransoms are demanded and paid in Bitcoin, untraceable by Interpol. Some observers go so far as to claim that the price of Bitcoin is tied to the volume of ransomware attacks. Hospitals and other health care institutions are among the main targets of these attacks: not only do they have to pay money, but their IT systems shut down, threatening the lives of patients.
  • In 2021, tech companies that used to donate CPU power to legitimate projects have had to stop doing so, citing the constant abuse from crypto miners. It is worth quoting the words of Drew DeVault:

Cryptocurrency has invented an entirely new category of internet abuse. CI services like mine are not alone in this struggle: JavaScript miners, botnets, and all kinds of other illicit cycles are being spent solving pointless math problems to make money for bad actors. […]  Someone found a way of monetizing stolen CPU cycles directly, so everyone who offered free CPU cycles for legitimate use-cases is now unable to provide those services. If not for cryptocurrency, these services would still be available. (source)

In return for this list of societal bads, so far, all the blockchain has to offer is a plethora of speculative financial assets: a casino. Which is also a societal bad, if you, like top innovation economist Mariana Mazzucato, believe that the economy is overfinancialized, and policies should be put in place to roll financialization way back. The blockchain is, overall, a net societal bad: it consumes resources to deliver a casino. Humanity would be better off without it. The picture gets even grimmer when you consider the opportunity costs: blockchain startups have gobbled an estimated 22 billion USD in venture capital funding from 2016 to 2021, very likely matched by various forms of government support, and that money could have it been used in more benign ways.

So, what’s going on here? Kranzberg’s First Law, yet again. The original group of developers that rallied around Satoshi’s Nakamoto White Paper had a libertarian ideology: they dreamed of a trustless society, where contact is reduced to a minimum and anonymised, and were obsessed with property rights. So, they built a technology that encodes those values, which in turn attracted more people than believe in those values. Code is law, they said. If someone can technically do something, that something is allowed, even moral, under some kind of tech version of social Darwinism. When the DAO was hacked in 2016, exploiting vulnerabilities in the Ethereum blockchain, the perpetrator bragged about it: if I stole your money, it’s your own fault, because code is law. I am just smarter than you, and I deserve to walk away with your money.

Trustless societies do exist – the mob is one of them. But they are not a good place to live. Economists and social scientists think of trust as social capital, and seek ways to build it up, via accountability and transparency. Again, the blockchain could conceivably be used for something good, but in practice almost all of its uses contribute to making the world a better place, while making money for the top 0.1% of crypto holders. This is because the tech itself embodied evil values, and because the social coalition behind it upholds these values. Don’t take it from me, take it from open source developer Drew DeVault:

Cryptocurrency is one of the worst inventions of the 21st century. I am ashamed to share an industry with this exploitative grift. It has failed to be a useful currency, invented a new class of internet abuse, further enriched the rich, wasted staggering amounts of electricity, hastened climate change, ruined hundreds of otherwise promising projects, provided a climate for hundreds of scams to flourish, created shortages and price hikes for consumer hardware, and injected perverse incentives into technology everywhere. (source)

Or writer and designer Rachel Hawley:

NFTs seem like an on-the-nose invention of an anticapitalist morality play: a technology that delivers exponential gains to those already at the top by convincing everyone to collectively imagine that free, widely distributed artwork is actually a scarce commodity, all while destroying the actual scarce resources of our planet. (source)

Or economist Nouriel Roubini, testifying to the U.S. Senate:

Until now, Bitcoin’s only real use has been to facilitate illegal activities such as drug transactions, tax evasion, avoidance of capital controls, or money laundering. (source)

Of how and why we are bad at supporting the right innovation

Why are the two most hyped technical innovation of the past 20 years, the blockchain and artificial intelligence, diminishing human well-being instead of enhancing it? Why are we investing in things that make our problems worse, when the world it facing environmental collapse? My working hypothesis is that the financial world will put money into anything that promises returns, with little humanitarian concerns. They lead the dance; and governments the world over have been captured into supporting anything that promises GDP growth. If I am right, it is important to decouple support to innovations from their growth implications, and throw our institutional support behind technologies that uphold human well-being over capital growth. Jason Hickel has some interesting thought in his book Less is More, and Mazzucato has forcefully made the point across the arc of her work.

Time will tell; and I am confident that better minds than mine will cast more light onto the matter. But this question can no longer wait, and if you are working in one of these two tech ecosystems, I suggest you ask your employer, and yourself, some hard questions.