'AI is still at the mainframe level', says OpenText's Chirag Patel
Computers still relies on humans for context, and that won't change in the near future
"Artificial intelligence is still limited to ‘narrow' AI," said OpenText's Chirag Patel at a recent Computing event. Instead of the general purpose AI that we see from Hollywood, where computers can be as good or better than humans at a variety of tasks, narrow AI can only do one thing - like playing chess, or Go.
It will take a long time before general AI is ready - more than 30 years, Patel, a senior solutions consultant at OpenText, expects.
"Computing began with massive mainframes," he said, "but now we're have smaller, more powerful devices in our pockets. In terms of AI, we're still at the mainframe level - it's in its infancy."
The most advanced use cases for artificial intelligence today are things like predictive maintenance, fraud detection and personalised advertising. Future concepts like curing cancer (a long-standing claim of IBM Watson) and a human-like workforce are still in the distant future.
One of AI's weak points is that it still needs human intervention to perform well, "because we can make associations that machines cannot - yet". An example is customer retention: a computer could use pattern recognition to recognise that people in a certain geographical area like or dislike a product, but would have trouble analysing why that is. "They are very good at processing and analysing data, but not at being innovative and contextual," said Patel.
OpenText describes this synergy as ‘Knowledge Engineering'. The company thinks that the combination of knowledge engineering and machine learning will be particularly important to derive value from unstructured data, by giving machines context. Patel said, "The five per cent of work that humans still need to do is very important."
While AI will prove to be very important in the coming decades, it isn't something that anyone can just jump into. Patel said, "We are very much at the start of this AI journey. Start simple and increase complexity."