Why your data is far more important than which AI you use
Be AI-agnostic, because the quality of your data will determine your competitive edge
It doesn't really matter if you use Google, Amazon or Microsoft's AI algorithms, according to experts.
The real differentiator is not whether you use Lex, Azure AI or Google's machine learning tools, but the quality of your own data, speakers at OpenText's Enterprise World 2018 conference said yesterday.
"The algorithms are nothing without the data," Mike Gualtieri, principal analyst at Forrester, said. "Algorithms get all the press, but the competitive advantage, the accuracy of those models, the decisions that the models can make, it all comes from the data, and the data that enterprises have is very specific to them and their customers."
While Google and the other tech giants investing fortunes into machine learning models have trained their tools on billions of images, pieces of text and other media, none of these AI systems have the specific data that companies have to hand, Gualtieri explained.
"They've got a lot of web content but do they have the transaction information? They don't have that so enterprises shouldn't think 'I'll just wait and Google will do all the AI for me'. Every enterprise has their own valuable information and from that they are going to be able to create the models using machine learning."
That makes having quality data enterprises' primary concern, OpenText argued, before picking an AI or machine learning tool that can turn that data into insights.
Industries could also benefit from sharing data with one another, the analyst explained, warinng that enterprises should be wary of sharing information that gives them an edge over their rivals. "That balance has to be found about what's competitive and what's good for everyone in the industry to share."
Lalith Subramanian, VP of engineering for analytics at OpenText, said that with the advent of emojis and text language, AI was having a tougher time analysing conversations.
"It's not English anymore, it's abbreviations and all kinds of things," he said. "To me one of the most fascinating areas of exploration for this company has got to be around 'what is modern conversational linguistics' because it's not about strictly structured language with very formal grammars. You're having instead abbreviations and words and vocabularies [are] changing."
While every vendor is eagerly advertising the apparent benefits of applying AI to business processes, Subramanian said it's unlikely that drivers for using the technology would be very different to the usual factors of cost-cutting, increasing revenue, or fear of regulatory infringements.
"[Customers] say 'we would like to improve our customer service and build a chatbot'. Okay, but what's it going to take for you to buy one of these systems with a chatbot?" he said.
"You're not going to buy it unless you can lay off 15, 20, 30, 40 people in your staff because nothing else is going to justify it. Because I can't tell you your customer stats are going to go up by X per cent. It's got to be a little bit more of a concrete use case which usually I define by fear of jail, or greed."
The ultimate law enforcement agency guide to going mobile
Best practices for implementing a mobile device programFree download
The business value of Red Hat OpenShift
Platform cost savings, ROI, and the challenges and opportunities of Red Hat OpenShiftFree download
Managing security and risk across the IT supply chain: A practical approach
Best practices for IT supply chain securityFree download
Digital remote monitoring and dispatch services’ impact on edge computing and data centres
Seven trends redefining remote monitoring and field service dispatch service requirementsFree download