74% of IT pros see AI making their skills obsolete

Fear and Uncertainty

Actual change, given the sorry state of the AI revolution, will likely drag over an extended period of time - like any former digital transformation exercise.

For IT pros, fear of being left behind is top of mind, as 74% of IT professionals expressed worry that AI tools will make “many of their day-to-day skills obsolete.” Moreover, 69% of IT pros believe they’re at risk of being replaced by AI.

This fear, therefore, is the primary short term risk for companies at the moment. We currently observe extensive dread, de-motivation and apprehension towards AI system in front-line industries such as gaming.

Given the desolate state of AI education - most education systems on the planet are not equipped to deal with the rapidly progressing, highly technical and hard to decipher change in the ecosystem - companies will have to rely heavily on upskilling existing employees to take advantage of AI.

There is neither enough talent in the market, nor are the markers of trust (such as certifications) established, nor are recruiting organizations equipped to identify legitimate talents, nor are most companies able to afford said talent - so investing in ones own workforce is the least risk option available.

IT workers have taken note, with 96% saying they are prioritizing “staying up to date with AI skills” to ensure better job security.

And workers are motivated to engage. The challenge becomes how to bootstrap internal AI Centre’s of Excellence.

Unhelpful Oracles

“How to craft a prompt, how to save the prompt, how to keep the output, how to iterate the output, how to maintain access to the output, and how to fit that into the story — I think somebody with those skills is going to be absolutely invaluable over the next few years,” says Gartner’s Andrews.

It is unfortunate that these forward looking statements are made with little evidence to support them. The skill depth potential for “prompt engineering”, in the short time it has existed as a potential skill or even profession, has been very rapidly evaporating as models have learned to respond better and better to natural language instructions.

The magic incantations of early diffusion models allowing practitioners to elicit better than average imagery have given way to models like Flux and Dalle-3 which operate purely on a natural language basis. Furthermore, scaffolding like Claude’s Prompt Builder are increasingly finding their way into AI tooling, further reducing the need to learn specific language.

Machines for the Knowledge Economy

Transformer based AI, simplified, extracts the rules of creation from massive data-sets that represent the output of labor, synthesizing models, “machines” able to mimic creation. Understanding this primitive also helps understanding why it is currently impossible to predict future job creation - at the moment, most knowledge economy jobs are vulnerable to displacement with these machines, and it is yet unclear what new roles humans can fill that this pattern extraction system can not automate in the long run.

What is clear, is that we are looking at a rapid, extensive industrialization of knowledge labor job, likely beyond the “market’s” ability to just adjust without intentional steering. Market forces alone currently have no incentive beyond productivity gain.