Hyperintelligent Entities

Shodan by Jim Hatama

How can you challenge a hyperintelligent entity?

In all our Dungeon Mastering lives comes a time when we throw our players into something crazy, something epic and beyond their grasp. It could be a friendly bout with a god or facing off against some kind of AI, but usually it involves the party, even if they are seasoned heroes, being in over their heads.

Writing a hyperintelligent entity is in of itself a challenge, writing it well, almost impossible.

We, as humans, can conceptualize a higher intelligence, but really comprehending it would be beyond our grasp. We deal with human-level intelligence in our every-day lives, and even the smartest of our friends rarely feel truly alien. They might be quirky, eccentric but we can still understand their motives relatively well on a general level. They will need to eat and sleep, go to the toilet, use devices that are built to be used by us, etc.

You get my point.

Writing an entity that is beyond human-level intelligence involves a lot of cheating on our part, and its portrayal needs to be handled with the utmost care lest the illusion breaks, and it is revealed that behind your entity sits a dork with glasses, pulling the strings Wizard Of Oz style.

Knowledge Isn’t Intelligence

People tend to confuse a large amount of information and memorization of facts to be a sign of intelligence, it’s unfortunately not.

In our day and age, with almost instantaneous access to every piece of information imaginable over the Internet, everyone of us would have the potential to be super-intelligent. We have done this for centuries, externalized knowledge and information from our brain onto other carriers, like books for example.

Our brains are but processing units for information that is mostly external to us. It has a relatively small buffer that can hold a small amount of knowledge or raw data. Mostly we only internalize very specific data that is repeatedly used by us day in day out, the reason for this is that the dynamic access to our memory is far faster than our hands can turn a page or type queries into Google (for now).

Without going too much into a philosophical debate, for the sake of the discussion and my point that I’m trying to make here, I will define intelligence as follows:

The potential amount of bandwidth and processing required to abstract solutions to problems as accurately as possible.

Knowledge and intelligence need each other. A high intellect without the necessary knowledge (data) will make mistakes because it can’t accurately assess the problem. Similarly, a low intellect with high amounts of knowledge will make mistakes because it can not process and abstract the solution, even if the data is sufficient.

Currently, with the Internet, we solved most of our knowledge-access problem, however we do not have the capability to process all this data as individuals.

But lets imagine there was an entity that could.

Believable Flaws

A hypothetical AI, with vastly superior intellect and access to all relevant data, would not see a human (or human level intelligence) as a threat. Similarly we do not see a lion or an elephant to be a threat, even though they are physiologically superior to us.

A hyperintelligent entity, with enough agency, could toy with the party by directing their actions in subtle ways. The only way to challenge a hyperintelligent entity is to create openings for the players that are believable. Depending on the entity, one can create believable scenarios where the party has an edge over the entity in some way.

The easiest thing to do is to give the entity human-based emotions or desires, but it’s a primitive and crude way of negating the problem. Why not simply make the entity human-level intelligence instead? It creates far less narrative dissonance than having an AI that goes into fits of rage that “clouds” its judgment.  Arrogance is tricky, but it can work if done well.

The worst thing you can conceivably do, is to make the entity appear incompetent.

With human-level intelligence, writing an incompetent antagonist is not as egregious, or at least can be reasonably ignored since they are “only human”. But presenting an incompetent godlike entity will clash a lot more with your portrayal. This includes not only their overall schemes but how they behave in combat. Remember, they are hyperintelligent, and therefore probably master strategists and tacticians.

Of course you are limited by your own intellect as a DM, so the key here is very in-depth preparation for the confrontation. You have to put effort into developing strategies that at least -seem- smart. This can be achieved by setting up traps, or having backup “deus ex machina” in case things go incredibly wrong for your entity. Don’t be afraid to try and exploit the game mechanics for your gain, it’s what the entity would do.

Hyperintelligent entities are also acutely aware of their weaknesses and will act accordingly to prevent players to exploit them.

There is an excellent video by Noah Antwiler (aka The Spoony One) on how to properly portray dragons and their abilities in D&D, if you want a specific example.


For the most part, if we move past human-level intelligence and grant the entity sufficient knowledge, the plans, reasoning and decisions of the entity will seem largely strange and alien.

Depending on the entity this might come from physiological differences, like not needing to sleep and eat, or its desires would be completely foreign to us and its reasoning to achieve the goal largely not understandable.

The trick is to never disclose too much about the inner workings of the entity, if you do, you have fallen into a trap by trying to explain a hyperintelligent entity with human-level intelligence.

Recent examples of this problem in a game was Mass Effect 3, in which the antagonist entity, The Reapers, a race of hyperintelligent AI, performs cyclically a genocide on the galaxy. The mystery we encountered in the original Mass Effect, concerning the motivation of The Reapers, was very cthulhian. It spoke of a motivation so far removed from human thinking, so alien, that we as humans would not understand it. It spoke as a god.

In Mass Effect 3 the “plan” and motivation of The Reapers is revealed and it turns out to be banal, morally bankrupt, and not at all worthy of a hyperintelligent AI. It presented us a typically -human- reasoning, because it was written by a human, it diminished the status of the entity. Was the player given the choice, he would have been easily able to argue against the plan and win.

From cthulhian monstrosity with unfathomable intellect to complete retards in one game. (c)Badspot@deviantart


It could be argued that hyperintelligent entities, their reasoning being so far removed from what we as humans understand, could be portrayed as insane.

We have never encountered another intelligence equal or above to our own, and it is hard to portray how these intelligences might think or act. However a truly alien intellect, especially in fantasy or science fiction, not being bound by the general laws of our universe, might come into existence “ex nihilo” without a proper basis in an evolutionary process that might shape their intellect.

Take for example AIs, they could conceivably be said to be created instantly, without the need for a lengthy evolutionary process of a slow, cumulative gain in consciousness and societal interaction. They are thrust into “our” universe and our societal structures without any kind of previous primer. Sure, they can learn to understand our system, our intellect, but their own intellect might be entirely alien and unfathomable.

Extrapolating reasons for actions the entity takes should be kept to a bare minimum, or sometimes even left as completely illogical, with a hint of reason that sometimes shines through. Writing insane entities is almost as hard as writing for hyperintelligent ones, writing both, is a task not taken lightly.

The trick is to establish both, the alien nature of the intellect, bordering on insanity, as well as reasons for minute details of the behavior. Madness or insanity does not necessarily have to do with murderous or dangerous intent, it might just be eccentric behavior, reclusiveness or other traits associated with a form of delusion.

In William Gibsons “Count Zero” we encounter an AI from the previous book, now different, reclusive, building boxes of unknown purpose with seemingly random objects inside. We do not understand its actions, but the author hints at a greater plan, a purpose for the seemingly random behavior. We never completely learn what the plan is/was, nor do we fully understand the motivation of the other Loa (other AI entities), but we do not have to, we understand and feel they operate on a different level than the reader or the humans in the story do.

Point Summary

Let us review:

  1. Write competent entities
  2. Take time to understand the entity and the consequences of its existence
  3. Never fully explain your actions (no exposition)
  4. Utilize the abilities of the entity to its fullest. You have time to prepare, the players do not.
  5. Do not shy away from bending the rules, but do not break them (players hate cheating DMs)
  6. Understand the difference between knowledge and intellect
  7. Never let the players “out-smart” your entity in a direct and equal confrontation
  8. Be aware of your entities weaknesses
  9. Create believable openings and flaws for players to exploit

In general those are good guidelines for the creation of any good antagonist, but with hyperintelligent entities they are doubly important because everything hinges on the portrayal of their behavior and their decisions.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: