What Star Trek (still) Has to Teach Us about AI, Language, and Being Human

My communication style is visual storytelling, so strap in!

No one is surprised when I say I want to talk about Star Trek and AI… But there are so many lessons to learn in Star Trek episodes! Today’s lesson surrounds how language is the power we need to wield carefully in this scientific reality.

In case you’re not as well versed as this trekie… You just need to know three things for today:

First - Data is a one-of-a-kind*[1] sentient android capable of incredible intelligence, association, analysis and mental capacity, who cannot speak in contractions.
Second - the computer on the Enterprise is an AI who accepts commands verbally, and can essentially perform anything which doesn’t require a physical action. This version of artificial intelligence doesn’t often breach into sentience, until the holodeck episodes, but how you execute commands is often a point.
Third - A holodeck is a room full of photonic projections and faulty safety features, and often became what I feel are the ‘filler’ episodes or to provide a bit of variety from the starship. These are my favorite episodes and later become the projection of all kinds of human flaw oriented jokes in series like Deep Space Nine, Lower Decks, and spin offs like The Orville. That part you didn’t necessarily need; consider it a tip ;)

Back to my point. Language is SO important when it comes to communicating intent, especially to an artificial intelligence.

The ideal example is in Season 2, Episode 3, Elementary, Dear Data of Next Generation, Geordi makes an attempt to provide Data with a challenging Sherlock Holmes holodeck program with very specific instructions to the holodeck computer:

“Create a challenge enough to beat Data.”

This bit of detail backfires when the Professor Moriarty character accidentally becomes self-aware, and literally is challenging enough to beat the android which the computer knew and understood as the objective for the request.

This isn’t the only time an instruction was taken too literally or to a fault, or otherwise caused an other than intended outcome. In another episode Geordi asked the computer to "show me" something and accidentally conjures a holographic person - then gives her personality and falls in love. But in the Holmes' episode, the instruction nearly destroyed the Enterprise and the entire crew because of the “beat the android” words.

A recent article of AI mishap with language manipulation is only hilarious if you don’t own the dealership whose AI bot sold a car for $1.00. Even early demonstrations of my own journey with Midjourney artwork and its ability to understand my intent and render limbs and phalanges are as hilarious as they are terrifying (which was mostly the intent):

“A computer screen with a figure crawling out from the ether”

Figure 1 Figure Note: Unedited images rendered by Midjourney 5.2 based off of simple prompt.

Figure 2 Figure Note: Same prompt in Version 6.

So - what does this have to do with today’s AI? Well - LANGUAGE continues to be important, because no matter the intent a literal interpretation is the current expectation of engaging with young, as well as mature AI.

How you describe what you need, the order in which you describe it, and the emphasis you place on details will guide the overall outcome. The more specific you can get, the better your results will likely be. The better your results are, the better the AI learns to do what your vision actually is in the future.

The cycle improves over time.

Now - I’m going to take this application one step further and introduce you to another favorite series of mine, Amelia Bedilia.

“From dressing the chicken to drawing the drapes, Amelia Bedelia does exactly what Mr. and Mrs. Rogers tell her to do. If things get a bit mixed up, well, that's okay.”[2]

Drawing the Drapes meant something different to Amelia...

She basically Poppins’ things and all is well in the end. Lesson: Words are important.

When things have multiple meanings or interpretations, whether intentional, cultural, or otherwise, people can and do miss hidden or even implied meaning. For this reason we must come to value and achieve mutual understanding that clear expectations are stated, and understood. This helps those who might not come to the same conclusion. Ever met someone who couldn’t ‘read between the lines’? (Spoiler - It’s me, hi.)

I have never been known to shy away from asking questions or speaking up in most settings when I need to. But I have known others who do not hold the same bold approach and leave questions unasked, and as a result do not hold the same vision.

Words are important whether you’re talking to an artificial or even just an alternatively intelligent (AI) recipient. Exchanges should be clear, simple, and structured based on quality and value.

Like Midjourney, short or simple exchanges will rely heavily on default styles and understanding by the receiver. If someone requires more context to understand, it may be necessary to provide more descriptive communications. However, more isn’t always better, because the value of specific words may not be weighted the same by the recipient. Communicate the main concepts, with enough context to understand.

If nothing else, always verify and provide opportunity for questions or clarity. The AI bots might not be able to accommodate this, but us humans can.

I hope you enjoy the creativity and progress you might find in clearly communicating with an AI.

Thanks for reading, friends!

Resource Notes:

*[1] Data actually is not one-of-a-kind but two. He has a Brother, Lore, who is another plot line. Potentially one-of-three if you consider his daughter Lal from season three.

[2] https://ameliabedeliabooks.com/books/9780064441551/amelia-bedelia - funny and after all these years, still hold value.

Page updated

Google Sites

Report abuse