We became smarter by talking to ourselves...


… and maybe this could be the case with AI too

In 2022, ChatGPT and other large language models (LLMs) made significant strides, demonstrating a surprising connection between language fluency and the ability to reason. Let’s examine this with an example:

The above answer demonstrates a form of reasoning. While not found verbatim on the internet, responses of this nature rely on previously established knowledge and understanding of logical fallacies.The model identified the logical inconsistency of dying from ‘old age issues’ at 24.

Incredibly, anthropologists might not be entirely shocked by this language breakthrough. History suggests a fascinating connection: within our own evolution, the Homo lineage saw a similar surge in both language and cognitive abilities. What’s more, studies even show how certain speech disabilities can impact our overall cognitive development. This points to an intriguing dance between language and the way we think!

Language has a magical power: it unlocks the doors of our imagination and lets us hold entire conversations within our own minds*!

Picture this:

“Grocery store now, or risk being late for work? But the afternoon might be better… or will I miss the gym? Ugh, maybe I can survive one day without a restock…”

That endless internal chatter? That’s language at work – not just for talking to others, but for shaping our very thoughts. Try to imagine thinking without a single word!

Now imagine a world where AI doesn’t just act, but thinks out loud!

Researchers reported last month at the Neural Information Processing Systems (NeurIPS) conference that are tapping into the power of our own internal monologues to help AI learn and solve problems!

Like us, they’re able to break down complex tasks into smaller steps, verbalizing their intent: “Okay, first find the key, then unlock the door…”

This isn’t just about efficiency; this kind of language-based thinking grants AI incredible flexibility. They can adapt to unexpected obstacles, come up with creative solutions, and even potentially learn by watching instructional videos – just like we do! The possibilities are exciting, hinting at a future where AI might think through problems as seamlessly as we navigate our own daily lives.

This glimpse into AI’s potential “inner voice” raises fascinating questions. Could a machine that thinks in words become more like us than we ever imagined? As AI continues to evolve, the lines between human and artificial reasoning may blur. Perhaps one day, we won’t just be observing how AI thinks, but engaging in genuine conversations with minds shaped, in part, by the very essence of our own.

*The fact that our internal monologues have no ‘mute’ button, can lead to a cascade of challenges! More on this in a future article 😉