IT Pro: When do you think accurate transcription of video will happen? Do you have a timeframe for that?
I think it'll be about a decade before it's really, really good and when it's real-time. We are learning and advancing at exponential rates, so there could be some breakthroughs which will allow it to happen sooner, but conservatively it's about a ten year horizon.
IBM has a project underway called BlueBrain. It's a ten year initiative to create a working model of a human brain not to simulate, but create... down to every synapse, every neural network, every chemical interaction.
IT Pro: So you think ten years is a pragmatic, conservative time frame even though it's an incredibly complicated field? It's not just a matter of transcribing what they say, there are vocal inflections, nuances, colloquialisms, visual cues etc.
Absolutely, it's very complex. I think the paradigms for how we parse language, videos, photos, we will start learning how to do that at exponential rates. It learns in exponential ways. It's not linear because of the connectivity that we've got. Look at the work, simple work, Microsoft has done with Photosynth or Sea Dragon. Photosynth takes photos from all over the world and can recreate 3D scenes by stitching the photos together. They can look at the EXIF data of different photos and analyse the pictures. Just 15 years ago, the technology to stitch photos together was really complex and really expensive. Today, any commodity piece of technology can do it. You can buy a camera that does it in real-time as you move your camera around it will stitch your photos together. The technology is evolving at such exponential rates, that yes it's very complex but I think we'll see some breakthroughs.
I'll give you one other quick example IBM has a project underway called BlueBrain. It's a ten year initiative to create a working model of a human brain not to simulate, but create. That's a big distinction. Down to every synapse, every neural network, every chemical interaction. The project manager believes that after a decade a consciousness will evolve from this artificial brain. Whether or not it will, I don't know. But the point is that we're learning to create neural networks, we're putting a lot of money in quantum computing which is one of the tools that turns out is very, very good at neural analysis.
Over the coming years as we see more and more breakthroughs in quantum computing, quantum networks, neural nets, AI, we finally have the computing horsepower necessary and all the connectivity to enable us to start analysing rich things like human brains too.
