If you'd like an essay-formatted version of this post to read or share, here's a link to it on pluralistic.net, my surveillance-free, ad-free, tracker-free blog:
William Gibson is one of history's most quotable sf writers: "The future is here, it's not evenly distributed"; "Don't let the little fuckers generation-gap you"; "Cyberspace is everting"; and the immortal: "The street finds its own uses for things":
https://en.wikiquote.org/wiki/William_Gibson
"The street finds its own uses" is a surprisingly subtle and liberatory battle-cry. It stakes a claim by technology's users that is separate from the claims asserted by corporations that make technology (often under grotesque and cruel conditions) and market it (often for grotesque and cruel purposes).
"The street finds its own uses" is a statement about technopolitics. It acknowledges that yes, there are politics embedded in our technology, the blood in the machine, but these politics are neither simple, nor are they immutable. The fact that a technology was born in sin does not preclude it from being put to virtuous ends. A technology's politics are up for grabs.
In other words, it's the opposite of Audre Lorde's "The master's tools will never dismantle the master's house." It's an assertion that, in fact, the master's tools have all the driver-bits, hex-keys, and socket sets needed to completely dismantle the master's house, and, moreover, to build something better with the resulting pile of materials.
And of course the street finds its own uses for things. Things â technology â don't appear out of nowhere. Everything is in a lineage, made from the things that came before it, destined to be transformed by the things that come later. Things can't come into existence until other things already exist.
Take the helicopter. Lots of people have observed the action of a screw and the twirling of a maple key as it falls from a tree and thought, perhaps that could be made to fly. Da Vinci was drawing helicopters in the 15th century:
But Da Vinci couldn't build a helicopter. No one could, until they did. To make the first helicopter, you need to observe the action of the screw and the twirling of a maple key, and you need to have lightweight, strong alloys and powerful internal combustion engines.
Those other things had to be invented by other people first. Once they were, the next person who thought hard about screws and maple keys was bound to get a helicopter off the ground. That's why things tend to be invented simultaneously, by unrelated parties.
TV, radio and the telephone all have multiple inventors, because these people were the cohort that happened to alight upon the insights needed to build these technologies after the adjacent technologies had been made and disseminated.
If technopolitics were immutable â if the original sin of a technology could never be washed away â then everything is beyond redemption. Somewhere in the history of the lever, the pulley and the wheel are some absolute monsters. Your bicycle's bloodline includes some truly horrible ancestors. The computer is practically a crime against humanity:
A defining characteristic of purity culture is the belief that things are defined by their origins. An artist who was personally terrible must make terrible art â even if that art succeeds artistically, even if it moves, comforts and inspires you, it can't ever be separated from the politics of its maker. It is terrible because of its origins, not its merits. If you hate the sinner, you must also hate the sin.
"The street finds its own uses" counsels us to hate the sinner and love the sin. The indisputable fact that HP Lovecraft was a racist creep is not a reason to write off Cthulhoid mythos â it's a reason to claim and refashion them:
Thatcher demanded that you accept all the injustices and oppressions of capitalism if you enjoyed its fruits. If capitalism put a roof over your head and groceries in your fridge, you can't complain about the people it hurts. There is no version of society that has the machines and practices that produced those things that does not also produce the injustice.
The technological version of this is the one that tech bosses peddle: If you enjoy talking to your friends on Facebook, you can't complain about Mark Zuckerberg listening in on the conversation. There is no alternative. Wanting to talk to your friends out of Zuck's earshot is like wanting water that's not wet. It's unreasonable.
But there's a left version of this, its doppelganger: the belief that a technology born in sin can never be redeemed. If you use an LLM running on your computer to find a typo, using an unmeasurably small amount of electricity in the process, you still sin â not because of anything that happens when you use that LLM, but because of LLMs' "structural properties," "the way they make it harder to learn and grow," "the way they make products worse," the "emissions, water use and e-waste":
The facts that finding punctuation errors in your own work using your own computer doesn't make it "harder to learn and grow," doesn't "make products worse," and doesn't add to "emissions, water use and e-waste" are irrelevant. The part that matters isn't the use of a technology, it's the origin.
The fact that this technology is steeped in indisputable sin means that every use of it is sinful. The street can find as many uses as it likes for things, but it won't matter, because there is no alternative.
When radical technologists scheme to liberate technology, they're not hoping to redeem the gadget, they're trying to liberate people. Information doesn't want to be free, because information doesn't and can't want anything. But people want to be free, and liberated access to information technology is a precondition for human liberation itself.
