If anyone thinks IBM has lost its cutting edge, especially when compared with tech titans like Google and Apple, then they probably aren’t aware of its advancements in cognitive computing systems.
A worker interacts with a robot at the IBM stand at the 2014 CeBIT technology trade fair in Hanover, Germany (Getty Images)
Four years ago, IBM’s cognitive computing system beat two human champions to win the US game show Jeopardy. More recently, the supercomputer that goes by the name ‘Chef Watson’ wrote a cookbook. And not just any cookbook, but one with surprising ingredient combinations people wouldn’t ordinarily think of, and yet, somehow work well together. These recipes, for instance, may just tickle your fancy:
- Indian turmeric paella
- Italian pumpkin cheesecake
- Turkish Korean anchovy Caesar salad
- Vietnamese apple and pork kebab
Knowing these were compiled by a machine, one might assume they were just random combinations a programme threw together, but that couldn’t be further from the truth. An incredible amount of data-processing and algorithms are behind those eclectic delicacies.
Chef Watson has been trained to understand content and flavour profiles, analysed thousands of recipes, and been fed huge amounts of culinary data, including research on the flavours that give people pleasure at a molecular level. One recipe, for example, calls for mushroom, strawberry, chicken and pineapple – an ingredient choice that at first glance seems incompatible, but shares significant levels of the flavour compound g-dodecalactone.
One imagines an AI brain to be analytical, logical, and basically, devoid of creativity. Yet, these recipes ooze with just that. It’s quite impressive, and leads one to ask what exactly cognitive computing does. Simply put, cognitive computing is designed to tackle especially human-esque types of problems, in other words, problems that are complex in nature. In this context, “complex” does not simply mean “very complicated”. That wouldn’t be anything special, because we all know computers are capable of processing extraordinarily complicated problems already. Complex problems are something else.
Complexity is characterised by uncertainty and ambiguity and relies heavily on context. Complex situations are dynamic and information-rich, where the data not only changes frequently, but can even be conflicting. Complex problems are the kind of problems where you might not be able to derive the “right” answers, but rather the “best” answers.
Complex situations, for example, have much to do with systems like psychology or language – in other words, the types of fields in which one senses an AI machine would be out of its depth. With complexity, it is not enough to just have a lot of data or clever algorithms. Much more is needed, which is the gap that cognitive computing is trying to fill.
This is what makes Cognitive Cooking with Chef Watson so extraordinary. It is essentially designed to deal with the type of problems typically characterised by the necessity for a human mind, to approach and come up with solutions. Problems like what to make for dinner. It bears mentioning the cognitive computing system wasn’t solely responsible for the dishes in the cookbook; there was also a team of professional chefs that helped iron out the recipes. Watson would work its algorithmic magic to come up with surprising combinations, then a human mind would help colour in the blanks.
However, the fact that the computer didn’t work alone to solve complex problems, but relied on human prompts and interaction, doesn’t make it any less of an interesting vision for future AI applications. In fact, it helps us envision a future where human and artificial intelligence work together, complementing one another, to deal with complex issues in a far more creative and resourceful way.
By Tasmin van den Heever
Follow me on Twitter: @tsvanden
Please note that the opinions expressed in this article are those of the author and not MWEB Connect (Pty) Ltd
Switching your ADSL data to MWEB is quick and easy. Simply switch your ADSL line and data to MWEB by selecting the ADSL line speed and data package that suit your needs.