Just over two years ago, the explosion of Large Language Models (LLMs) among the general public shook up our vision of the future of work. Today, with the benefit of hindsight, it’s time to take stock of their real impact and their potential to “replace” humans.
The initial shock
I still remember my first interaction with ChatGPT. Like many people, I was flabbergasted by the tool’s capabilities. The question immediately arose: what’s left for humans? Despite some major and obvious shortcomings – 2 years on, ChatGPT is still unable to count the number of letters in a word – the machine seemed capable of coding, writing and analyzing, sometimes better and faster than we could.
Today’s reality: impressive but nuanced
Two years on, these tools are still objectively impressive, but we can see their limitations more clearly. They excel at doing 80% of the work on many tasks, but the remaining 20% still requires human expertise.
The chainsaw metaphor
To understand the impact of LLMs, I find the analogy with the chainsaw apt. Its invention didn’t make lumberjacks disappear, but it profoundly transformed their profession. Today, a lumberjack who insists on using only an axe would simply no longer be competitive.
This analogy also helps us to understand the coexistence of different levels of use: just as there are professional lumberjacks and occasional chainsaw users, we are seeing the emergence of expert LLM users and Sunday users.
Very uneven adoption
In the technical teams
Within the technical teams, particularly the developers, I see a clear divide :
- On the one hand, developers who have integrated LLMs into their development environment and use them on a daily basis.
- On the other, those who use them only occasionally, or not at all
This difference in adoption is already translating into significant productivity gaps.
Personally, I save a lot of time when structuring Python scripts or writing SQL queries of varying complexity.
In the trades
Distribution is even more contrasted on the business side:
- Some employees have integrated these tools as naturally as Google into their daily lives.
- Others seem totally unaware of their existence.
- Between the two, the vast majority are still hesitant and uncertain about how to use them.
The question of training
Faced with this uneven uptake, companies’ instinctive response is often to train their staff. However, I have mixed feelings about this approach.
Take Google, for example: we didn’t need any training to learn how to use it. It was adopted naturally, through practice and word-of-mouth. I think the same will be true for LLMs. In some specific cases, knowing the right way to write a prompt can be an advantage, but in reality, that’s not where the challenge lies.
On the other hand, there is a real training issue, but it lies upstream: the ability to structure one’s thinking and clearly formulate one’s needs. Without this fundamental skill, even the best LLM, combined with the best prompt, won’t be able to help effectively.
But is it really the role of the company, and more specifically the IT department, to provide this fundamental training? It’s a question worth considering. Even if LLMs are considered “technical tools”, their effective use calls for skills that go far beyond the technological framework:
- The ability to analyze and synthesize in order to properly formulate needs
- Critical thinking to assess the relevance of answers
- Mastery of language and clarity of expression
- The ability to structure thought logically
- A general knowledge enabling answers to be contextualized
- Writing skills to reformulate and adapt outputs
These skills are more a matter of basic education than professional training. The IT Department can certainly provide the appropriate tools, guarantee their secure use, share best technical practices, facilitate exchanges between users, or suggest relevant use cases…
But it would be a strategic mistake to try to compensate for shortcomings in these fundamental skills by providing technical training. It would be like trying to train someone in the advanced use of a word processor when they have difficulties with grammar and syntax.
LLMs won’t replace us in the immediate future, but they will profoundly transform the way we work. Like the chainsaw for the lumberjack, they will become indispensable tools for staying competitive.
So the real question isn’t whether they’ll replace us, but rather how we’ll adapt to working with them. Because while LLMs may not replace workers, they will certainly replace those who don’t know how to use them.
The challenge for companies is not so much to train people to use these tools, but to create an environment that encourages their natural adoption, while developing the fundamental skills that will enable them to use them effectively.
Leave a Reply