David Weinberger: what information was

There are many things I admire about my friend and colleague, David Weinberger: his intellectual curiosity, his generosity with his time and guidance, his sense of humor… One facet of David I most admire is his willingness to think in public. Most people who speak for a living (as David does, and as I aspire to do) use well-worn and carefully roadtested material. David is brave enough to put new ideas in front of audiences and work through new ideas, live and in public. And we’re lucky enough at the Berkman Center lunch today to hear his new talk, “What Information Was: Bits, Links and the Iron Rule of Irony”, an exploration of issues that David is starting to think about and wrestle with.

David starts with the provocative question, “How did we become the information age?” We’re moving out of that age and into a new one, one we haven’t named and don’t even understand yet. So we’re at a good point to reflect on this closing age and ask, “Why did information become the central metaphor?”

Information has been a cradle to grave metaphor for this age. David tells us that, if he were to stand up and say, “DNA is not information” – we’d probably think I was anti-science and an idiot. But DNA doesn’t come labeled with base-pairs. When we look at DNA, we see information… but DNA is a squiggly little molecule, a physical thing. It’s helpful to consider it as information and analyze it that way… but it’s a molecule.

Ray Kurzweil represents the “grave” side of the metaphor. Kurzweil’s “Age of Spiritual Machine” asks “when will we have computers large enough to model the brain, and allow Kurzweil to pour himself into a computer and survive his own death… .” Our willingness to consider this idea – the idea that a computer running a model of Kurzweil’s brain, the idea that this even makes sense to us, shows us that we think of ourselves as information.

In epistomology, we’ve traditionally considered sensation, perception, and judgement. In the past century, we’ve added “sense data” to this paradigm, and in the last fifty years, added information. We tend to consider information to be a basic constituent of the human mind. Stephen Wolfram argues that the universe itself is made from information – he suggests that the univese can be thought of in terms of cellular automota, reductions to pure information.

Despite the fact that we’ve reconsidered huge aspects of our culture in terms of information, we’re extremely bad at answering the question, “What is information?” Weinberger cites Ronald Day, who mentions that he’s discovered roughly 200 definitions of information. There’s a technical definition for the term, but that’s almost never what we mean.

There’s a conventional history of information, which moves from the Jaquard loom through Babbage, through Hollerith, through Turing and Shannon. But that history is a reading back into the 19th century of a concept that we didn’t have, a Shannonian definition of information. Instead, David proposes we look at Babbage and see how he considered information, long before Shannon’s information theory. Babbage uses the word “information” 28 times in his autobiography. Initially, he’s talking about information as “something you didn’t know, and now you do.” The second use is as “the contents of a table” – a set of data that was useful to the railroad industry.

In 1948, Claude Shannon took over the term and introduced a new definition for it. He did this for “noise” as well, and redefined them with highly technical meanings which allowed us to discuss the carrying capacity of different channels. In this definition, information is “a sequence of choices from a finite set of elementary symbols” which can be transmitted.

The history of information starts with simple definitions, takes a detour into complex mathematics, and then becomes a metaphor for… well… everything. So what enabled information to take over the world? Its utility, and its politics. But David’s interested in the implications of the metaphor:

Information scales. Information allows corporations to grow to new sizes. But the secret of the information age is that information works by reducing the amount of information – you simplify individuals to the simple categories you decide are important. Information helped companies only because we made the decision to strip things down.

In that paradigm, we might reduce an employee as a few database fields – name, title, social security number, salary. Now, we might represent an employee as a Facebook page – a vast set of connections, an abundance of information. Links aren’t about stripping down – they’re about expanding the universe. In an age of abundance – an abundance of good stuff and of crap – we’re actually better at managing the crap than the good stuff. We still use email because we’ve figured out how to avoid the crap.

We’re less aware of the good stuff – we believe that good ideas are fairly scarce, and this turns out not to be true… and this may be what’s killing newspapers. Institutions that depend on scarcity start to fail in an age of abundance. These institutions used to separate signal from noise. But that’s the wrong way to think about things today – we’ve got an abundance of signal, not of noise.

Information is a resource. We can query information and we can fetch from it. Alternatively, we can navigate it, entering a space. We thought about this a lot in the information age – think of a movie like Tron where we enter into data, becoming an avatar. We also thought about ourselves as being engulfed and threatened by information, like Katherine Hepburn in “The Desk Set“. These days, we routinely think of ourselves as entering the web space. The web and the real world are now so integrated, we can’t bring our children out of this space – they’re on their phones, in virtual space at he same time as the real world.

We used to measure our ability to query and fetch information in terms of precision and recall – did we get what we were searching for? Did we get all of it? In a world where we query trillions of pages, we’ve begun to think about relevance and interestingness. Precision and recall are generally less important, especially in the gigantic world of the web. Relevance has to do with whether information meets your needs; interestingness is obviously an idiosyncratic characteristic. We’re no longer using objective metrics – we’re using deeply personal ones.

Bits apply to everything. Nothing escapes being “bitified”. And we have a sense for what bits are – David quotes Gregory Bateson: “bits are a difference that make a difference”. A bit measures some sort of difference. While we talk about “atoms versus bits”, we don’t talk about “atoms versus links”. Bits are useful because they can apply to everything – “plain or peanut M&Ms, Kurzweil is or is not in the machine”. With these distinctions, we can build models that are coextensive with the world. They apply to everything, or almost everything.

