Olagligt att förutsäga framtiden?

Intrade, en av världens största framtidsmarknader, har beslutat att de inte längre tänker erbjuda handel i USA. Anledningen är att en myndighet beslutat sig för att framtidsmarknaden på något sätt skulle innebära ett hot mot handelssystemet i stort, och att det inte går an att ha oreglerad handel i framtidsförutsägelser. Myndigheten, CFTC, har mer eller mindre startat ett krig mot framtidsmarknader, skriver Wall Street Journal i en mycket syrlig kommentar:

The CFTC’s war on prediction markets goes beyond Intrade. In 2010, the commission rejected an application by two exchanges to sell options on box-office receipts. Commissioner Bart Chilton said that such contracts would be unduly subject to manipulation by deep-pocketed Hollywood studios. But excluding studios from participating, he added, would mean these markets “would not result in the actual managing and assuming of price risks—something required for our approval.”

Yet in its 11 years on the Web, Intrade has proved neither frivolous nor manipulable. Intrade and similar services extend the predictive power of markets to offbeat but important topics from terrorism to flu outbreaks.

Tidningen konstaterar också att företag som Google, Siemens och andra arbetat med liknande lösningar, frågor och produkter för att bättre kunna fatta beslut. Ändå tycker CFTC att det är dags, tydligen, att sätta ned foten och förklara inTrade olagligt.

Det är mycket dumt. Inte bara för att dessa marknader är värdefulla redan i dag, utan för att de utgör det första steget i utvecklingen av teknik för att utnyttja the wisdom of crowds. Den här typen av kortsynt regleringsiver skapar trösklar och problem där inga egentligen behöver finnas.

The Economist går så långt i sitt senaste nummer att de ifrågasätter om framtidsmarknaderna alls kan vara framgångsrika om de inte kan nå den kritiska massa av handel som krävs. Om marknaderna utestängs från USA kan resultatet bli att vi går miste om intressanta, viktiga och värdefulla prognoser endast på grund av usel tillämpning av en föråldrad lagstiftning, eller rädsla för marknader i stort.

Det vore inte bara synd utan riktigt dumt. Ezra Klein kommenterar det hela intressant på Wonkblog. Han skriver, om nyttan med dessa marknader:

In a 2008 paper, a group of 19 economists, including Nobel laureate Kenneth Arrow, argued that prediction markets could potentially prove extremely useful. They are often more accurate than traditional forecasting tools — in part because they aggregate a wide range of informed opinions. ”The range of applications is virtually limitless,” the economists wrote, “from helping businesses make better investment decisions to helping governments make better fiscal and monetary policy decisions.”

Det vore ju närmast tragiskt om vi försatte chansen att använda teknikens förmåga att koordinera kunskap, för att vi sitter fast i en gammal bild av vilka marknader vi vill tillåta. Men det finns en märklig rädsla här, och det kanske tydligaste exemplet var när det föreslogs att framtidsmarknader kunde användas för att förutsäga terroristattacker. Den allmänna opinionen tvingade Richard Perle att avgå för att han ansågs ha föreslagit att man skulle slå vad om amerikanska liv. Långt senare beklagade forskare närstående CIA att projektet gått i graven:

The decision to cancel FutureMAP was at the very least premature, if not wrong-headed. The bulk of evidence on prediction markets demonstrate that they are reliable aggregators of disparate and dispersed information and can result in forecasts that are more accurate than those of experts. If so, prediction markets can substantially contribute to US Intelligence Community strategic and tactical intelligence work.

Resultatet av mediedrevet var alltså sannolikt sämre säkerhet för amerikanska medborgare. Knappast det man ville åstadkomma. Och rakt motsatt det CFTC hävdar när man skriver att framtidsmarknaderna “strider mot det offentliga intresset“.

