Archive for August, 2006

Robohack

Friday, August 25th, 2006

The Financial Times last week ran stories about a news agency, Thomson Financial, claiming to have developed a software program able to automatically generate news reports. Here is one of the FT stories. Two aspects are interesting:

First, auto-generation of news stories is not new. In 1991, I saw the service developed by Prodigy Communications, a joint venture between IBM and Sears, Roebuck to provide online services. Prodigy’s software-reporter took input feeds from news agencies, such as Reuters and AFP, and turned these feeds into stories, complete with graphs, taking care to vary the tone, the voice (active or passive), the sentence length, etc, within and between stories.

Second, although the results are not new, the motivations may be. Prodigy was an online service aimed at consumers in the pre-browser days of the Internet. They wanted to provide up-to-the-minute news stories without the expense and trouble of employing round-the-clock journalists. (The Prodigy consortium had originally included a media company, CBS, and their model of news reporting was from radio news.) Thomson, on the other hand, seems driven by the desire to provide fast news to auto-traders, ie to get financial market stories out as quickly as possible:

“The question we asked is how we can deliver information fast enough for our clients to make an immediate decision to buy or sell,” says Matthew Burkley, senior vice-president of strategy at Thomson Financial.

. . .

In an effort to tap into the boom in automated trading, based on financial algorithms which set rules that determine whether shares, bonds, currencies, derivatives and commodities should be bought or sold, Thomson and Reuters are also trying to help their customers’ computers read the news.

Computerised trading reduces the need to buy information screens and terminals, especially if they replace actual traders. Companies would instead pay for feeds of information to the computers. Thomson will soon offer a feed of automated reports that can be traded on within a second of earnings releases, and is trying to develop a range of sophisticated automated feeds for its hedge funds.

Reuters is creating “machine readable news”, on which computers can automatically trade. Tom Glocer, Reuters chief executive, said this was “one of my favourite new projects”.

Already, a third of all US equity trading is driven by so-called “algorithmic trading”, according to latest estimates, as traditional fund managers follow the lead of hedge funds and embrace complex automated trading strategies in an effort to improve efficiency.

The figure is set to rise much further in the coming years as fund managers, along with brokers and exchanges, strive for ever-greater speed and control over the trading cycle amid heightened market competition and consolidation.

Algorithmic trading uses mathematical models, which are being developed by PhD-armed staff at a growing number of investment banks and specialist trading firms, as a means of trading large blocks of shares as quickly as possible. Yet once the computers have written the news, plugged it into their trading programmes and, with any luck, made some money on the trades, there remains one function human beings should be better at: going out and celebrating.”

(Note: This post was generated by a human being.)

Predicting footballer misbehaviour

Wednesday, August 9th, 2006

Prediction markets were used initially for political races and for economic variables, but recently they have found application in other domains where information may be distributed.  Applications now range from predicting influenza epidemics to predicting court appearances by Australian Rules Footballers.  

ACE Newsletter

Monday, August 7th, 2006

The latest issue of Leigh Tesfatsion’s regular newsletter on Agent-Based [tag]Computational Economics[/tag] (ACE) is now available, here.

Computational Social Choice

Friday, August 4th, 2006

Ulle Endriss of the ILLC is organizing the First International Workshop on [tag]Computational Social Choice[/tag], in Amsterdam, the Netherlands, from 6-8 December 2006.   In creating artificial societies of self-interested and goal-directed [tag]software agents[/tag], we need to define ways for them to make joint decisions, whether collaboratively or not.   This is an exciting area of research, bringing together people from [tag]computer science[/tag], [tag]political theory[/tag], and [tag]economics[/tag].

The deadline for papers is 2006-10-01.

Evolutionary ideas in economics

Tuesday, August 1st, 2006

The Economist magazine has a review article on [tag]evolutionary economics[/tag], which I think is free-to-view.

Source:

“The Cambrian age of economics.”  The Economist, 20 July 2006.