Web 2.0 – From the Labs (Google, IBM, Microsoft)

Web 2.0 – From the Labs
Peter Norvig, Ph.D., Google
Richard Rashid, Ph.D., Microsoft
Jim Spohrer, IBM Jim:
IBM has worldwide labs. What you might not now is that we collaborate with research facilities around the world in research and development activities. We’re working on measuring the spin of a single atom which will allow microscope to do 3D imaging of molecules. Blue Jean is the fastest computer in the world and is working on the protein holding problem.

IBM is working to establish service science. 50% of our revenue comes from the services business. 50% of GE’s revenue comes from their services business. Product companies are turning into services companies. The whole global economy is doing this. Why? The labor foce is shifting into services. But the educational world isn’t doing enough to teach about services innovation.

We’re working on statistical machine language translation. We’re using named entities and world clusters to allow the computer to understand meaning better. Our machine translation takes one language in and outputs another language. We’re using a statistical approach which takes into account phrases, templates, discriminative training and lots of data. Our Arabic to English translation is not quite fluent, but it is completely understandable.

Demo of named entities: We break text into sentences and then match sentences against patterns. We discard noisy data and regularize over names. We also use the relations of concepts and the nesting sets of concepts to understand the concepts.

Microsoft is working on empowering the individual through the democratization of science and enabling new forms of social information. We have terraserver which is a 98 terabyte web database. Also, skyserver.sdss.org which is a server and services that includes information from worldwide astronomy databases (such as iova.net and skyquery.net). We’re working with the NIH (National Instititute of Health) to create a central database of their published works called PubMed. Also, wwmx.org and wallop.com.