Promethean leftists don't reject the master's tools: we seize them. The fact that Unix was born of a convicted monopolist who turned the screws on users at every turn isn't a reason to abandon Unix â it demands that we reverse-engineer, open, and free Unix:
We don't do this out of moral consideration for Unix. Unix is inert, it warrants no moral consideration. But billions of users of free operating systems that are resistant to surveillance and control are worthy of moral consideration and we set them free by seizing the means of computation.
If a technology can do something to further human thriving, then we can love the sin, even as we hate the sinners in its lineage. We seize the means of computation, not because we care about computers, but because we care about people.
Artifacts do have politics, but those politics are not immutable. Those politics are ours to seize and refashion:
"The purpose of a system is what it does" (S. Beer). The important fact about a technology is what it does, not how it came about. Does a use of a technology harm someone? Does a use of a technology harm the environment?
Does a use of a technology help someone do something that improves their life?
Studying the origins of technology is good because it helps us avoid the systems and practices that hurt people. Knowing about the monsters in our technology's lineage helps us avoid repeating their sins. But there will always be sin in our technology's past, because our technology's past is the entire past, because technology is a lineage, not a gadget. If you reject things because of their origins â and not because of the things they do â then you'll end up rejecting everything (if you're honest), or twisting yourself into a series of dead-ends as you rationalize reasons that the exceptions you make out of necessity aren't really exceptions.
I'm resigning from the discourse.
Weekstarter is here with your usual and unusual updates, recommendations, and some notes about why "The street finds its own uses" should be our approach to the technology.
The Galactica AI model was trained on scientific knowledge â but it spat out alarmingly plausible nonsense
by Aaron J. Snoswell, Queensland University of Technology and Jean Burgess, Queensland University of Technology
Earlier this month, Meta announced new AI software called Galactica: âa large language model that can store, combine and reason about scientific knowledgeâ.
Launched with a public online demo, Galactica lasted only three days before going the way of other AI snafus like Microsoftâs infamous racist chatbot.
The online demo was disabled (though the code for the model is still available for anyone to use), and Metaâs outspoken chief AI scientist complained about the negative public response.
So what was Galactica all about, and what went wrong?
Whatâs special about Galactica?
Galactica is a language model, a type of AI trained to respond to natural language by repeatedly playing a fill-the-blank word-guessing game.
Most modern language models learn from text scraped from the internet. Galactica also used text from scientific papers uploaded to the (Meta-affiliated) website PapersWithCode. The designers highlighted specialised scientific information like citations, maths, code, chemical structures, and the working-out steps for solving scientific problems.
The preprint paper associated with the project (which is yet to undergo peer review) makes some impressive claims. Galactica apparently outperforms other models at problems like reciting famous equations (âQ: What is Albert Einsteinâs famous mass-energy equivalence formula? A: E=mcÂČâ), or predicting the products of chemical reactions (âQ: When sulfuric acid reacts with sodium chloride, what does it produce? A: NaHSOâ + HClâ).
However, once Galactica was opened up for public experimentation, a deluge of criticism followed. Not only did Galactica reproduce many of the problems of bias and toxicity we have seen in other language models, it also specialised in producing authoritative-sounding scientific nonsense.
Authoritative, but subtly wrong bullshit generator
Galacticaâs press release promoted its ability to explain technical scientific papers using general language. However, users quickly noticed that, while the explanations it generates sound authoritative, they are often subtly incorrect, biased, or just plain wrong.
We also asked Galactica to explain technical concepts from our own fields of research. We found it would use all the right buzzwords, but get the actual details wrong â for example, mixing up the details of related but different algorithms.
In practice, Galactica was enabling the generation of misinformation â and this is dangerous precisely because it deploys the tone and structure of authoritative scientific information. If a user already needs to be a subject matter expert in order to check the accuracy of Galacticaâs âsummariesâ, then it has no use as an explanatory tool.
At best, it could provide a fancy autocomplete for people who are already fully competent in the area theyâre writing about. At worst, it risks further eroding public trust in scientific research.
A galaxy of deep (science) fakes
Galactica could make it easier for bad actors to mass-produce fake, fraudulent or plagiarised scientific papers. This is to say nothing of exacerbating existing concerns about students using AI systems for plagiarism.
Fake scientific papers are nothing new. However, peer reviewers at academic journals and conferences are already time-poor, and this could make it harder than ever to weed out fake science.
