Lately I’ve noticed more and more frustration directed at generative AI and LLMs. People often mention issues like hallucinations, factual errors, and slop. I can relate to that as I’ve found myself thinking the same way many times. Yet, when you consider the technology and the sheer scale of how it has been implemented, it is nothing short of impressive.

And frankly, generative AI is very useful. It can speed up work, help with creativity, eliminate repetitive tasks, and facilitate research. But it is not what AI companies claim it to be. And that’s one of the issues.

What they are marketing is not just software. It is the idea of artificial general intelligence (AGI), machines that can reason, understand, and act across any field like humans do. We are nowhere near that point. What exists today is software that predicts patterns in data and produces a reasonable-looking result. The output is tuned to sound confident and human. It looks very polished but it is not to be confused with intelligence.

The business model behind the hype

Most AI companies are operating at a loss (it has been reported that OpenAI had a net loss of US$13.5 billion during the first half of 2025). Training and running large models is expensive and they depend on venture capital and complex transactions to cover the cost. This creates pressure to tell a bigger story than "this is a useful tool". Investors want massive growth, so the message becomes "this is the road to general intelligence and AI workers replacing professions x and y".

The reality is harder to square with those claims. If current AI were as capable as advertised, these companies probably would not need to keep hiring at the pace they are. The headcount growth shows how dependent they still are on human expertise, even as they market the idea of machines replacing it.

This pressure is reinforced by the surrounding ecosystem. Nvidia invests in OpenAI. OpenAI has signed a multibillion-dollar contract with Oracle. Oracle buys hardware from Nvidia. Money and influence circulate within a small group of companies, each benefiting from keeping expectations high.

In this environment, usefulness is not enough. The technology must be presented as truly revolutionary, even if that does not match reality.

Why it feels more powerful than it is

Generative AI often feels more capable than it really is. The answers are fluent and confident, which makes it easy to mistake style for substance. Many people now turn to ChatGPT instead of Google. The experience is a lot smoother, but not necessarily more accurate.

For someone outside a field, the results can look like real expertise. For someone inside, the inaccuracies and gaps are obvious. The risk is not that the tool is useless, but that its limitations are ignored because of how it is marketed.

A tool, not a solution

Generative AI does not replace expertise, research, or judgment. It is a tool. Used well, it can make us better. Used carelessly, it can cause serious harm.

The hype makes us lose sight of this. We do not need AI to be intelligence in order for it to be very useful. We need it to be what it already is: a practical tool that improves whatever it is we do when paired with skill.

So, where does that leave us?

AI companies will continue to make claims about intelligence to secure funding and keep growth stories alive. The rest of us should take a simpler view. Generative AI is very useful, but it is not intelligence. The value lies in understanding its limitations and using it well, not in comparing it to the promises around it. And if those promises are ever fulfilled, the impact will be undeniable.