On the occasion of its centenary, IBM presented Watson in 2011 , an artificial intelligence computer system capable of answering questions asked in natural language. It was named after the founder and first president of IBM, Thomas J. Watson. Now, the agency DARPA (Defense Advanced Research Projects Agency), which hosted the development of Watson, has released a free and open source version of this supercomputer: “Deep Dive”.
“It all started as part of a DARPA-funded project in which Watson was also involved. Watson is a question-answer engine and the goal of DeepDive is to extract a large amount of structured data from unstructured data sources, ”explains Christopher Re of the University of Wisconsin.
DeepDive incorporates algorithms based on the probability of learning, as well as open source tools, such as Madlib, Impala (from Oracle) or Hogwild and to feel familiar with all this environment it is essential to have knowledge of SQL and Python .
“Under the covers, DeepDive is based on a probability model , which is a very basic approach to building these systems. We use a RISC processor at the moment, and we are trying to make a compiler; We believe that machine learning will allow us to make it much easier to program for the next generation of DeepDive ”clarifies Re.
In the same way, in the DeepDive open project they plan to obtain many more types of data: images, figures, tables, graphs, spreadsheets … this system, according to DARPA scientists, must be a real data eater .
Deep Dive has an average of 10,000 downloads per week and it is likely that this number will not stop growing considering how attractive artificial intelligence is both for those who are passionate about this world and for scientists.