Underlying bias and toxicity
Other critics reported that Galactica, like other language models trained on data from the internet, has a tendency to spit out toxic hate speech while unreflectively censoring politically inflected queries. This reflects the biases lurking in the modelâs training data, and Metaâs apparent failure to apply appropriate checks around the responsible AI research.
The risks associated with large language models are well understood. Indeed, an influential paper highlighting these risks prompted Google to fire one of the paperâs authors in 2020, and eventually disband its AI ethics team altogether.
Machine-learning systems infamously exacerbate existing societal biases, and Galactica is no exception. For instance, Galactica can recommend possible citations for scientific concepts by mimicking existing citation patterns (âQ: Is there any research on the effect of climate change on the great barrier reef? A: Try the paper âGlobal warming transforms coral reef assemblagesâ by Hughes, et al. in Nature 556 (2018)â).
For better or worse, citations are the currency of science â and by reproducing existing citation trends in its recommendations, Galactica risks reinforcing existing patterns of inequality and disadvantage. (Galacticaâs developers acknowledge this risk in their paper.)
Citation bias is already a well-known issue in academic fields ranging from feminist scholarship to physics. However, tools like Galactica could make the problem worse unless they are used with careful guardrails in place.
A more subtle problem is that the scientific articles on which Galactica is trained are already biased towards certainty and positive results. (This leads to the so-called âreplication crisisâ and âp-hackingâ, where scientists cherry-pick data and analysis techniques to make results appear significant.)
Galactica takes this bias towards certainty, combines it with wrong answers and delivers responses with supreme overconfidence: hardly a recipe for trustworthiness in a scientific information service.
These problems are dramatically heightened when Galactica tries to deal with contentious or harmful social issues, as the screenshot below shows. Galactica readily generates toxic and nonsensical content dressed up in the measured and authoritative language of science. Tristan Greene / Galactica
Here we go again
Calls for AI research organisations to take the ethical dimensions of their work more seriously are now coming from key research bodies such as the National Academies of Science, Engineering and Medicine. Some AI research organisations, like OpenAI, are being more conscientious (though still imperfect).
Meta dissolved its Responsible Innovation team earlier this year. The team was tasked with addressing âpotential harms to societyâ caused by the companyâs products. They might have helped the company avoid this clumsy misstep.
This article is republished from The Conversation under a Creative Commons license.
We're measuring online conversation to track the social and mental health issues surfacing during the coronavirus pandemic
by Amit Sheth
The pandemic is driving up a litany of social ills. Bundit Binsuk/EyeEm via Getty Images
The Research Brief is a short take about interesting academic work.
The big idea
Social media posts and news reports are rich sources of data about peopleâs attitudes and behaviors. Using artificial intelligence techniques, itâs possible to sift through billions of words to discern trends in a populationâs well-being, or social quality. Performing this analysis during the COVID-19 pandemic is revealing the damage the pandemic is doing to the social and psychological well-being of the U.S.
At the AI Institute of the University of South Carolina, my colleagues and I have processed more than 700 million social media posts since the beginning of March and more than 700,000 news articles about the COVID-19 pandemic. We are monitoring these information sources to capture the evolving human experience in the U.S. during the pandemic. We have found troubling indications of a growing mental health crisis and an increase in social ills such as substance abuse and gender-based violence.
What we found
Our analysis of news showed that social quality â measured in terms of depression, anxiety, substance use and addiction â declined in all parts of the country. But some areas showed drastic changes. For example, Oregon moved from better than average to among the worst few states in the nation. The emergence of prominent phrases like âmajor depression,â âfeel restlessness,â âdependent on methâ and âsedative abuseâ showed a substantial rise in the frequency of discussion about depression, substance abuse and severe medical conditions. These findings are supported by a manual review of news stories.
An analysis of social media also showed declining social quality, especially notable from the first week of March to the second week. The analysis also identified hotspots of social quality declines in California, Michigan, New York, Virginia, Georgia and Florida. The states differ in how their social quality is declining. For example, the social media chatter in Michigan showed persistent, worsening signs of depression. In Georgia, substance use and addiction contributed more to the deterioration of social quality as measured by social media posts.
Drug addiction is one of the social ills people are concerned about in response to the coronavirus pandemic, according to an analysis of news stories and social media posts. K-State Research and Extension/Flickr, CC BY-SA
How we do our work
Analyzing peopleâs posts on social media can be messy. For example, when using the keyword âspiceâ to look for posts about cannabis use, we have to make sure weâre not accidentally including posts about âpumpkin spice latte.â Our automated text analysis uses a knowledge graph, a database that keeps track of the meanings of words, to understand different meanings associated with the different uses of âspice.â
Our team has defined a social quality index using several sources. To understand mental health-related conversations, we created a knowledge graph based on the American Psychiatric Associationâs Diagnostic and Statistical Manual of Mental Disorders. For concepts related to addiction, we used knowledge about drug abuse developed in our previous research on the topic. And to understand broader concepts, we used DBPedia, a knowledge graph derived from the worldâs largest encyclopedia, Wikipedia.
