There’s a lot of Generative AI optimism going around ever since OpenAI launched ChatGPT. “We’re going to solve world problems”, “AI will advance all research exponentially”, “we’re building super intelligence”, … Estimates for when these transformative breakthroughs would arrive vary from ‘any day now’ to ‘in a few decades’. The developers of the Large Language Models (OpenAI, Anthropic, …) are claiming the more optimistic timelines and future capabilities. And I have a suspicion that that is more marketing hype than anything else.
And I’m not alone in this view. Microsofts CEO thinks we’re in a “Dot-com-like bubble” and that AI is generating almost no value. As powerful as generative AI tools are for personal productivity (if you use it right), it’s hard to find business problems to solve with it. And when finding one (like helpdesk chatbots) some companies manage to do such a bad job of it that I actually wish I could go back to the phone menu and waiting for an operator to be free. I’ll talk about my frustrations with the Proximus chatbot and phone assistant another time.
Disruptive
The vision being sold by Generative AI companies is that these AI systems will become so capable that they will be able to replace most humans in the workplace. Companies could run entirely on AI Agents (Large Language Models running in an automated loop and interacting with computer systems or each other to finish more complex tasks) and they will be more productive and we’ll have infinite growth … and we’ll all benefit from it. Well, okay, it sure sounds disruptive. But does it actually make sense? Let’s break these utopian ideas down.
The chance that ‘all humanity’ will benefit from the elimination of human labor is very slim. I’m sure that the technology companies will benefit from it. Company owners will benefit from it – for a while. But will the people that used to be employees benefit? As a society we’re constantly disproving this point. From the gig-economy to the richest people gathering more and more wealth to the more entrepreneurial people complaining about “people living off their hard earned income through taxes”. I think a more realistic picture is that such AI-run companies will siphon even more wealth to an ever smaller group of super-rich people who don’t want to share. How such a societal system is meant to be sustainable is unclear to me. Who will buy what all these companies are making if there are no consumers?
Another big blocker I see for the current Generative AI tools to become so capable that they can run a company unsupervised (or minimally supervised) is that they are not creative… not really. Sure I use them too when I have writers block and staring at a blank page. They’re ideal to summarize a text for me to read. Or any of the other ways in which a human can use them as a productivity tool. But they are only creative if their output is curated by a person. When we humans are creative we develop all kinds of ideas in our head, often very wild. The true creative genius can curate their thoughts and ideas and only let those filter through that are work in reality, are truly novel and/or pleasing to other humans. Current AI systems have no such built-in system to separate the brilliant from the stupid. And I doubt that just making the models bigger and stronger will “automagically” give rise to such a capacity.
And then there are the physical limits. Training and running the current AI systems is already a strain on water, energy and hardware availability. Using these models constantly to replace humans in the workplace will strain the limited resources we have even further.
Solutions
Some of these problems can and will be solved. One proposed solution I saw involved quantum computers and nuclear fusion to overcome the current limitations. I sincerely doubt that this is what we should wait for. Nuclear fusion is progressing very slowly and a breakthrough will probably come on a complete different timescale that the limitations in Generative AI that we need to overcome. And quantum computing… there are voices claiming that AI will solve the problems for which we previously thought we’d need quantum computers. I’m not counting on either of these to solve Generative AI’s issues any time soon.
When making predictions about the future, we have a tendency to extrapolate from the current trends. We often don’t consider that the future will bring paradigm shifts. Obviously you can’t predict or plan for a paradigm shift but still, that is where the solutions lie. DeepSeek has introduced such a paradigm shift and their models can do the same work at a reduced energy and hardware cost.
I also believe that the current Large Language Models are only a part of what you’d need to build AGI (Artificial General Intelligence) or even super-intelligence. You’re going to need a few more AI models of different types to complement each other. The first things that come to mind are AI systems that: can fact-check; do a reality check (ask if the current thought will work in the real world); work with tasks, goals and priorities; can develop and fine-tune a theory of mind; …
When looking at business applications, the technology is actually moving too fast. Businesses need more time to identify those processes that that would gain from being automated with generative AI. Product teams need more time to find worthwhile applications for Generative AI in their software products. So we need more time to learn how to use these tools.
Other AI
I’ve very specifically mentioned “Generative AI” throughout this text. That’s the AI sub-field that is moving so fast at the moment. But there’s more to AI than ChatGPT. For a long time now, we’ve been using AI systems for analytics and predictions; for industrial control and navigation; for image identification and classification; … These AI sub-fields have all proven their worth for years now and will continue to do so.
AI is useful and brings business value. We’ll find use cases for the new Generative AI tools as well – on top of the few we already have. Just don’t get stuck in the current hype. You don’t HAVE to use (Generative) AI but if you’re struggling with one of your business processes, consider AI as one of the possible solutions.
AI Mindset 
Leave a comment