Some secondary thoughts on the Internet waking up, and stuff like that…
The internet is a network with lots of nodes (~2 billion computers) and links, with lots of messages being passed around in parallel, so there is an obvious similarity to a brain. There are less messages than in a human brain (~100 billion cells), but AFAIK the information contained in each message is typically greater than that in a neuron firing (1 bit to ~1000 other neurons).
See here for an estimate at the total processing power of the internet (ignoring message delays & bandwidth), and here for some calculations of how fast a human brain is.
For the purpose of assessing whether the internet is intelligent, do we include people in the system or not?
The internet must surely be considered to be at least as intelligent as the most intelligent user currently connected. Think of a single person using
a single computer. As a combined unit they must be at least as capable as either alone. Adding more computers can’t hurt.
Well it doesn’t do much by itself (yet) so it would have to be considered to be fairly stupid at the moment. Or am I being a bit unfair?
What would make it (more) intelligent? You’d probably want some AGI software running on each node or at least on one. Perhaps something like SETI@home (e.g. using BOINC) would help here?
More, and more efficient nodes, links and messages, would obviously enhance any intelligence that is there, whether humans are included or not.
What do you think? Does anyone know of any projects along these lines?
Is this a competitive approach to increased intelligence? (c.f. cyborgs, AGI)
Two billion computers is fiftyfold less than the number of neurons in a human brain. Also, the human brain has many kinds of neurons, each carrying out specialised functions. Contrast this with today’s computers which (apart from a difference in how many calculations per second they perform) are all essentially the same. Computers are not as densely interconnected as neurons are, and each computer can only process one incoming bit at a time.
Rodolfo Llinas compared the Internet to a jellyfish. A jellyfish does not have a brain or central nervous system. Instead, it has a loose network of nerves in its skin, called a ‘nerve net’. According to Micheal Chorost, “with its approx. 2 billion relatively unspecialised, homogenously organized nodes, we can think of the Internet as a sort of super-jellyfish. It doesn’t have much differentiation and specialisation”.
The Internet is a vast collection of resources. It retains a vast amount of information, but it does not understand that information. It is difficult to say whether it is ‘intelligent’ or not. There is, as yet, no reason to suppose an intelligence as smart as a human’s has emerged on the Internet but maybe we shouldn’t limit intelligence to ‘humanlike’. Maybe the Internet possesses a kind of intelligence that is nonhumanlike?
On the question of whether or not we should include people as part of the Internet, I think that we should. In Vernor Vinge’s 1999 paper about the Singularity, the author outlined several ways in which an intelligence might become recursive (ie, an intelligence creates something more intelligent which creates..and so on). One of these was “Large computer networks (and their associated users) may “wake up” as a superhumanly intelligent entity”. There was also “Computer/human interfaces may become so intimate that users may reasonably be considered superhumanly intelligent” Clearly, these scenarios are not exclusive. Perhaps we should expect some combination of the two.
Elkhonon Goldberg has compared Google to the frontal lobes. The frontal lobes gets many ‘votes’ from different parts of the brain and its job is to select what is most important at any time. This is roughly analogous to what search engines do. Google aggregates the votes of many humans and selects what is most important, putting it at the top of the search page. Chorost compared the pagerank algorithm to hebbian learning, saying “a highly ranked page will garner more page views, thus strengthening its ranking. Just as neurons that fire together wire together, pages that link together ‘think’ together. If many people visit a page over and over again, its page rank becomes so high that it effectively becomes stored in the collective human/electronic longterm memory…so we can say Google functions as a primitive hippocampus, the part of the brain that decides which short-term memories are worth converting into long-term ones”.
Remember, that it is not Google itself which is intelligent. Rather, (as Chorost put it) “It’s a combination of four things: Human declarative knowledge, human choices about that knowledge, a computer system that collects votes about those choices, and a high-speed far-flung communications network that integrates them all”.
There is an excellent book called ‘The Fourth Paradigm’. It’s a collection of research papers that look at data-intensive science: throwing gargantuan amounts of data at supercomputers/ computer networks, using algorithms to find patterns in that data, and converting the result into some kind of format that human patter-recognition capabilities can quickly glean useful information from. In short, its all about ways of optimizing human and machine intelligence and getting them working together as efficiently as possible. Definitely worth a look if you are after insights into how the internet plus its associated users’might become a superhumanly intelligent entity.
In the internet, each node can send a message to any other node using its IP address or (for well known machines) their Domain Name.
The message may go via lots of other nodes to get there, and the delay can be of the order of 0.1 or 0.2 of a second.
There are lots of different message types (TCP/IP, UDP, higher level protocols such as email, web). I’d expect the data in a message to at least balance the complexity of a neuron firing.
A computer can process more than one byte or word at a time due to multiple processors, RISC architectures, and probably other techniques.
There are lots of different servers performing different tasks, so although they may be similar in their hardware, they are specialised at a higher level. Human brains all look pretty similar (not that I’ve seen many 🙂 but we’re all very different people.
IMHO, the internet will probably win the race to greater than human intelligence as cyborgs and/or AIs will be connected to it before any reach a threshhold level of intelligence.
The internet is already self-improving if we count people as a part of it, it is just not all that fast yet. For example: Open Source, Online Brain Training, Online Education, Google search, Wikipedia
I personally think that irrespective of its intelligence, the internet is at least currently a force for good – people can meet others from completely different backgrounds.