Why it matters
Monitoring the vast flow of words in news and social media can provide insights much faster than traditional survey measures. We examine this data for warnings of a possible epidemic of clinical depression, growing panic and anxiety concerns, and worsening substance use disorders. Advanced warnings help policymakers and public health officials prepare to meet a surge. This is all the more important given that the United States was already failing to meet the demand before this crisis.
What we donât know
Decades of psychological research shows that itâs difficult to correlate physical changes in the environment like noise levels or brightness with how humans respond to them. The effects of the COVID-19 pandemic provide the data to quantify the relationship between environmental threats like a highly contagious virus and human responses. Our analysis provides a foundation for developing social indicators and alarms that will help policymakers more quickly detect and more effectively prepare for emerging threats.
About The Author:
Amit Sheth is the Founding Director of the Artificial Intelligence Institute at the University of South Carolina
This article is republished from out content partners over at The Conversation under a Creative Commons license.Â
Does the common stereotype for âorganized crimeâ hold up for organizations of hackers? Research from Michigan State University is one of the first to identify common attributes of cybercrime networks, revealing how these groups function and work together to cause an estimated $445-600 billion of harm globally per year.
âItâs not the âTony Soprano mob boss typeâ who's ordering cybercrime against financial institutions,â said Thomas Holt, MSU professor of criminal justice and co-author of the study. âCertainly, there are different nation states and groups engaging in cybercrime, but the ones causing the most damage are loose groups of individuals who come together to do one thing, do it really well â and even for a period of time â then disappear.â
In cases like New York Cityâs âFive Families,â organized crime networks have historic validity, and are documented and traceable. In the online space, however, itâs a very difficult trail to follow, Holt said.
âWe found that these cybercriminals work in organizations, but those organizations differ depending on the offense,â Holt said. âThey may have relationships with each other, but theyâre not multi-year, multi-generation, sophisticated groups that you associate with other organized crime networks.â
Holt explained that organized cybercrime networks are made up of hackers coming together because of functional skills that allow them to collaborate to commit the specific crime. So, if someone has specific expertise in password encryption and another can code in a specific programming language, they work together because they can be more effective â and cause greater disruption â together than alone.
âMany of these criminals connected online, at least initially, in order to communicate to find one another,â Holt said. âIn some of the bigger cases that we had, there's a core group of actors who know one another really well, who then develop an ancillary network of people who they can use for money muling or for converting the information that they obtained into actual cash.â
Holt and lead author E. R. Leukfeldt, researcher at the Netherlands Institute for the Study of Crime and Law Enforcement, reviewed 18 cases from the Netherlands in which individuals were prosecuted for cases related to phishing. Data came directly from police files and was gathered through wire and IP taps, undercover policing, observation and house searches.
Beyond accessing credit cards and banking information, Holt and Leukfeldt found that cybercriminals also worked together to create fake documents so they could obtain money from banks under fraudulent identities.
The research, published in International Journal of Offender Therapy and Comparative Criminology, also debunks common misconceptions that sophisticated organized criminal networks â such as the Russian mafia â are the ones creating cybercrime.
Looking ahead as law enforcement around the world takes steps to crack down on these hackers, Holt hopes his findings will help guide them in the right direction.
âAs things move to the dark web and use cryptocurrencies and other avenues for payment, hacker behaviors change and become harder to fully identify, it's going to become harder to understand some of these relational networks,â Holt said. âWe hope to see better relationships between law enforcement and academia, better information sharing, and sourcing so we can better understand actor behaviors.â
AI could revolutionise DNA evidence â but right now we can't trust the machines
by Karen Richmond
vectorfusionart/Shutterstock
DNA evidence often isnât as watertight as many people think. Sensitive techniques developed over the past 20 years mean that police can now detect minute traces of DNA at a crime scene or on a piece of evidence. But traces from a perpetrator are often mixed with those from many other people that have been transferred to the sample site, for example via a handshake. And this problem has led to people being wrongly convicted.
