Minimum device Complexity for reading JPGs? BMPs? MP3? Text?

[dcarson]

Real time adds complexity. Taking minutes or even longer to unpack a jpg can be fine but audio or video takes a lot more speed. From my interesting quote file.

A friend of mine and I just calculated that if he stored his half-hour TV program (he does infomercials) on 400-byte/meter paper tape with at a width of 8 bits, his commercial (saved as an AVI) would take 16.1 million meters of paper tape. To be played back in real time, it would need to travel in excess of over twenty thousand miles per hour.

[Peter Knutsen]

Originally Posted by dcarson

Real time adds complexity. Taking minutes or even longer to unpack a jpg can be fine but audio or video takes a lot more speed. From my interesting quote file.

Yes. Wanting stuff done in realtime is often what costs Complexity. Barring the need to juggle around large amounts of data (in a sufficiently fast CPU cache-like storage), even a puni computer can do anything - given enough time.

Originally Posted by dcarson

A friend of mine and I just calculated that if he stored his half-hour TV program (he does infomercials) on 400-byte/meter paper tape with at a width of 8 bits, his commercial (saved as an AVI) would take 16.1 million meters of paper tape. To be played back in real time, it would need to travel in excess of over twenty thousand miles per hour.

How fast is that relative to lightspeed?

[jeff_wilson]

Originally Posted by vicky_molokh

Another one came up yesterday: complexity for 'reading' voice. As in, identifying the language spoken and/or converting it to plain text. Former is less important than latter, but knowing both would be nice.

As Anthony says, this began about 1997 though it is still not 100% reliable today. I don't know of any apps that identify multiple languages from audio.

Originally Posted by vicky_molokh

Edit: voice synthesis from text is another interesting one. And translation. UT plants translation at C3-C5 depending on fluency level achieved, but I keep hearing urban legends about pocket audio translator devices existing in the early 90s/late 80s, though I'm very sceptical of their fluency level.

Ray Kurzweil invented general text-to-speech in 1976 with a table full of specialized devices, but that would still fall under C-1.

Real-time voice translation depends on the above, not-yet-100% speech-to-text technology, but I've seen a recent demo where a guy gave a speech in English with a ~1/12 word error rate conversion to text, and a ~1/8 character error rate translation to Chinese, plus a voice recognizably his read the Chinese text despite him not speaking Chinese himself. So, C5 might represent that performance on something like a mini today.

I recall voice translators sold by Sharper Image or DAK-style ads back in the day, but the ads said you typed the English word and got a foreign word in text plus a synthesized "bon.jour." or whatever. Someone may have floated an early version that let you record your own voice for input comparison or output or both, but vocab would have been tiny, <1000 words, possibly <250 as I dimly recall.

[johndallman]

Originally Posted by Peter Knutsen

How fast is that relative to lightspeed?

Really slow. About 0.003% of the speed of light.

[Peter Knutsen]

Originally Posted by johndallman

Really slow. About 0.003% of the speed of light.

Thanks for calculating that.