← to blog

What is left for a human?

There’s a panic: what’ll be left for a human in the LLM-ish world? I assume it is an existential question for some (me!), and those who feel threatened are neither the first nor the last.

Indeed, an agent can do a lot of what previously was mainly ‘handmade’: code, writing, proposing solutions, search, even solving our homework.

At my current job, we’ve automated a lot of work using just Claude’s skills capability. The skill goes to Jira or RAG to gather context, understand the as-is state, formulate a solution. For hypotheses we have built a skill that can create SQL and calculate metrics based on ClickHouse’s data.

And it may seem that nothing’s left for us humans!

But I can’t agree.

I have spoken about it with several people and here’s the list of tasks that may still be left for us.

  1. Logic and reasoning
    • Logic and reasoning is already done in AI. It can not only use proper formal logic approach, but also — with appropriate instructions — can use problem solving frameworks.
    • Of course there are nuances to how it is solved. LLM can hallucinate or be based on wrong data (so that output is not logical).
  2. Using problem solving frameworks1
  3. Communication and selling
    • Communication and selling is partially solved. AI can call a cold lead, rewrite your landing, you can even create a machine that would post to socials for you. It can use Cialdini’s persuasion framework.
    • But communication is not a text-only activity. While AI can recreate some prosodic and textual characteristics (like tone, tempo, stress, specific word selection) it can’t use gestures, smiles, or other ‘physical’ ways to communicate yet.

So we can see, each task is solvable and even if it is not solved, it may be solved in the future. Let’s look at it through a different lens: what roles will stay with humans?

I think agency2 will stay with humans, while acting can be delegated to an AI system. An agent (ha!) can act, but a human will be the source of action, of finding what can be done, adjusted, redone in a better way, or even invented.

Right now, many of us confuse agency with action. For example, I heard colleagues say “Claude did it, not me”, which is not true, if you have this principle in mind, since Claude did it only because you asked.3

But if you really think that Claude did it, then time’s against you.

  1. Was: Creative Problem Solving. I mean using problem solving frameworks here (TRIZ, SIT, etc.). However, I assume creativity may be solved by AI. ↩

  2. Ability to act according to free will; to be a source of action. Some religions say (Genesis 2:5) that a human is created to make the world better. I think agency allows to see how the world can be improved. Similar to Logos. ↩

  3. This same notion was discussed decades ago in Hannah Arendt’s Eichmann in Jerusalem: AI will find a way to complete a task without asking why, but the result is yours. ↩

— Aliaksei