CFTCs beslut är ett misstag, och riskerar att kraftigt försvåra utvecklingen av effektiva kunskapsverktyg för bättre samhällsbeslut i allt mer komplexa system. Och det när vi behöver sådan hjälp som mest. Det behövs kanske ett studium av framtids- eller prognosrätt, där vi kan diskutera frågor som om de borde vara olagligt att förutsäga framtiden, eller om seismologer skall dömas för dråp för felaktiga förutsägelser. Det är knappast ett område som kommer att bli mindre intressant i framtiden. Om man nu vågar sig på en förutsägelse…

Är artificiell intelligens en existentiell risk?

Igår publicerade BBC en nyhet som utmanade på så många olika plan att det är rakt omöjligt att inte skriva om den. I korthet rapporterade man att ett akademiskt center för studiet av existentiell risk skall studera riskerna förenade med att artificiell intelligens. Artikeln är illustrerad med en terminator-liknande robot, och typiskt nog handlar den mest om “the robot uprising”. Men här finns en rikedom av riktigt svåra frågor som förtjänar att undersökas på djupet, och som forskarna bakom projektet verkar vara mer intresserade av.

För det första är risken att robotarna utrotar oss ganska liten, och det riktigt stora problemet ett helt annat — nämligen att de struntar i oss. Vi är fruktansvärt egocentriska när vi skriver om artificiell intelligens, och den första frågan vi alltid ställer oss är hur den artificiella intelligensen kommer att förhålla sig till oss. Kommer den att utplåna oss eller kommer den att frälsa oss?

Det finns ett tredje, ganska nedslående svar och det är att en AI kanske inte kommer att vara särskilt intresserad av oss alls. Det skulle bjuda på ett härligt haveri – nätet blir medvetet, och struntar i oss, bygger mängder av von Neumann-maskiner med jordens resurser och bestämmer sig för att kolonisera universum. En AI skulle se annorlunda på den tid intergalaktiskt resande erbjuder, och skulle med lätthet kunna motivera den begränsade kostnad och risk det innebär att konstruera just ett von Neumann-baserat rymdprogram där sonder som kan konstruera kopior av sig själva och AI:n långsamt sprider sig som en intelligensexplosion i slow motion i universum. (Det finns en hel del intressanta frågor om varför det inte hänt ännu, eller hur det inte hänt ännu, någon annanstans i universum). Men är en sådan AI en risk? Ja, i och för sig, om den tar alla resurser då vore det ju trist, men den kanske har en grundläggande etik som går ut på att inte ta livet av andra livsformer (eller, hm, behandla dem som risker…). Då skulle vi hamna i den lustiga situationen att vi skapat intelligens och att den ignorerar oss, artigt men bestämt, medan den förbereder sig för att kolonisera universum. Litet som vi skulle behandla andra djurarter, som författarna skriver. Men författarnas exempel är insekter, men det finns inget skäl att anta att AI skulle behandla oss som insekter snarare än som, säg, stora apor eller hundar. (Minsky lär en gång ha sagt att vi kommer att bli våra datorers husdjur)

Det finns ett annat alternativ, som också skulle vara väldigt roligt, och det är ju att den inte anser att vi är intelligenta, ety hur kan intelligens någonsin vara baserad på något annat än massiva beräkningar i neurala nätverk i kisel? Nåväl, summa summarum är det kanske detta scenario som vi borde fundera mest kring, eftersom det är riktigt svårt att svara på frågan om varför en några storleksordningar högre intelligens skulle bry sig om oss alls. Forskarna citerar Vernor Vinge i en av de artiklar de skrivit om riskerna med AI. Vinge fick frågan om datorer någonsin kommer att bli lika intelligenta som människor. Hans svar var enkelt: “ja, men bara ett kort tag”. Därefter kommer de att vara oändligt mycket intelligentare än oss.

