Something weird happened recently: My mum asked me if I had heard of this thing called machine learning (ML). Other than thinking she is hanging out with the wrong crowd, it got me thinking how mainstream ML has become. However, this “new oil” hasn’t always been so hot, with some false starts involving Turing and visions set at Dartmouth in 1956.
Executive SummaryEngineers and data scientists with no experience in insurance can transform the industry simply by getting hold of the data and adopting cloud-first strategies, writes Lakshan DeSilva, CTO of Intellect SEEC. Here, he explains how the evolution of machine learning from not working to neural networking has gotten us to this point.
The first false start began when computers sought to translate English to Russian using artificial intelligence (AI) techniques, an effort that ultimately failed due to the intricacies of language. Minor progress was made in the 1970s by constraining the use case (“Micro-Worlds”), but when the capabilities were not able to be applied outside of the lab setting, this second attempt was eventually deemed a failure. A new approach in the 1980s—where Expert Systems attempted to mimic the work of doctors, teachers and chemists to create a clever computer—showed promise, but failed given its linear approach. One expert system was not able to build the next expert system.
Member Only Content
To continue reading, purchase this article or become a member.
*Already have an account? Click here to login