'Artificial intelligence is not intelligent, all it does is pattern match'
There has been a lot of attention paid to AI lately, and especially the generative kind.
But what exactly is it, and what do employers need to know before deploying it in the workplace?
“Generative AI is able to create new versions of things and that can go from being as helpful as helping you generate new text for a concept to as unhelpful as some of the deep fakes that are happening right now where you see the president saying something absolutely horrible on video, when the president never actually said that,” says George Westerman, senior lecturer at MIT Sloan School of Management in Cambridge, Mass.
It is just the latest facet of AI technology and computing that has evolved over the past 10 years, says Westerman.
“The algorithms for deep learning and other things have existed but it’s just the power of computing and the amount of data that we’ve had over the last decade or so that’s made it so much more possible to do now than they ever could before.”
The dangers of relying on AI output
While AI can be great at such things as preparing draft documents and creating descriptions of events, there are some things that HR departments especially should be wary, when using it.
“The most common things like all those that make work more efficient but obviously, employers should be very careful because of data issues. What comes to mind is that employees, even unbeknownst to them, even outside the tech sector, they have a lot of data that they carry, from personal data to payroll data; benefits, the position history, the prior employment, things that HR wrote about them and assessment that they received and all those words which were typed before can now be used by generative AI,” says Ori Freiman, post-doctoral fellow at McMaster University’s Digital Society Lab in Hamilton.
Employees who use the text output of such tools as ChatGPT should always double-check its facts, as it sometimes creates “hallucinations,” says Freiman.
And because of its nature, “one of the challenges we have with generative AI is that it doesn’t have all the knowledge that we have,” says Westerman.
“The way I say it is that artificial intelligence is not intelligent, all it does is pattern match. All it does is base predictions on other things that’s been done in the past. These algorithms don’t actually know what they’re saying and that means that they often say things that just aren’t right and you have to be careful with them.”
A small percentage of HR departments surveyed reported they are currently using AI technologies.
The best ways HR can benefit
There are some areas however, in which HR departments can use AI, according to Freiman.
“We all can think about generative job description, and writing assessments about employees, things like that. That’s easy to come across these kinds of things.’
However, there are some important caveats around privacy, says Westerman.
“The number one thing that people really need to understand is, don’t put anything company confidential or individually confidential into one of these engines because if you do, it’s out there for the world to see. Anytime you put anything into one of these generative engines, it becomes data that can be used elsewhere.”
When using AI as a recruiting tool to parse resumes of potential candidates, there is a potential for bias, says Freiman.
“You need to ask the right questions about how this algorithm was trained. Can you say whether the training dataset is representative or not representative? Is it localized or not localized? But even if it is accurate, we don’t want to perpetuate biases that already exists. It’s a whole world of AI ethics that goes into the different processes of HR from attracting the candidate to evaluating them to how they work, from paying benefits all the way to how the workers are monitored.”
Perhaps its best application is in summarizing text, says Westerman.
“For example, position descriptions, and figuring out what the competencies are in those position descriptions or looking at resumes and making sense of what kinds of skills you have in an organization, or even making sense of the latest HR regulations that are coming down, this idea of summarizing, it can be very good at that.”
Generative AI can also be used as a great benefits tool, according to Freiman.
“There are other ways to use chatbots, which are not ChatGPT or applications of closed systems that you can use for the employees, for example, so a benefit chatbot that you can ask about the benefits that you’re entitled to have. It’s easy to do because the answers are there.”
A university research study found that for those using the technology, productivity actually rose.
AI will become more universal
In the future, we can expect to see the technology embedded in a lot of the tools organizations are already using, says Westerman.
“Microsoft is already putting it into their search engines. The digital marketing groups are already putting that into their [tool] so I can expect that these tools will eventually be incorporated in all the tools that we already own for HR or for office productivity.”
“They’re far too useful to not be incorporated all over the place so we’ll start seeing them everywhere ad I think that will be a good thing, as long as we know how to use,” he says.
While there may be some excellent ways HR can employ generative AI, Westerman provided one cautionary tale.
“I asked ChatGPT to write a bio on me a couple of weeks ago, the bio it wrote had a book that I never wrote. I did it again earlier this week, and that book thing had been fixed but there was another error in there. You have to be careful with these things that they can be tremendously helpful but at the same time, they can also be wrong.”
For HR professionals, it’s a good idea to reach out before implementing such technology, says Freiman.
“It’s very important to assess the tools before you start working them and assess the tool in a professional way with the wider community of HR — colleagues and peers — and within the organization, from IT for example; from the legal department. It is always important to do risk assessment because things do go wrong, it happens all the time.”