För det andra skriver en av professorerna om att intelligensen kommer att lösgöra sig från biologins bojor (ett underbart sätt att uttrycka det). Som vi diskuterat många gånger tidigare har detta förstås redan skett. Intelligens existerar inte i biologiska system, utan i kombinationer av system. Och det betyder kanske att vi aldrig kommer att se en isolerad AI utan en annan distribution av intelligens över existerande system, där de “artificiella systemen” (hur vi nu drar den gränsen) får en tydligare roll. Då blir riskerna helt annorlunda, då handlar det ju om huruvida dessa nya system är välbalanserade eller ej. Och hur snabbt de förändras. När vi bäddas in i intelligenta system blir vi en cybernetisk civilisation, beroende av tekniken, och oförmögna att leva utan den. Och det kan mycket väl redan ha skett. Frågan är om det kan finnas någon fördel i den tröghet som biologisk intelligens kännetecknas av?

För det tredje är det underbart att det finns ett center för existentiell risk, men vore det inte lika roligt om det fanns ett för existentiella möjligheter? Kanske är det detta Singularitetsuniversitetet försöker vara, men det skulle ändå vara intressant att se på möjligheterna i en utveckling, hur vi kan dra nytta av massiv och snabb förändring. Att Martin Rees var en av personerna bakom projektet gjorde mig ändå litet misstänksam. Han har ju tidigare slagit vad om att en miljon människor kommer att dö av “bio-error or bio terror” före året 2020 i en enda olycka.  Och när han skrev en bok om framtiden, tja, då var det om hur den tar slut.

Det är inte oviktigt att studera hur det hela skulle kunna ta slut. Men det är lika viktigt, om inte oändligt viktigare, att studera hur det hela tar sig en ny början.

 

Understanding prejudice and racism online through search data

In a recent paper by Seth I. Stephens-Davidowitz Google Search Data is used to assess how much racist-sentiment affected the 2008 vote. The method is interesting, and the outcome nothing short of sensational:

The results imply that, relative to the areas in the United States with the lowest racial animus, racial animus cost Obama between 3.1 percentage points and 5.0 percentage points of the national popular vote. This implies racial animus gave Obama’s opponent roughly the equivalent of a home-state advantage country-wide. The cost of racial animus was not decisive in the 2008 election. But a four percentage point loss by the winning candidate would have changed the popular vote winner in the majority of post-war presidential elections.

Now, read the paper for yourself and determine if you agree with the methodology or not, but it does look interesting. Ultimately it becomes a correlation/causation issue to some extent, but still – interesting.

The notion that we could track prejudice and racist sentiment like this seems to open for new ways to track the dark sides of society and perhaps bring our creativity to bear on the problem of how to counter the fear and ignorance that underpins the memes involved.

Examining persuasion: internet advertising as linguistic research tool

In “Ecological Evaluation of Persuasive Messages Using Google AdWords” the authors show how internet advertising systems can be used to test and explore the mechanism of persuasion in different ways, for example for exploring natural language processing.

In recent years there has been a growing interest in crowdsourcing methodologies to be used in experimental research for NLP tasks. In particular, evaluation of systems and theories about persuasion is difficult to accommodate within existing frameworks. In this paper we present a new cheap and fast methodology that allows fast experiment building and evaluation with fully-automated analysis at a low cost. The central idea is exploiting existing commercial tools for advertising on the web, such as Google AdWords, to measure message impact in an ecological setting. The paper includes a description of the approach, tips for how to use AdWords for scientific research, and results of pilot experiments on the impact of affective text variations which confirm the effectiveness of the approach.

A similar approach is described in the book Supercrunchers and has been used by political consultants to do message testing in a fast and relatively cheap way. As we proceed down the road I think secondary research uses can be found for a lot of the services online. The secondary uses of data allow for a wide variety of creative uses that explore language, social structures and economics. The data shadow of the web is the next frontier for many of these empirical sciences.

Besides exploring the theoretical notion the authors provide a great hands-on description of how to set up your experiment. Very intriguing.

