Some good observations on TechCrunch by Vinod Khosla recently (here and here) on a potential renaissance of Artificial Intelligence in the upcoming years:
“Location, social networks, identity, and personalization are just the tip of the iceberg. There will be countless new types of data streams and new ways to make data useful. Fundamental data utilities and applications will be built on these and a lot more information and data reduction will be extracted from this hairy-looking big ball of often-unstructured data.”
This could play out in three dimensions:
- Natural interaction (voice, text, gesture) that finally is beginning to work as computing power has come down so drastically that many of the mainly statistics-based approaches can actually work and scale, producing meaningful results.
- A new metaphor for interacting with information and services based on an intelligent assistant, popularized primarily by Apple’s Siri.
- The availability of large data sets (primarily the explicit and, more importantly, implicit online behavior of hundreds of millions of users) feeding meaningful and non-trivial pattern recognition approaches that discover meaning in very large amounts of data.
Some caveats apply:
- It is unclear how far the intelligent assistant metaphor will scale. Anthropomorphizing the machine is useful because it drives end-user habit formation, but can quickly backfire where naive and unrealistic expectations cannot be met.
- If innovation is driven by collecting more data about more people then the open issue of how to deal effectively and on a global scale with issues of online privacy will rear its ugly head. In one potential scenario we may see more innovation and more customer value available to users in jurisdictions with less than stringent privacy rules and regulations. And vice versa.