How have you been forming ideas lately?

AI has obviously changed how we think about, consume and interact with information, but what action can we take to maintain our human edge?

Described by machine learning founding father and recent Nobel Prize winner, Geoffrey Hinton, as "a not-very-good expert on almost everything”, LLMs essentially draw on massive amounts of information and attempt to formulate the most correct answers to questions.

But what do humans do well that AI struggles with? Well, while AI is great at remixing what it already knows, humans excel at inventing what's never been known.

Anchor and leap ideas.

In AI terms, ideas can be thought of in two categories:?

“Anchor ideas” are the ones AI knows particularly well. They're the well-worn paths of accepted information within the vast wilderness of human knowledge.

“Leap ideas” on the other hand are those that require creativity and invention. For example, try asking an AI to imagine a new form of government that’s never existed before.

Where humans still excel and AI struggles, at least for now, is in the realm of leap ideas – connecting seemingly unrelated concepts and imagining things that have never existed before.

"AI can mirror humanity's collective thoughts, but true novelty is when we step beyond what's familiar." – Michael Dempsey

AI can’t replace human thinking.?

What sets humans apart from AI is our ability to leap into the unknown, and breakthroughs often happen at the intersection of ideas.?

In 2024, a team at MIT created a new material by combining principles from spider silk and quantum physics.

It’s highly unlikely that AI could have achieved such a feat because it required connecting dots from wildly different fields, with vast amounts of experimentation and extrapolation of ideas.

It’s a similar case with the recent revolution in brain-computer interfaces. It wasn't neuroscientists or computer engineers alone who cracked it. It was a chef who applied food texture principles to neural implants.?

Now, we have BCIs that feel "natural" to the brain.

Similarly, in a recent challenge, OperAI’s “O1” could optimise existing rocket designs but couldn't conceptualise a fundamentally new propulsion method.?

This is still the domain of the human brain.

Is there a novelty barometer for AI?

In 2024, GPT-4.5 processed over 1.8 trillion parameters. That's basically like having a brainstorming session with the collective knowledge of humanity.?

But here's where it gets interesting.?

According to reports like the 2024 AI Index from Stanford HAI, AI is being used increasingly in the generation of novel ideas, such as imagining new ideas for patents. How this works is by analysing existing patents, identifying gaps, and suggesting new ideas that might be patentable.

So this isn't about AI replacing human creativity. It's actually about AI helping to push human creativity further.?

Here’s another example.

New York-based AI company, Runway, applied its models to test the novelty of its own existing product concepts. This led to insights that ultimately resulted in a pivot from "video generation" to "a new camera for the AI age" and a valuation increase to $1.5 billion in 2023.

It’s not that we are less original than we think.

The real value isn't in having an idea that AI is incapable of generating.?

It's in using AI to rapidly iterate through the obvious, so that humans can focus on the truly groundbreaking.?

This is where human creativity still remains supreme.

Why anchor ideas matter Don't write off anchor ideas just yet, either.?

According to the 2024 World Intellectual Property Report (WIPO) , a substantial proportion of recent patents build on existing technological frameworks rather than introducing entirely new concepts .

Google Glass flopped in 2015. Recently however, AI-powered analysis of consumer data led to its rebranding as "NeuroLens." Same core idea, but with a twist.?

It's now projected to hit ￡5 billion in sales by 2026.

In this case, AI had a different advantage: a lack of human bias. AI doesn't remember embarrassing product launches or market crashes the same way that humans do.

It also never forgets.

A recent study showed that AI models helped identify 150 "forgotten" pharmaceutical compounds that human researchers had overlooked. Three of these are now in clinical trials.

Meanwhile, at Moderna, they used AI to revisit discarded mRNA therapies. This helped a human scientist to spot a connection that led to the company’s groundbreaking new Alzheimer's treatment.

So the real magic happens when human intuition and AI-powered analysis meet.?

What’s more, as AI gets smarter, human creativity becomes more valuable, not less .?

Startups are catching on.?

Y Combinator's 2024 batch showed a 70% increase in companies focusing on "AI-resistant" problem spaces.

And companies with "leap ideas" in their mission statements outperformed their sectors by 3.7x in 2024.?

The future belongs to those who can balance between anchor ideas and leap ideas.

It's not man vs. machine.

It's man with machine vs. the unknown.

That's where true innovation lies.