AI fails: why AI still isn’t ready to take your job

The fear of AI-fuelled job loss is spreading. No matter what industry you’re in, AI-powered bots and software are taking a crack at it. AI seems to be ringing the death knell for all manner of jobs, tasks, chores and activities. From doctors, to customer service, to household assistants, no job feels safe.

Naturally, this has people worried about the future. But is AI ready to take over our jobs, or even likely to do so ever?

Prevalent AI fails would suggest not. Here’s how recent artificial intelligence mishaps show that AI still isn’t ready to take our jobs just yet.

From hospitals…

AI is going everywhere, and even doctors are feeling threatened by the new tech. For example, it is now infiltrating hospitals to help with oncology, clinical trial matching and genomics. It all sounds rather impressive. Sadly, the AI hasn’t yet lived up to the claims.

It was reported last year, following some leaked documents, that one such AI supercomputer had been poorly trained to assist with cancer diagnosis. The program failed to perform its basic function, instead making several incorrect and unsafe recommendations. In fact, it was suggested that the program isn’t usable in most cases.

In this example, then, even the best medical AI was still unable to perceive and understand things the same way a human can. Plus, being a good doctor isn’t just about diagnosis and treatment. There’s a distinct need for the more human traits that enable a good, reassuring bedside manner — something AI isn’t likely to achieve for a while yet.

In the future — if properly trained — AI may well find a permanent place in hospitals. However, it’s unlikely that AI will be capable of taking the entire role of a doctor. Rather, it will be a tool to assist them. It’d be more like a glorified computer than a human doctor — so, hospital computers might be out of a job. Doctors? Not so much.

When it comes to medical care, AI still needs to be propped up by human flexibility, empathy and understanding.

…To hospitality

Hospitality is another industry that artificial intelligence appears keen to get stuck into. But, once again, left alone, AI really isn’t very good at it.

Take bartending, for instance. A required, integral and basic ability for a bartender is the capacity to pick up a glass. Simple enough, unless you’re an AI robot. Not to mention, bartending is more than handling glasses. If AI fails at that, it can’t hope to recognise and manage drunken behaviour, or chat to regulars, or adapt service to suit a stag group or a corporate event accordingly.

Where AI might succeed in hospitality is in logical, fact-based tasks — such as streamlining the check-in process at a hotel. What it can’t do, is offer human understanding and a welcoming smile.

In other words, large areas of hospitality need the human touch to make them, well, hospitable.

Meanwhile, in business…

Hospitable service isn’t just for hospitality, either. Customer service jobs are also proving tempting for artificial intelligence. In fact, customer service is one of the best examples of AI being unable to take jobs, but proficient in supporting human team members.

There have been multiple AI fails when it comes to customer service. For example, Fabio the Pepper robot lasted only a week in his customer service role at a Scottish supermarket. Why? He confused and scared customers. Fabio’s failure highlights an important lesson for AI in customer service: you can’t bank on flashy tech alone to create a great customer experience.

This kind of AI customer service flop happens online and in apps too. Many chatbots have, at some point, failed to understand basic messages, or attempted to do too much at once. But chatbots have started to settle into their customer service role — alongside human employees, rather than instead of them. It’s here that the failures, and the successes of AI, demonstrate that AI can’t necessarily take over human jobs — but they can support them.

AI isn’t rising to supremacy in other areas of business, either. Take, for example, Amazon’s secret AI recruiter, which had to be scrapped after showing a distinct bias against women. The problem is, artificial intelligence doesn’t have a moral compass. It does what it is told to do but doesn’t comprehend whether it’s the right thing to do. Only the human touch provides that level of understanding, and it is humans that need to guide and support AI in business.

Putting your feet up: home assistant fails

Often, AI fails at home too. This might not threaten jobs, but it does give more insight into the abilities and shortfalls of current artificial intelligence. In particular, AI home assistant failures show the shortfall in voice recognition.

It turns out, even established AI assistants have bad days, and Amazon’s Alexa provides us with the perfect example. The AI assistant ordered an expensive doll house and cookies when asked by a six-year-old. When the instance was reported on the news, several other Alexa appliances attempted the same thing, at the request of the TV. It could not differentiate child and adult voices, and it listened rather too closely to the television set.

But Alexa is still pretty good, and popular in a lot of homes. Plus, newer AI assistants will be able to do even more, right? Well maybe, but if LG’s Cloi (pronounced is anything to go by, we’re still a long way off. At the unveiling of the helpful smart home assistant, Cloi, at CES 2018, the AI repeatedly failed to respond to requests. This made for an awkward and embarrassing unveiling — one that demonstrates that we still have a long way to go before AI voice recognition is reliable enough to enter the workplace.

Artificial intelligence is supposed to enrich our lives and make things easier. That’s the goal with our AI assistants and home, and it should be the goal with AI at work too. AI still has a long way to go. Even when it is doing well, it needs the human touch to augment its ability.

The hype, the headlines and the human

The rumours of the impending AI takeover have been greatly exaggerated. Headlines and hype have fuelled the fear of an AI-pocalypse.

The reality is far less imposing. AI will be able to handle some of our daily tasks. It might be able to handle a few of our jobs. But it can’t do anything without humans behind it, not just to maintain and monitor it, but to provide the human touch that AI can’t.

Artificial intelligence is becoming capable of an increasing number of tasks. But only humans have the empathy, understanding and flexibility that not only guide AI, but make jobs worth doing.

Note: we originally published this article here: