There is so much being written about AI that many companies are joining the AI bandwagon just because it seems to be the right thing to do.
During the recent EngineerIT panel discussion on the subject, “Has AI changed the face of automation testing”, editor Hans van de Groenendaal presented the question to two experts in the AI landscape, Nkgaphe Tsebesebe of the CSIR (a PhD student, mostly focusing on machine learning to perform diagnostics of communicable diseases) and Johan Steyn who calls himself a human centred AI advocate.
“This topic is of interest to me,” Johan said. “I worked for about eight or nine years in quality assurance, testing with several global vendors and for one of our banks. The past two years, my focus has been on ethical AI to safeguard our next generation.”
Both participants agreed that AI is an incredibly powerful technology and will solve many problems but companies should not create problems to justify their big spend on the technology. Common sense remains one of our most powerful tools.
“I think it's difficult for smaller businesses to afford AI although the cost is coming down,” Johan said. “I often see that we're trying to fix problems with AI that just would require common sense.”
He recently met with the retail banking leadership team of one of South Africa’s large banks. “The meeting was scheduled to be for an hour to discuss use cases in retail banking. I'm a customer of this bank, which now is also my customer. At the start of the meeting, I asked if I could make a quick call? I called this bank's contact centre number and I put my phone on speaker mode – and continued the meeting when my call was answered. We waited 45 minutes. Obviously, the meeting didn't progress. I asked ‘When last did you call the contact centre?’ I believe I made the point: AI is not the be all and end all of everything. It's sometimes simple things that are the solutions we need.”
Nkgaphe said AI may have limitations but, if you’re considering AI from technical perspectives, it is more of an umbrella term.
“It can include machine learning, deep learning and the data sciences in it. If you want to implement machine learning and you don't have enough data, you have reached the limitation.
“The biggest limitation of AI is the amount of data available to process a particular task. If there is not enough data to solve a particular problem, AI will not give us what we are seeking. Without enough data, at the end of the day, AI ends up guessing.”
The podcast discussion, covering a range of interesting AI-related topics, is available at www.engineerit.co.za – free for you to enjoy!
