The evolution of computer science: Connecting top-down and bottom-up

David Krakauer from the Santa Fe Institute asks “what is intelligence?” and discusses the rift in the field of computer science between the top down, symbolic approach to cognition (how can we make a machine play chess?)  and the bottom-up, inferential approach (how can we evolve a general-intelligence machine?).  He suggests that the singularity – when machines will outpace human beings – will occur only when machines master both aspects.  But is it a good idea to trust them?

This is a typical SFI talk, sweeping over evolving brain size, poetry, the Turing test, Evolution, the Matrix, Blade Runner and Doctor Strangelove. Fantastically, Krakauer mentions my work on a cultural singularity that I blogged about here!

Ultimately, his point is very similar to Adam Curtis’ documentary All Watched Over by Machines of Loving Grace, which is also brilliant.

See Krakauer’s talk here.

Physicists get linguist envy?

So I wrote a post a couple of weeks ago on my Hungarian friend’s blog in which I wrote about, amongst other things, why some linguists have physics envy, but I just read a new scientist article in which it seems physicists can have linguistics envy too!

Murray Gell-Mann, a nobel prize winning physicist (who discovered quarks), has taken it upon himself to try to work out the origins of human language:

Another pet project is an attempt to trace the majority of human languages back to a common root. Since the 19th century, linguists have been comparing languages to infer their common ancestry, but in most cases, Gell-Mann says, this kind of analysis loses the trail 6000 or 7000 years back. He says most linguists insist it is impossible to follow the trail any further into the past and – this is what truly rankles with him – “absurdly, they don’t even want to try”.

Gell-Mann heads SFI’s Evolution of Human Languages (EHL) programme. The EHL linguists say they can go even further back by classifying language families into superfamilies and even into a super-superfamily. “What we’ve found,” Gell-Mann explains, “is tentative evidence for a situation in which a huge fraction of all human languages are descended from one spoken 20,000 years ago, towards the end of the last ice age.” The team does not claim to account for all languages, though, and remains agnostic about whether they can eventually do so. “All of this just comes from following the data,” he says.

I love that attempting to trace the majority of human languages back to a common root can be described as a ‘pet project’.

If anyone’s interested here’s a paper he wrote on the subject:

Murray Gell-Mann, Ilia Peiros, George Starostin. Distant Language Relationship: The Current Perspective.