Source – networksasia.net
Artificial intelligence is the buzzword for 2018 and it is being used everywhere and often to describe fairly mundane automation and analytics-driven processes. One of the earliest adoptions has been speech to text, for example in call centres. AI enables the speech of both the call centre operator and the customer to be converted into text files and then an algorithm scans that text looking for keywords that may indicate an area of risk for the enterprise or sentiment analysis of the customer.
“All AI today is narrow-focused, in other words, we have a system and we give it a specific challenge or task, a series of input data and a whole bunch of training data which is usually labelled or pre-classified,” said Charles Sevior, CTO Unstructured Data Solutions, APJ and Greater China, Dell EMC, in an email interview with Networks Asia. “We are at the early stages of having computers accurately interpret unstructured data in a time so fast that it is considered to be “just like a human” – so-called AI.”
In the interview, Sevior further talks about AI, such as the critical things that organizations should know before applying AI and machine learning. He also answers question regarding the Internet of Things and Big Data.
We’ve seen the videos of robots working in warehouses. How far are we from the machine revolution?
Autonomous capabilities are becoming known as ‘artificial intelligence’ but of course these systems are a long way from being intelligent or sentient. The definition of ‘machine revolution’ is blurry – we have had highly automated processes and production lines with robotics for many years. These processes reduce the need for manual repetitive labour and elevate the human functions to that of control, supervision, intervention and process refinement. This will continue to evolve for many years, so we view this as an evolution, not a revolution.
Do enterprises understand what AI means for now?
AI is the buzzword for 2018 and it is being used everywhere and often to describe fairly mundane automation and analytics-driven processes. The key is the definition and difference between AI and BI. Business Intelligence has been the process of rapidly analysing massive amounts of Structured Data, derived from operations & transactions that is in a machine-readable format.
We have been using SQL databases and data warehouses for decades. This then evolved into Hadoop – which enabled a new type of low-cost general-purpose computing to dramatically increase the data capacity at much lower cost using the HDFS storage protocol. Hadoop also made it possible store the original source data – so called “semi-structured data” – and then to define the database schema at time of Read rather than Write. This moved the ETL process out of the critical data path providing greater scope for Data Scientists to ask new questions of data – for both fresh and historic information.
Now we have a tool that helps us to direct computer analytics processes at the 80% of the world’s information that is truly Unstructured Data. This encapsulates all the information stored globally that is sourced from video, pictures, audio, radar, LIDAR and other type of sensors (such as seismic, thermal etc).
The only way we know what is in this unstructured information is to apply human intelligence – to look, listen and interpret – and then to annotate that data by entering Metadata. The moment of a great World Cup goal is marked by a human video replay operator and labelled with information about the teams playing, goal kicker, assist, goal keeper, time, date, venue etc. We get the benefit of that metadata in the form of video highlights.
Now we are becoming familiar with so-called AI processes such as object recognition, logo detection, speech-to-text, video sequence labelling and facial recognition that can do some of these functions via an automated process. We are at the early stages of having computers accurately interpret unstructured data in a time so fast that it is considered to be “just like a human” – so-called AI.