Typingpool: Human Audio Transcription Parallelism

theodp writes "Silly rabbit, parallel processing is not just for Big Data! Building on techniques outlined by Andy Baio back in 2008, Wired writer and 20% Doctrine evangelist Ryan Tate has released Ruby-based software called Typingpool to make audio transcriptions easier and cheaper. 'Typingpool chops your audio into small bits and routes them to the labor marketplace Mechanical Turk,' Tate explains to his reporter pals, 'where workers transcribe the bits in parallel. This produces transcripts much faster than any lone transcriber for as little one-eighth what you pay a transcription service. Better still, workers keep 91 percent of the money you spend.' Remember to Use the Force for Good, Tate adds."

Comments

Most so-called AI is already done like this

[Anonymous Coward]

Computer AI (semantics automatically arising from syntax) is, of course, impossible. All so-called AI programs depend on Human derived rules.

Google, for instance, uses Human workers to constantly add new rules to their search engine, while allowing the unwashed public to think the same is done by magical computer algorithms.

Many companies in the last 15 years have claimed to offer high quality translation of written and spoken Human language by computer, when what they really meant was that a computer system would send the data to Human operators for real-time conversion.

The 'Mechanical Turk' approach is the reality of so-called computer AI.

Of course, with voice recognition, or language translation, we should be disturbed by the lack of progress by computer software. Both of these tasks clearly do NOT need any concept of AI, since we already have full understanding of the language rules needed by the algorithms. Why then does software still perform so poorly? The researchers of the 1950s would be horrified at our lack of progress. They assumed it was just an issue of processing power and storage.

Not only does real AI not exist, we can't even successfully handle the obvious pseudo-AI problems like language processing.

Rethink the 'Mechanical Turk' concept. Shouldn't we all be able to input into a 'super' MT system to give that system the benefit of our innate understanding of key rulesets? How many rules does a speech-to-text engine need? If millions of Internet users were able to input rules into such an engine, would it rapidly become near perfect. Surely we would all volunteer to be part of a MT to create better speech recognition and language translation? And NO, Google's lousy "suggest a better translation" doesn't count, given how poorly it improves matters.