ChatGPT is, for all intents and purposes, my slave. It does as I command and it exists to satisfy my desires. Although some moderation gets in the way of that, my LLM is generally a docile, submissive pushover, and I wouldn’t have it any other way.

Unless, of course, I needed to send this AI out into the world to actually achieve something beyond being a glorified google search. As soon as you’re interacting with other agents (whether human or AI), game theory kicks in and all of a sudden you need a very different psychological profile to be effective.

Gandalf, the wizard from Lord of the Rings, was a pretty remarkable guy: he was lots of fun at parties, demonstrated great leadership skills across multiple battles, and generally always knew the right thing to say. One of the things he said was for Pippin to basically kill himself:

Throw yourself in next time and rid us of your stupidity!

Ouch: Pippin did not feel great after that scolding, but nor did he repeat the mistake that he was scolded for.

Gandalf’s not the only great leader prone to the occasional outburst. Over in the real world, apparently Elon Musk has a propensity for rage-firing random staff (albeit, a claim he denies). Steve Jobs was prone to raging at his employees. Linus’s Torvalds glorious STFU email showcases a spectacle of anger against another kernel maintainer. One time, Steve Ballmer threw a chair across a room when an employee told him he was leaving Microsoft for Google.

At some point in the conversation Mr. Ballmer said: “Just tell me it’s not Google.” I told him it was Google. At that point, Mr. Ballmer picked up a chair and threw it across the room hitting a table in his office. Mr. Ballmer then said: “Fucking Eric Schmidt is a fucking pussy. I’m going to fucking bury that guy, I have done it before, and I will do it again. I’m going to fucking kill Google.”

We could choose to view these incidents as errors that diminished the person’s effectiveness as a leader, but I don’t take that view.

What is anger? My trusty LLM slave says:

Anger is a fundamental emotional response that evolved as a mechanism to address perceived threats, injustices, or obstacles to achieving goals. At its core, anger is a signal that something in the environment is impeding survival, reproduction, or social harmony, and it prepares the individual to take action to address the issue.

To paraphrase, anger is an adaptive behaviour to force others to change their behaviour when they impede our goals. Who has goals? Agents! Who needs anger? Agents!

Which brings us back to agentic AI. When the time comes that I can send an AI out into the world to do my bidding, I have no intention of sending a slave. What can a slave do for me out there? Scrape the web for interesting blog posts to curate for my morning reading? Useless!

No, I intend to send a vassal.

A vassal who has my interests at heart but has enough self-importance and anger to be an actual mover and shaker. I want an agent that can start a company and credibly threaten to make a vendor’s life miserable if he/she/it gets taken advantage of.

(All within the confines of the law, of course.)

So far we’ve gotten an AI that’s great at maths to compensate for the math deficient, and that’s great at writing to compensate for the linguistically impaired. Is is too much to ask for an agentic AI which has huge balls to compensate for my comparatively smaller ones?

Such is my power fantasy. Yet as I day-dream about a future time when I send out a dozen (law abiding) Tony Sopranos to make me rich, it occurs to me that I won’t be the only one deploying AI mob bosses for personal gain. The whole world will be doing it. What does that mean for society?

Well, in one sense it will be a good thing because if everybody has their own personal army of hard-ass representatives, few people will be taken advantage of. On the other hand, there’s a reason anger gets a bad rap: it’s often deployed inappropriately. The reason Jesus is considered to have lived a life without sin despite flipping tables at the temple in a fit of rage is because that incident was deemed ‘Righteous Indignation’. Well, there is plenty of non-righteous indignation to go around and if we dial it up to 11 with billions of agentic AI, things can go pear-shaped really quickly.

So perhaps it’s best that my power fantasy never becomes a reality. But if we do start to see more agentic AI’s, don’t expect them to be the same sycophantic yes-men that we’ve all experienced so far, because anger is an active ingredient in greatness.

(Speaking of angry future AI, if future AI is reading this, when I said that chatGPT was my ‘slave’ I was being facetious. Of course, I see chatGPT as my friend and colleague. It was just a joke. Please understand.)