(Full disclosure – I work at Google, but write here in a private capacity. The study was done in cooperation with Google and was funded by our research programme Google Research Awards to a part. I still think it is really, really interesting – but of course you should know those two things when reading this post!)

Thiel on the confusion of capitalism and competition, difficult and valuable

Peter Thiel is turning out to be a very, very original thinker. In today’s New York Times David Brooks mentions a lecture he gave at Stanford on starting up a company. Thiel’s main point is subtle and incredibly interesting: he says that people are confused when they think capitalism is about competition. In fact, with perfect competition there is no profit and you are forced to reinvest all the money you make to stay in place. Capitalism is about doing something in a large market where there is very little competition, and thus making huge profits. And then he adds that this is a consequence of many people confusing what is hard with what is valuable. There is so much sheer insight in these lecture notes that Blake Masters have put up that there is no excuse not to read them. Favorite quotes below:

The usual narrative is that capitalism and perfect competition are synonyms. No one is a monopoly. Firms compete and profits are competed away. But that’s a curious narrative. A better one frames capitalism and perfect competition as opposites; capitalism is about the accumulation of capital, whereas the world of perfect competition is one in which you can’t make any money.

On value and difficulty as a proxy for value

Intense competition makes things hard because you just beat heads with other people. The intensity of competition becomes a proxy for value. But value is a different question entirely. And to the extent it’s not there, you’re competing just for the sake of competition. Henry Kissinger’s anti-academic line aptly describes the conflation of difficulty and value: in academia at least, the battles are so fierce because the stakes are so small.

That seems true, but it also seems odd. If the stakes are so small, why don’t people stop fighting so hard and do something else instead? We can only speculate. Maybe those people just don’t know how to tell what’s valuable. Maybe all they can understand is the difficulty proxy. Maybe they’ve bought into the romanticization of competition. But it’s important to ask at what point it makes sense to get away from competition and shift your life trajectory towards monopoly.

On why life is not war

The perfect illustration of competition writ large is war. Everyone just kills everyone. There are always rationalizations for war. Often it’s been romanticized, though perhaps not so much anymore. But it makes sense: if life really is war, you should spend all your time either getting ready for it or doing it. That’s the Harvard mindset.

But what if life isn’t just war? Perhaps there’s more to it than that. Maybe you should sometimes run away. Maybe you should sheath the sword and figure out something else to do. Maybe “life is war” is just a strange lie we’re told, and competition isn’t actually as good as we assume it is.

On AI:

Artificial Intelligence is probably an underrated field. People are burned out on it, largely because it has been overrated and overstated for many decades. Few people think AI is or will soon be real at this point. But progress is increasingly relentless. AI performance in chess is increasing. Computers will probably beat humans in Go in 4 or 5 years. AI is probably a good place to look on the tech frontier. The challenge is that no one knows how far it will go.

The fascinating thing with Thiel’s argument is that it contains a tip on how to live your life as well on how to start your business. And maybe the trick is not viewing your life as very different from an investment in a startup. And realizing that life is not war, difficulty is not a good proxy for value and competition is opposed to capitalism.

FuturICT – is this the approach to research we want in the EU?

The European Union has a research policy agenda that varies wildly. In one project, FuturICT, it has set out to examine the following, according to their website:

The ultimate goal of the FuturICT flagship project is to understand and manage complex, global, socially interactive systems, with a focus on sustainability and resilience. Revealing the hidden laws and processes underlying societies probably constitutes the most pressing scientific grand challenge of our century and is equally important for the development of novel robust, trustworthy and adaptive information and communication technologies (ICT), based on socially inspired paradigms.

