The use of AI can be risky, as was proved a few days ago. Most of us keep marvelling at the wonders of AI. All too often it seems to be pure magic and set to outperform humans. We are astonished and impressed by the way that AI can now hold “intelligent” conversations, answer questions, and perform uncanny creative tasks, such as painting (see another blog).

But from time to time we are made aware that some AI behaviour is not as it should be and that caution is needed when using it for decision-making processes, for example when recruiting or making investments, or dealing with medical decisions.

The focus of such warnings is usually “bias”. Manifestations of bias seem to be common. Clearly, if the data that the AI system is trained on contains gender or political or health-related biases (for example against smokers), then AI will be equally biased, resulting in unfair or un-ethical outcomes.

And when such effects come to light they erode people’s confidence, be they job seekers, patients or consumers. They become wary of employers, or insurance companies, or mortgage lenders who use AI. This will also have reverberations in other areas where AI is establishing a presence and where we are expected to entrust our lives to AI. Think of driverless trains and buses, but also of online help systems, where you get support and advice not from a human, but from a bot. This can certainly be a rather frustrating experience.

A crass case of AI going horribly wrong was reported widely a few days ago. Kentucky Fried Chicken (KFC), the fast-food chain, had sent out a promotional message to customers in Germany, suggesting to commemorate Kristallnacht with a dinner of chicken topped with extra-generous amounts of cheese.

In a statement KFC Germany blamed the message on a bot that had initiated an automated push notification. Such notifications are linked to national calendars with the aim of stimulating sales on special occasions. No-one had alerted the AI-driven bot that Kristallnacht (1938) is a day of national shame and certainly not cause for celebration. It in fact marked the beginning of the Nazi’s devilish plan to annihilate the Jewish population.

The message went out to everyone using the KFC app – causing a huge storm of outrage and bewilderment that such a blunder could have gone undetected.

In a statement issued to the Jerusalem Post on Thursday, the fast-food chain apologized for the error, explaining that they “use a semi-automated content creation process linked to calendars that include national observances. In this instance, our internal review process was not properly followed, resulting in a non-approved notification being shared.”
“We understand and respect the gravity and history of this day, and remain committed to equity, inclusion and belonging for all.”

But the damage is done, the embarrassment, the hurt and the outrage enormous. This will surely be a wake-up call for everyone to be extra cautious when using AI.