[Tech Thoughts] Don’t forget: Working with AI means a need for human oversight

This week, South Africa withdrew its draft national artificial intelligence policy, owing to a glaring issue — the reference list of the policy featured some sources fabricated by AI.

This phenomenon of fabricated sources showing up in otherwise normal documentation is currently a bit of a problem, as there’s been enough hallucinated AI sourcing that a database exists of legal proceedings derailed by AI hallucinations.

While some Tech Thoughts readers may find it funny that people who are uncritically using generative AI are getting a comeuppance, it must be emphasized that, barring an apocalyptic AI-killing event occurs, we will all have to deal with and perhaps work alongside some form of AI automation sooner rather than later and regardless of our feelings on the existence of AI in whatever form it takes.

This means people need to train themselves to both work alongside and distrust AI outputs at the same time.

Why does it matter?

The International Labour Organization (ILO) estimated back in February in a report on Generative AI and the Philippine labor market’s exposure to it that some 12.7 million workers in the Philippines (or a quarter in employment) are exposed to generative AI.

While not exactly serving as job replacements, AI is serving to automate certain tasks in the jobs people already have.

The ILO said only 3.6% of jobs fall into the highest GenAI exposure category with the elevated risk of job displacement. “Rather than outright automation, the most significant impact of GenAI on the Philippine labour market is likely to be the transformation of jobs, potential gains in productivity and enhanced employment quality,” it added.

Moreover, the ILO says this exposure to GenAI in the workplace isn’t gender neutral, but rather impacts the jobs of women more than men. “Women face double the rate of GenAI exposure compared to men, reflecting their greater concentration in high-exposure occupations. Women with an advanced education face particularly high potential of GenAI disruption,” the ILO went on to say.

What the ILO is pointing out here is that if AI is going to be inescapable, then we had best learn to work alongside it — though we might best be served by treating it as a clock in need of repair that won’t always be correct, so we have to double-check its outputs.

Burnout and ‘brain fry’

Labor in an AI-addled world may mean doing the work of two people, or at least, the cognitive equivalent of trying to do something while being asked to check someone else’s work, repeatedly, till the job is done.

Specifically, you’ll need to do your own work, and when using AI, you’ll need to check its outputs to make sure they’re actually usable for your current task.

As researchers in a study published by Harvard Business Review put it when they studied how AI was affecting work: “Contrary to the promise of having more time to focus on meaningful work, juggling and multitasking can become the definitive features of working with AI.”
This “brain fry,” or the cognitive exhaustion from intensive oversight of AI agents (basically, checking the work of AI) is a real thing. The researchers defined it as the “mental fatigue from excessive use or oversight of AI tools beyond one’s cognitive capacity.”

This AI-associated mental strain carries significant costs in the form of increased employee errors, decision fatigue, and an intention to quit.

At the same time, using AI to handle the repetitive tasks of certain jobs can be said to reduce burnout, or the emotional strain of doing work. The researchers posited that offloading the toil of doing repetitive tasks to AI lets workers focus on doing more creative work — stuff they actually want to do. By offloading the boring stuff, it seems, some employees “reported higher work engagement and motivation scores; more positive emotional associations with AI; and fewer negative emotional associations with AI than others.”

Burnout and brain fry are two sides of a coin of working. Whereas burnout is an emotionally-driven tiredness or fatigue, AI brain fry is mentally-driven cognitive strain.

The ‘future’ of work

Without going too far into specifics, I am currently working with someone on trying to use AI to make some aspects of work easier. I admit, however, that I am dreading the potential brain fry of checking a ballpark figure of 100 AI outputs.

It will involve fact-checking AI outputs for accuracy, while also doing the creative tasks of fixing style issues so that the output doesn’t read like a robotic printout.

The creative stuff I can do, but the task of checking an AI’s work does not sound fun at all. It is a necessary task however. If left alone, AI is bound to get something wrong, and that hurts me and Rappler’s integrity as a whole.

As South African Minister of Communications and Digital Technologies Solly Malatsi said regarding the AI policy blunder they created: “This failure is not a mere technical issue but has compromised the integrity and credibility of the draft policy.”

If you work with AI, it means the responsibility of making sure AI doesn’t do stupid things will unenviably be on the shoulders of the human watching it. So stay eagle-eyed and sharp, and remember, welcome to the “future” of work. – Rappler.com