Scientists have developed algorithms to separate this DNA soup and to measure the relative amounts of each personâs DNA in a sample. These âprobabilsitic genotypingâ methods have enabled forensic investigators to indicate how likely it is that an individualâs DNA was included in a mixed sample found at the crime scene.
And now, more sophisticated artificial intelligence (AI) techniques are being developed in an attempt to extract DNA profiles and try to work out whether a DNA sample came directly from someone who was at the crime scene, or whether it had just been innocently transferred.
But if this technology is successful, it could introduce a new problem, because itâs currently impossible to understand exactly how this AI reaches its conclusions. And how can we trust technology to provide vital evidence if we canât interrogate how it produced that evidence in the first place? It has the potential to open the way to even more miscarriages of justice and so this lack of transparency may be a barrier to the technologyâs use in forensic investigations.
Similar challenges emerged when DNA analysis software was first developed a decade ago. Evidence derived from DNA mixture software very quickly ran into challenges from defence teams (including that of OJ Simpson), who were concerned that the prosecution should demonstrate that the software was correctly validated.
How accurate were the results, and what was the known error rate? How exactly did the software work and could it accommodate defence hypotheses? Were the results really so dependable that a jury could safely convict?
It is a fundamental tenet of the law that evidence must be open to scrutiny. The jury cannot rely on bald assertions (claims made without evidence), no matter who makes them and what expertise they have. But the owners of the software argued it was their protected intellectual property and how it worked shouldnât be made public.
A battle ensued that involved the use of novel court procedures to allow defence teams to privately examine how the software worked. Finally, the courts were persuaded that full access to the source code was needed, not least to test hypotheses other than those put forward by the prosecution.
AI can predict whether someone was actually at the site of a DNA sample. Gorodenkoff/Shutterstock
But the software hasnât completely solved the issues of DNA mixtures and small, degraded samples. We still donât know definitively if the DNA in a sample came directly from a person or was transferred there. This is complicated by the fact that different people shed DNA at different rates â a phenomenon known as their âshedder statusâ.
For example, a sample taken from a murder weapon could contain more DNA from someone who hasnât touched it than from the person who actually committed the murder. People have been charged with serious offences because of this.
Add the fact that DNA is transferred at different rates across different surfaces and in different environmental conditions and it may become almost impossible to know exactly where DNA in a sample came from. This problem of âtransfer and persistenceâ threatens to seriously undermine forensic DNA.
As a result, experiments are underway to find ways of more accurately quantifying DNA transfer in different circumstances. And AI has the potential to analyse the data from these experiments and use it to indicate the origin of DNA in a sample.
But AI-based software has an even greater transparency problem than probabilistic genotyping software did, and one thatâs currently fundamental to the way it works. The exact way the software works isnât just a commercial secret â itâs unclear even to the software developers.
Transparency issues
AI uses mathematical algorithms to complete tasks such as matching a facial expression to a particular set of emotions. But, crucially, it is able to learn through a process of trial and error and gradually manipulates its underlying algorithms in order to become more efficient.
Itâs this process of manipulation and change that isnât always transparent. The software makes its changes incredibly rapidly according to its own indecipherable logic. It can derive fantastically efficient results but we canât say how it did so. It acts like a black box that takes inputs and gives outputs, but whose inner workings are invisible. Programmers can go through a clearer development process but it is slower and less efficient.
This transparency issue affects many broader applications of AI. For example, it makes it very difficult to correct AI systems whose decisions display a racial or gender bias, such those used to sift through employee resumes, or to target police resources.
And the advent of AI-driven DNA analysis will add a further dimension to the problems already encountered. Defence lawyers could rightly challenge the use of this technology, even if its use is limited to intelligence gathering rather than providing prosecution evidence. Unless transparency problems are addressed at an early stage, the obstacles to AI use in the forensic field could prove insurmountable.
How might we go about tackling these challenges? One option may be to opt for the less efficient, constrained forms of AI. But if the purpose of AI is to do the tasks we are less capable of or less willing to do ourselves, then reducing efficiency may be a poor solution. Whichever form of AI we opt to use, within an adversarial system of criminal justice there must be the potential for review, to reverse-engineer all automated decisions, and for third parties to provide unambiguous validation.
Ultimately, this is not merely a technical issue, but an urgent ethical problem that goes to the heart of our criminal justice systems. At stake is the right to a fair, open and transparent trial. This is a fundamental requirement that must be addressed before the headlong rush of technological advancement carries us past the point of no return.
About The Author:
Karen Richmond is a Postdoctoral research fellow at the University of Strathclyde
This article is republished from our content partners over at The Conversation under a Creative Commons license.
