a blog about things that I've been thinking hard about

Can the Internet Think?

16 November, 2005
every neuron is a voting system

The synaptic inputs to every neuron are like voters that vote for and against the decision to fire.

A small number of important websites are explicitly or implicitly vote-based.

In the brain, the internal "voting" neurons are more numerous than the neurons that connect externally.

To make the internet more like a conscious brain, we need more vote-based websites. Many, many more.


What is the "Internet"?

I want to start by defining what I mean by the Internet. The Internet isn't just a network of computers, it's a network of computers plus all the people who use those computers. So if I ask questions like "Can the Internet Think?" or "Is the Internet conscious?", I'm not asking questions about artificial intelligence, because the system under study already includes human intelligence. What I'm really asking is whether the Internet can be thought of as a single conscious entity.

A similar question is whether an ant nest can be considered a single organism. An ant nest is a system of tunnels in the ground plus a large number of ants. Each ant is an organism, but the nest as a whole can be considered a type of "super-organism", the components of which are the individual ants. An ant nest "reproduces" when it produces queens and drones which fly off and mate and start new nests. Other ways that an ant nest resembles a single organism is that it takes actions to maintain itself and defend itself against attackers.

There is also an analogy between an ant nest as super-organism and a person's body as super-organism, considering each cell of the body as an organism in itself (so that the ant nest could even be considered a "super-super-organism"). Re-applying this analogy to the consciousness of the Internet, we can suggest that our brains are "super-neurons", where each neuron "thinks" in such a manner as to contribute to the overall thought processes of the brain.

What it Means for the Internet to "Think"

If the Internet was a sentient conscious being, then the following statements would make sense:

In some ways these statements represent the understanding of a very naive Internet user, and we might feel compelled to explain to such a person that the Internet doesn't really "know" facts or "believe" things, and you can't "tell" it stuff, that actually it's other people doing the knowing, the believing and to whom you must be doing the telling. Or, at the very least, we would insist that attribution of knowing, believing and being told apply to specific websites, rather than to the Internet as a whole. So it makes sense to say things like:

Attention and Belief

Two critical features of a conscious being are attention, i.e. the ability to focus on a particular issue, topic or item of information, and a coherent system of belief and knowledge. Thought requires attention, in that you must be thinking about something at a particular time, and it operates upon belief and knowledge, in that your thinking is based upon your existing beliefs and knowledge, and the outcome of the thinking might be to alter those beliefs in some way, or to add to your knowledge.

So does the Internet have attention, and does it have beliefs and knowledge?

The skeptical answer, as I have already suggested, is that only individual Internet users have these attributes, or perhaps groups of individuals acting through particular websites, and it makes no sense to apply them to the Internet as a whole.

However, there are certain websites and Internet facilities which make some attempt to provide determine global attention and global belief or global knowledge. In all cases these sites provide some method for an unlimited number of Internet users to contribute to the attention or beliefs and knowledge encapsulated in the site's operation. Some major examples are:

A common feature of many of these systems is some type of voting, which may be implicit (as in Google) or explicit (as in Digg and Slashdot), or something in between (as in Delicious, where tagging links might be intended to be a vote, or it might be intended as nothing more than a personal bookmark). Google and Wikipedia are oriented towards the valuation of sites on an ongoing basis, whereas Digg and Slashdot are more news oriented. Delicious allows for a mixture of both.

There has been such an explosion of different voting systems and "social" websites recently that one might think we have too many of them. But we would do well to compare this to the human brain, where each neuron is in effect a voting system which receives "votes" from all those neurons connected to it pre-synaptically, and the votes are for or against that neuron firing an action potential. So the human brain consists of hundreds of millions of voting systems, all connected to each other, with each receiver of votes receiving them from thousands, tens of thousands or even hundreds of thousands of voters. The information flow in the voting far exceeds the amount of information flow coming in from the original inputs (from the senses) or going out to the final outputs (mostly to the muscles). If the Internet wants to be more like a human brain, it needs more voting systems.

The Internet's Attention

The sites I have mentioned provide something recognisable as attention, belief and knowledge. But will they develop so as to encompass the unique attention, beliefs and knowledge of the whole Internet?

Firstly I will consider attention. Sites like Digg and Delicious provide a notion of common attention, where possible links of interest are submitted, some users vote on submitted links, and the more popular links rise to the top, where they then capture the attention of large numbers of users who use the sites to find interesting stuff.

One feature of Internet "attention", which is probably always going to differ from human "attention", is that it will never be restricted to just one thing at a time. Delicious allows users to provide whatever tags to their bookmarked links, and each tag defines a potential target of attention. Users interested in "webapplication" might be completely separate from users interested in "knitting".

But if we consider attention within a given topic, whether that topic be narrowly or broadly defined, and consider it across different attentional websites, there are reasons for different attentional websites to converge to each other on each topic. If attentional website A presents an interesting story on "knitting", and attentional website B has not yet presented the same story in its "knitting" topic, then any user of site B who also uses site A will be motivated to bring this new story to the attention of site A users. So on a per-topic basis, there is going to be a tendency towards a unification of attention across all of the Internet.

The Internet's Knowledge

Something similar will happen with knowledge, although it is much easier to agree on what is "interesting" than on what is "correct".

An example of a topic where the Internet has apparently "learnt" something new is given in Internet Killed the Alien Star on Tech Central Station, where Douglas Kern writes that exposure of all the evidence for and against visitation of Earth by aliens has made most people realise that the evidence is too weak to be convincing, and most of the UFO enthusiasts have lost interest. I give this example cautiously, because if you go to Google, or even to Delicious's "popular" search, and search for "UFO", you will still mostly find sites run by people who believe in UFOs and not by skeptics

Wikipedia can certainly be seen as an attempt to create global unified knowledge. However it achieves this by disallowing "original" ideas, and by giving preference to knowledge backed up by authoritative sources (which are often not even Internet-based).

Google provides an indirect indication of authority, but on any given topic there is no guarantee that it picks the most reliable source. Google has also been accused of ignoring new sites altogether (in its dreaded "sandbox"), which limits its usefulness to those looking for good new information or ideas.

In as much as these sites or others can start to provide unified "best" knowledge about any topic, they will tend to converge in the same way that attentional sites would be expected to converge. For example, if a Wikipedia page on a certain topic is being edited, one would expect at least some of the editors to look up the topic on Google, just in case there is some relevant information that has been missed. And if you look for a topic on Google, and the Wikipedia page for that topic has been well written, then it will probably rank highly in the listings.

Is the Internet Just Starting to Wake Up?

Unified attention and unified belief and knowledge are both works in progress. More seems to be happening on the attention front, and it may be that the attention problem has to be solved before the knowledge problem can be solved. The solution to determine the "best" knowledge about a given topic, or the "best" answer to a given question, may depend on an ability to attend to all the relevant considerations. When attentional mechanisms can make all the relevant facts and ideas available to all the people interested in solving a particular problem, then this will speed up the process of determining a stable consensus "correct" or "best" solution to that problem.

There is a singularitarian aspect to all this, which is that, as the Internet develops the ability to "think" for itself, it will be able to think about its own ability to think, and it will think of ways to make its own thinking better. This positive feedback cycle may trigger an explosive acceleration of the Internet's total intelligence, possibly sooner than anyone expects.

Vote for or comment on this article on Reddit or Hacker News ...