Occasionally, things that are deadly serious in the real world can sound more than a little silly on paper. An example, to my mind at least, are the proposed robotic urban combat drones being developed for the USAF. Why? Because apparently they will look like frisbees – armed and armoured frisbees, sure, but still remarkably like that thing your dog loves to grab out of the air. The drones will have pretty serious capabilities despite their comedy stylings, and will be tele-operated by soldiers on the ground as they hunt down insurgents.

We’ve created so many new words in the last few decades that we’ve had to invent a word to describe the invention of new words – neologism. The flux and change of language is a stumbling block for researchers trying to teach computers to understand human communication, so some of them wrote a piece of software that goes out and discovers the meaning of new words by itself. The program, aptly named ‘Zeitgeist’, crawls through Wikipedia looking for new words, and then examines the links to and from the page where the word is found to try and determine its meaning.

In the ongoing efforts to make the world a safer place, governments are turning to technology to replace fallible and inefficient human operatives. They’ll surely be interested in the efforts of Australian scientists to develop algorithms to detect “unusual behaviour in an open environment” in video surveillance footage. While the aim is basically laudable, one can’t help but wonder how error-prone these systems will be (given the poor success rates of facial recognition software here in the UK), and what misuse such technology might be put to by unscrupulous governments or corporations.

Lockheed Martin are the proud winners of a NASA contract to build their proposed new Orion crew capsules. The contract is to the tune of $4billion or so, and aims for a first manned flight for 2014 with a Lunar mission by no later than 2020 – which still seems a long way off, but will probably be upon us before we have time to think.

Software is getting better and better at doing things that were thought to be the province of conscious intelligence – crossword puzzles, for example. A program called WebCrow has just beaten a number of human participants in a bilingual crossword challenge against the clock. WebCrow works by searching the web using terms from the puzzle clues, and trying words of the right length from the results as answers. This isn’t true artificial intelligence, granted, but it is being hailed as part of “a trend to use the web as a shallow source of human knowledge for artificial intelligence.”