Prediction 16
Duration 18 years (02002-02020)
Predictor
Gregory W. Webster
Challenger
TBA
Add your voice to a conversation with the bettors:
Bookmark this bet, and share it with friends:
All the tools are so close...wearable computers, voice recognition is growing in its capabilities. Information overload is hitting more and more people who are no longer capable of keeping up. With the advent of a simple AI or even tools like Google's search engine, a supplemental memory is inevitable.
Sign in to challenge Gregory W. Webster to a bet on this prediction!
That by 2020 a wearable device will be available that will use voice recognition capability and high-volume storage to monitor and index conversations you have or conversations which occur in your vicinity for later searching as supplemental memory.
Does this bet require that such a device store things in "ascii" equivalent? I'm involved in a project to create such a device that would store video/audio/biometric data digitally(with authentication such that the data would be suitable as evidence in court). We have no plans nor do we feel an immediate need for voice recogition(except perhaps to index the data). Our higher priority is face recognition.
The project is described at www.badgecam.com
We are a bit further along than is indicated on the page(we need to update it).
RJB
This bet is way to conservative. The technology is available now. BBN has a conference call organizer that listens and transcribes conference calls including who is speaking and what the subject is and portable devices currently exist with enough memory and computing power to store hours of voice data in mp3 format.
Does this prediction say that the portable device must do the voice recognition or is it ok to transfer the information to a central/powerful system where the voice recognition and indexing can be done?
I had already been cobbling together some hardware do something similar to this for my own use (I suppose that means that I can't patent the idea and make a billion dollars selling it.) The technology is already there for a prototype version of this.
How could you win? A more interesting bet would be cybernetic memory enhancment.
If it's that certain perhaps the bettor should bring the end date nearer.
I can recognize voices stored in my video camera.
I can record hours and hours with my high capacity tapes.
I can monitor people talking by me and organize this by time.
I can later see this tapes and remember what happened.
I can attach my camera in my head, shoulder or hand.
I can by this camera in almost any major department store.
I can recognize voices stored in my video camera.
I believe the term "voice recognition capability" was meant to imply that the device would respond to YOUR voice, not the other way around.
From the wording of the bet it's not just recording conversations, it's saving them in a way that is meaningful to the machine. That way, the machine can search the conversations for particular text - in the same way they can currently search text documents. This does not exist.
The device would be incredibly useful (And raise privacy issues) but I don't think it will come about. There is voice recognition technology around at the moment - and some of it even understands most of the words - but it has to be trained, and therefore can only convert one voice to text. For this device to work, it would have to translate all sides of any conversation into meaningful (to the machine) data, regardless of accent, different pronunciations of words, and changes in sound as a result of changing emotions. I don't think that device will come in the first half of this century.
doowhacker said:
"From the wording of the bet it's not just recording conversations, it's saving them in a way that is meaningful to the machine. That way, the machine can search the conversations for particular text - in the same way they can currently search text documents. This does not exist."
"The device would be incredibly useful (And raise privacy issues) but I don't think it will come about. There is voice recognition technology around at the moment - and some of it even understands most of the words - but it has to be trained, and therefore can only convert one voice to text. For this device to work, it would have to translate all sides of any conversation into meaningful (to the machine) data, regardless of accent, different pronunciations of words, and changes in sound as a result of changing emotions. I don't think that device will come in the first half of this century. "
Having trained quadriplegics in use of advanced voice recognition and seen the instant return on a modest investment in hardware and software, I must disagree. We're just about there.
Moore's Law alone will assure that the necessary computing power for speaker-independent, untrained, highly accurate voice recognition will exist in the next 10 years. For some time, now, Dragon Systems has had a solid-state voice recorder intended to be carried on the person, then docked later to the user's computer, where the voice notes are automatically recognized by Dragon Dictate and transcribed into text - searchable text analogous to ASCII. Currently, Dragon is advertising 90 percent accuracy on voice which the system hasn't been trained to recognize. That's better than I can manage, most days.
So the elements of the system DO exist - the only things lacking are incremental performance improvements in recognition quality.
The only privacy issue I can see is interception and/or theft of one's own files by an intruder into one's computer system. If someone tells you something, capturing the text of that conversation electronically is no different than writing one's self a note to avoid forgetting what was said - except that a computer is used, and not a pencil. The person speaking to you ordinarily would expect you to remember what he or she said to you, so there's no privacy issue there.
For this device to work, it would have to translate all sides of any conversation into meaningful (to the machine) data, regardless of accent, different pronunciations of words, and changes in sound as a result of changing emotions. I don't think that device will come in the first half of this century.
I agree. It will come though.
First, the device would have to store every single thing we've ever said, since birth. It would keep the sound. Through continued use, it would re-interpret all recorded speech throughout it's lifetime.
So, let's say it can understand the entire range of sounds from the user, from my voice. What happens when I talk to someone? That persons device, lets say yours, knows you. If you let me, our devices will trade text transcripts of the conversation. If not, I have my text and your sound; you have your text and my sound.
Regulation of such a device would be under debate for many years. I don't think we'll see it by 2050.
While technology to make this possible might be available, it falls foul of the "who would want it" test.
The beauty of spoken conversation is that it is open to interpretation by inflection and knowledge of the other party. I can tell a friend he's a **** **** over something inconsequential with a smile on my face and it will be taken as meant, but if he could somehow recover this later it might be taken out of context and even used as the basis of a slander accusation (or would it be libel once it was electronically stored?).
At the risk of sounding frivolous, there is already a similar product, my wifes brain!
what a pussy bet...c'mon this is sooo mid-decade at the latest
Please sign in to comment.