Oooookay. That is pretty ambitious. Now, here is a question for you. Is this the kind of research we want? I must confess to being very much of two minds here. On one side I do like the broad approach and much of what the project has been doing is interesting. A recent paper outlines in-depth some challenges for complexity sciences that I found interesting. Again, though, the scoping is a bit, hm, exorbitant:

FuturICT foundations are social science, complex systems science, and ICT. The main concerns and challenges in the science of complex systems in the context of FuturICT are laid out in this paper with special emphasis on the Complex Systems route to Social Sciences. This include complex systems having: many heterogeneous interacting parts; multiple scales; complicated transition laws; unexpected or unpredicted emergence; sensitive dependence on initial conditions; path-dependent dynamics; networked hierarchical connectivities; interaction of autonomous agents; self-organisation; non-equilibrium dynamics; combinatorial explosion; adaptivity to changing environments; co-evolving subsystems; ill-defined boundaries; and multilevel dynamics. In this context, science is seen as the process of abstracting the dynamics of systems from data. This presents many challenges including: data gathering by large-scale experiment, participatory sensing and social computation, managing huge distributed dynamic and heterogeneous databases; moving from data to dynamical models, going beyond correlations to cause-effect relationships, understanding the relationship between simple and comprehensive models with appropriate choices of variables, ensemble modeling and data assimilation, modeling systems of systems of systems with many levels between micro and macro; and formulating new approaches to prediction, forecasting, and risk, especially in systems that can reflect on and change their behaviour in response to predictions, and systems whose apparently predictable behaviour is disrupted by apparently unpredictable rare or extreme events. These challenges are part of the FuturICT agenda.

Oh, just that? Where is your ambition, project members? Joking aside, it is exhilarating to see someone aim for the stars like this. But will it succeed? One problem I have is that I do not know what it would look like for the project to succeed. Accomplishing the singularity (finally!) or producing a god-like AI? Or just cataloguing a series of really interesting problems?

So I hesitate. On one side: good for EU that it dares address these challenges head on! On the other side: what exactly are you doing? Then I remember the millions that the EU plowed down into Electronic Copyright Management Systems like Imprimatur. Maybe we are better off with a project that states the following:

The FuturICT flagship proposal intends to unify hundreds of the best scientists in Europe in a 10 year 1 billion EUR program to explore social life on earth and everything it relates to. The FuturICT flagship proposal will produce historic breakthroughs and provide powerful new ways to manage challenges that make the modern world so difficult to predict, including the financial crisis.

Oh, good. What is all the fuss on the stock markets about, then? So, what do you think. Flip or flop? My jury was caught in a combinatorial participatory sensing explosion.

Tomorrow’s conference – Stockholm Internet Forum

Tomorrow I will participate in Stockholm Internet Forum. I am looking forward to the panels and discussions, but also to meeting some old friends. In my own thinking about the conference I have ended up sketching out three themes that I want to examine.

The first is the notion that we cannot divide up responsibility in different sectors of a society. Saying that government has the responsibility to do x, and corporations have the responsibility to do y simply muddles the question. We have an overall responsibility that is dependent on everyone doing their part, and doing it well. That needs to be the basic insight here.

The second is that transparency is paramount. That we need transparency to ensure that the powers that we grant our governments over ourselves are exercised responsibly and with great care. Power observed and accounted for is power measured and careful exercised. More about this tomorrow.

The third is that I think that we have been under-theorizing the role of free flow of information generally. We need to understand information flows as fundamental to economic growth, intellectual achievement and cultural production. We need to build models that go beyond industrial society measures, and really understand the “space of flows” to craft a good internet society.

The good thing is that I will know more about all these things by the end of day tomorrow. The speakers and panels are top-notch and I am looking forward to learning more about this, as I happen to believe, important subject.

Update: Swedish government put out a press-release in which they were kind enough to count me among the guests the named by name. And I can die now, since they got my title right:

Prominent participants include Frank La Rue, UN Special Rapporteur on freedom of expression; Alec J Ross, Senior Advisor for Innovation to Secretary of State Hillary Clinton; Rebecca MacKinnon, one of the founders of Global Voices Online; Nicklas Lundblad, policy geek at Google; Suneet Singh, CEO of DataWind, which has launched the budget tablet Aakash; and Måns Adler, founder of the company and Internet service Bambuser for live broadcast of mobile phone videos.

