l Speech Recognition Software Images:
SchmasBot
Image by schoschie
See-through drawing of a robot Schmas. At the time I drew this (around 1995), the technical specs (a couple of DSPs running at 440 MHz for processing video and audio; 4 RISC CPUs running at 200 MHz as system processors; 64 MB RAM and 4 MB ROM; a 2.5" 2 GB harddisk; an olfactory molecular probe capable of detecting 8000 different volatile substances of max. 400 g/mol) were slightly futuristic; today, they’d make you yawn. Except for the molecular probe, maybe, but things like this are basically around.
The only thing that today’s technology is not able to do is: the SchmasBot is supposed to be able to translate 200 different languages, using audio speech. Natural speech recognition is still somewhat in its infancy, apart from stuff like the answering machines used in call centers for special purposes like train/plane bookings etc. These are much easier to do as they have a very limited vocabulary. Also, I doubt you would be able to fit the software required to recognise and synthesize natural speech for 200 different languages into 2 GB of harddisk space (that would be 10 MB per language… hm…. with strong compression… maybe, but then there would be nothing left for any of the other subsystems).
Oh, and the SchmasBot would probably only run for a few minutes with all that hardware built in, only powered on two 5V/6.8Ah rechargeable cells.