We can only do this because bits are unlike atoms. We believe that atoms actually exist. Bits don’t have that same, objective reality. The holes in a punchcard are bits – the holes in your shoes are not. Bits have to be in a system, a system that is highly regularized or standardized. If you punched holes in a card at random, they wouldn’t be bits – they’d be noise.

There are a hundred billion neurons in Ray Kurzweil’s brain. We could model his brain in a swirling cloud of dust, somewhere out in the universe – a piece swirling to the left could be a 1, to the right, a 0. Somewhere, if we consider the left and right spin of clouds of dust, we have a model of Ray Kurzweil’s brain as it was in the ten minutes when he first met his wife and fell in love. But this is absurd – reverse the spin, and we’ve got an entirely different representation. Bits are not real.

When we turn the continuous, real-world into bits, we need to decide what our resolution is. That resolution is different between a satellite map and mapping each rock on a coastline. Bits depend on our resolution, what we’re intending to do. This is true in computers as well – we decide to measure certain voltages as positive or negative, ones or zeros, based on a defining line.

Bits are about reducing distinctions to the simplest possible states – black or white, yes or no. They simplify. The web, by contrast, is a web of links. They agree, they amplify, they endorse, they denounce, they connect. Those links aren’t as simple as on and off – they build an enormously complex and intricate world, an abundance of rich, linguistic human intentions.

Information explains communication. At the very beginning of his critical essay on information, Shannon makes it clear that he’s not talking about semantics. But he introduces a very clear, formal model: Information is created by a sender, encoded, put through a channel, decoded and delivered to a recipient. Warren Weaver, who is credited as developing this model with him, acknowledges at the beginning of his seminal book that communication has a much broader sense – not just written and oral communications, but music, theatre, ballet and all human behavior. So why did the Shannon-Weaver Mathematical Model – an extremely narrow, formal model – turn into what we think communication is?

We adopted a conduit metaphor for communication before information theory, a vision of communications that looks like tin cans with a string between them. Communication isn’t a signal that’s a vibration in the air – it’s an act within a world where people share context and concerns. So why is the tin can model so powerful for us?

Descartes solved the mind-body problem, culminating a long tradition in western philosophy. He explains that we live in mental images, not just in the real world. It’s a lonely view of the world: each of us live by ourselves, in our own mental images of the world. In that space, communication has to be the act of communicating a worldview into another person’s heads. This is, “strictly speaking, a pathological, schizophrenic metaphysics.”

A less schizophrenic vision is that two individuals share a world, share perceptions and concerns. There is something interesting and relevant to them, and by talking, one may now see the world differently than before. Yes, there are vibrations in the air. But the tin can model of communications strips out everything that’s important – it’s a bad, incomplete, way of understanding communications.

In Shannon and Weaver’s model, communication is content, transmitted through a medium, disrupted by noise. That makes some sense – the model was developed in part in response to studies of communication during wartime. If you want to understand communications and start from the battlefield, then communication looks like a challenge that needs to be overcome – how the hell do we ever communicate? It’s a model based on examining the failure of something, not based on examining how communication actually takes place. That’s something we often do in our culture – we study failure, not a functioning system, and that study can lead to weird distortions in our understanding.

Hyperlinks, on the other hand, assume a path through an existing world. They’re generative paths – making a new hyperlink increases the abundance of that world. In the age of links, we assume that communication is possible, not a challenge. This model breaks the information age understanding of communication.

We can build models of anything using bits. We’re all familiar with the critiques of models – we’ve just seen models of a financial system lead towards massive disruption. The Department of Energy required a 10,000 year model to evaluate storage of nuclear material at Yucca Flats. This helps show the absurdity of models – we assume regularity and predictability, but there’s all sorts of possibilities that we can’t model. (David shows us dinasaours destroyed by asteroids – “didn’t plan for that!”) Models exclude that which doesn’t fit. They inherently deny the abundance of the world, the overflowing, uncapturable abundance of the world. And they’re purely formal – they leave something critical out: the body.

A bit is a measurement, the measurement of a difference. Any other measurement measures something in particular – weight, length. Bits just measure difference, which means they apply to everything. But the world never shows itself to us merely as a difference – it only presents itself in particular ways, differences between things we perceive in light, warmth, texture. The pure formality of bits comes from the fact that they are exactly how the world is not.

Returning to the Shannon-Weaver model, David asks us to focus on noise – the disruption in the system. But the noise is the world. The world shows up in Shannon’s diagram as the problem with the system. “In the system of abundant hyperlinks – this abundant system which is beyond systemization – we’re embracing, not avoiding the noise”. On the internet, we know that this information is interesting to at least one person because they put it there. It is a web of noise – that’s where it gets its strength.

Moving from the age of information to the age of noise on the web, the problem might be that the web is not noisy enough – we’re not appreciative enough of the differences.

I was catching up with David’s words and wasn’t able to accurately transcribe the questions. But I caught John Palfrey’s question, which asked David whether this new project represented a shift from the normative to the descriptive. Intercepting, and making the question much less polite, I rephrased, “So what?”

David argues that there is no “so what” – he’s intrigued by the idea that we embraced this broken model of the world. But it’s not a prescriptive or polemic project, he argues. I beg to differ – I think he’s starting to articulate a vision of a linked age, rather than an information age, and starting to think about the implications of how that age might work. And I suspect that he’ll be talking about those implications in the next few talks…

David offers an outline of his talk on his blog, though he tells me that the outline is an earlier version of the presentation. J’s of J’s scratchpad offers her notes from the talk as well – might be useful as a complement to my account here.