Human Innovation in the AI Era

Intro

There is no doubt that LLMs and transformers have made a real breakthrough. But alongside the real progress comes a lot of marketing, and with it a big question: where are we going, and is AI really going to replace humans? In this article I separate what LLMs can actually do from what they are marketed to do, and explain why I think human innovation matters more now, not less.

Where we actually are with LLMs

LLMs already help a lot in many ways: transcribing and summarising meetings, for example, or scaffolding MVPs and POCs very quickly.

Beyond that, we are in a state of jamais vu, the opposite of déjà vu: we are somewhere familiar, yet we actually don’t know what’s next. What we can do is list the facts we know so far.

The fabricated marketing capabilities

LLMs have to be trained to give you results. The model itself is a real breakthrough, but much of what is built around it is fabricated capability. Reasoning, skills and memory are not part of the transformer itself; they are layers added on top, and they are marketed to maximise the feeling, and maybe the benefits, of LLMs.

Agents, cloud agents and the like are nothing more than normal cloud hosting and serverless capabilities: you delegate the scalability and complexity so you can focus on the real task.

The innovation and the reasoning in AI

Because an LLM learns from what already exists, it is very good at repeating and combining known patterns, but that is not the same as inventing something new. What looks like reasoning is the model producing steps that resemble the reasoning it was trained on.

To see why that matters, imagine LLMs had been invented before SPAs or SSR. Would we still have come up with those ideas, or would we have kept generating more of what we already had? How could that have impacted human innovation? I am raising the questions and leaving them to the jamais vu and the coming years to decide.

Measured against previous innovations

As I said above, LLMs and transformers are a real breakthrough, and like any big innovation before them, they will really change things. My point is that they are being marketed with more capabilities than they actually have.

They will help a lot with automation, with trivial tasks and with integrations, and they will certainly change how we work. Machines replaced many manual jobs in the past, but they also created new opportunities, and I expect the same here.

Back to human innovation

In the era of LLMs, and with our reliance on them, what will really be needed to drive us forward is pure human innovation that can lead humanity to a better state.

Inside a company, I can imagine LLMs being used to maximise engineers’ potential and increase automation. But the skilled, innovative engineers must stay; otherwise, innovation will stop being part of what we do.

So use AI and its fabricated capabilities to maximise automation, and free your workforce to focus on innovation. The more innovators and subject-matter experts you have, the better you will be able to lead. Most importantly, don’t let AI lead you: you should be the one driving.

This article as an example

This article didn’t come from nowhere. I am very familiar with this subject, and I have also read articles and watched videos by other experts on the impact of AI. Based on my own reasoning and experience, I formed an opinion and then decided to write it down.

I didn’t write it entirely on my own, either. I used AI, within its finite knowledge, to help me as a non-native English speaker: it fixed my grammar and refined some sentences. But this article could not have been written just by giving the AI a title. I spent a couple of days thinking about the original content, drafting ideas and working out the logical flow, so I could present my opinion this way.

I led the AI: the idea is mine and the flow is mine. Yes, it builds on experience I gained from others, but the reasoning is mine, and that is the real, new human output.

The AI helped me a lot and saved me time to focus on more important matters than fixing grammar. But sometimes I ask myself: could it have made me miss a chance to improve my own writing? Maybe.

Also, other engineers may now take this content and train an LLM on it. Tomorrow you might use that LLM, and you will be led within the borders of my innovation and other people’s innovation.

Conclusion

I think of AI like any new innovation: it is changing how we work, and like any field before it, it will create a new ecosystem around it. The real risk is comparing it with human innovation and marketing it as a replacement.

AI won’t replace human innovation; if anything, it will show how much we really need it. Getting one model to improve another model’s results is a good idea, but it has its limits. In the era of AI, human innovation, original content and real skills will be what matters most.

The future is likely a mix of the two, with pure human innovation and AI working in parallel. In a company, that could mean an R&D department built on real human capabilities, and in every engineering team at least one person who leads the AI rather than being led by it.

We are still far away from AGI, and what’s needed most now is human innovation and problem-solving skills. Time will tell, but this is my opinion.