There’s a fundamental obstacle to working with AI.
If you tried using ChatGPT you might have stumbled on it.
Chat GPT is what is called a LLM or Large Language Model. It’s a specialized subset of AI that deals specifically with language understanding and generation.
And as with any language, you have to know how to use it.
You wouldn’t want to speak Italian to a Chinese person would you?
For the same reason we need to understand how to speak to Chat GPT and other LLMs.
Speak to them as you’d speak to human beings.
It’s still early days so a lot of people think in terms of dealing with machines.
But what if you reframed it?
In fact there are a lot of commonalities between LLMs and humans…
Here’s some interesting takes from a recent research paper that might change your perspective on how to deal with AI:
- Unpredictable learning: Just like humans, LLMs learn and adapt in ways that can’t be anticipated. Their behaviors emerge as a byproduct of increasing investment, making them just as unpredictable and intriguing as their human counterparts.
- World representations: LLMs, like humans, develop their own understanding and representations of the world. This uncanny ability enables them to generate human-like responses.
- Steering challenges: Ever tried changing someone’s mind? It’s not an easy task, right? Similarly, there are no reliable techniques for steering the behavior of LLMs. They’re just as stubborn and opinionated as any human you’ve met.
- Black box minds: When it comes to understanding the inner workings of both LLMs and humans, experts often hit a wall. The thought processes and mental models that drive their behavior remain a mystery, adding to the complexity of working with them.
- Surpassing human performance: Don’t underestimate LLMs! They have the potential to outperform humans on certain tasks, just like some individuals stand out in their expertise.
- Independent values: LLMs, like humans, don’t need to express the values of their creators or the values encoded in web text. They develop their own perspectives.
These quirks, these nuances, if you understand how to use them, can give you an unfair advantage when dealing with AIs.
So, what does this mean for you? When working with LLMs, treat them as you would humans. Embrace their unpredictability, understand their limitations, and appreciate their unique perspectives.
That’s how you make the most of them and have them work for you.
This is especially true when it comes to research work.
As I started using ChatGPT for my analyses and copywriting research, I’m more and more convinced that the future of customer research lies at the exciting intersection of human intuition and AI intelligence.
It’s like tapping into a reservoir of data on humans with laser-like precision, rather than performing clunky and inefficient Google searches.
Am I getting too nerdy? What would you love to learn more about?
If you find any value in these emails I’d be super happy if you wanted to share it with a friend.
Quote and reflection of the day:
“Samurai training involved the following kind of psychological exercise. Whenever, the teaching went, you feel yourself frightened or despairing, don’t say to yourself, “I am afraid,” or “I am despairing.” Say, “There is someone who is afraid,” or “There is someone who is despairing. Now, what can he do about this?””
- Robert Moore, King, Warrior, Magician, Lover
Seeing yourself from the outside helps create space for better decision making. You don’t make it about you anymore, but at the same time you commit to assuming responsibility for it.