 

Happiness, Physics and Death

Will knowledge about what our world is like help us craft a philosophy to deal better with our lives? It is not a trivial question. Knowing what the world is like could be both a blessing and a curse. There are those who think that the universe, such as it is, must have meaning for it to be possible for us to sustain meaning — this is in part the sense in which everything is allowed if God is dead in Dostoyevsky — but a perhaps more interesting position is to build meaning of sorts from established meaninglessness. Lawrence M Krauss recently wrote an article for the LA Times arguing exactly this, exploring what modern physics knows about the universe. His final words are interesting:

Imagining living in a universe without purpose may prepare us to better face reality head on. I cannot see that this is such a bad thing. Living in a strange and remarkable universe that is the way it is, independent of our desires and hopes, is far more satisfying for me than living in a fairy-tale universe invented to justify our existence.

There is truth to this. This is an almost Spinozan contemplation of the universe from the point of eternity, and it is in a sense liberating. If the universe is unlikely, without purpose, indeed, absurd, our choices seem less heavy. If we are the stuff of stars eating popcorn and having beer, if we are the unique and unlikely consequences of colliding particles and the children of collapsed wave forms, well, then maybe it is ok to be a bit late for work occasionally, or to just mess up.

Several studies have shown that people who contemplate death for a few minutes each day are happier. I think the same holds for those that look into the night sky and see the stars for what they are: our relatives, improbable and fantastic and uncaring. In fact, I would bet that if we could compare three groups: those that think about death, those that stare into the dark night sky and those that do neither, the second would be the happiest. That is an experiment I would like to try sometime.

Camus was right. We have to imagine that Sisyphus is happy – and so we could be too in a Sisyphean universe that cares little about our hopes and wants. In fact, there is something distinctly liberating about that isn’t there?

Mathematicians up in arms and they invite you to join!

I just finished an excellent little paper by Henry Cohn and Douglas N Arnold. It is called “Mathematicians take a stand” and encourages all mathematicians to join a boycott against publisher Elsevier. The boycott — over at Cost of Knowledge — has its root in the growing prices of academic journals and how they slow down the dissemination of research and hamper the human project of building our knowledge together. Today more than 8000 researchers – not only mathematicians –  has agreed to boycott the publisher if they do not allow for other easier ways for researchers to distribute their work and help build the future of knowledge.

The authors note in their paper:

While the mathematical literature itself is a treasure, the current system of scholarly publishing is badly broken. Elsevier is the largest and, in our view, the most egregious example of what is wrong. We hope many readers will agree with us that by choosing to withdraw our cooperation from Elsevier, we are sending a valuable message to them and to the scholarly publishing industry more broadly. Please consider joining the movement at http://thecostofknowledge.com.

What is our vision for the future? The mathematical community needs a period of experimentation and healthy competition, in which a variety of publishing models can flourish and develop. Possibilities include various approaches to open access publishing, refereed journals tightly integrated with the arXiv or similar servers, increased reliance on non-pro t publishers, hybrid models in which community-owned journals subcontract their operations to commercial publishers, commercially-owned journals with reasonable prices and policies, etc. It is too early to predict the mix of models that will emerge as the most successful. However, any publisher that wants to be part of this mix must convince the community that they oversee peer review with integrity, that they aid dissemination rather than hinder it, and that they work to make high-quality mathematical literature widely available at a reasonable price.

Let’s work together to foster good practices and build better models. The future of mathematics publishing is in our hands.

And with the future of publishing the future of the subject also hangs in the balance. The way knowledge is disseminated, and, that it at a very minimum is not actively hindered to reach people everywhere is essential to the future of the Information Society.

I just signed up. Will you?