At its Analytics Experience conference in Amsterdam, analytics software supplier SAS has disclosed research that suggests the vogue for artificial intelligence (AI) is streets ahead of user adoption.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Its report, Enterprise AI promise study: path to value, is based on a telephone survey, conducted in August 2017, of executives from 100 European organisations in sectors including banking, insurance, manufacturing, retail and government.
The survey found that although most organisations have begun to talk about AI, few have begun to implement projects. The hindrances to adoption come, the survey indicated, from a shortage of data science skills, as well as limits in how data science is structured inside companies and organisations.
In an interview with Computer Weekly at the conference, SAS co-founder and CEO Jim Goodnight said of the current vogue for AI: “We’ve been doing machine learning for close to 20 years. Back in 2002, we were doing credit card fraud detection for one of the big banks in London, using neural networks. And if you include logistic regression, we’ve been doing that since 1977.”
The new thing now, said Goodnight, is “the fact that we do these big jobs using massively parallel computing”.
He added: “I’m not a dyed-in-the-wool believer that AI is the next great thing. It will be a tough slog, a lot of model building, and lots of trials and errors.
“We are years away from a machine being able to think. We can train models to forecast what words you are seeing or estimate an image on a TV screen. But even then, you have to train it to identify every single object, so it’s down to probability.”
Olivier Schabenberger, chief technology officer at SAS, said in a group press interview: “Who would have thought five years ago that we would be talking about AI? It’s a 30-year convergence between neural network technology with big compute that allows us to solve problems that we could have tackled before but with unsatisfactory accuracy. If your speech-to-text accuracy is 20%, you get frustrated; if it is 60%, you productise. That is what has changed and it has happened since 2012. We did not see that coming.
“In analytics, we are seeing a shift away from more craftsmanlike models that are time-consuming to a more automated factory style of modelling. So, yes, a data scientist could beat that sort of model, but you don’t have time to invest in finding that model.
“When we talk about automation, we don’t just mean large systems that autonomously run and replace some process. We think about it at the smaller level of an analyst going from exploration to building, verifying and deploying a model so it can become a useful asset – that is automation.”
The SAS survey found that 20% of executives felt their data science teams were ready for AI, while 19% had no data science teams at all. A total of 28% planned to recruit fresh data scientists and 32% said they would seek to build up AI skills in their existing analyst teams.
Almost half of the respondents distrusted so-called “black box” AI, in which a system cannot explain its results.
The study also looked at enterprises’ readiness for AI in terms of infrastructure: 24% felt they had the right infrastructure in place, 24% felt they needed to update and adapt their current platform, and 29% had no platform at all.