Why is Generative AI Magic?

• Generative AI is real magic: machines are creating beyond their programming, and experts don’t fully understand it.

• The article breaks down three key breakthroughs that unleashed today's AI explosion.

• Leaders who ignore generative AI now risk being left behind by 2025.

Let's take a simplified walkthrough of the "generative AI" surge so you can hold the high ground at your next drinks party.

Here's a simple visual map of the key advances:

What's the big deal?

You've probably noticed more hype around this AI craze compared to previous tech fads—even compared to the last AI wave. The excitement is justified because generative AI is truly magical.

Let's quickly explore three key advances that sparked modern "generative" AI.

AI has been around since the '50s

Modern AI, synonymous with "machine learning," dates back to the 1950s when IBM engineer Arthur Samuel coined the term. After early promise, AI experienced a lengthy quiet period known as the "AI winter," limited by insufficient computing power and data availability. More background on this era can be found in this article.

Jump #1: Deep Learning (2015)

By the 2000s, advancements in computing and data sets allowed significant progress in machine learning, initially used behind the scenes for speech recognition, recommendations, ad targeting, and spam filtering.

However, early machine learning wasn't truly "learning"—it relied on scientists creating models based on complex maths for computers to make predictions.

This changed with deep learning, where computers could independently identify patterns and "learn" from data.

In 2015, major cloud vendors made deep learning widely accessible, fueling consumer products such as voice assistants (Alexa, Siri, Google), translation services, self-driving cars, and image recognition apps.

Yet, the initial hype waned when businesses realised deep learning required significant human input to label data—a method known as "supervised learning." For instance, a fraud detection system my team built required manual review of thousands of advertisements to train the AI, ensuring accuracy but demanding substantial human effort.

Jump #2: Self-Supervised Learning (2018)

In 2018, breakthroughs allowed computers to generate their own data labels, dramatically reducing human effort. Tech giants like Google, Facebook, and OpenAI quietly integrated this "self-supervised learning" into existing products.

However, widespread adoption remained limited due to prohibitively high computational costs.

Jump #3: Generative AI—the Magic (2022)

As researchers advanced self-supervised learning, they made a remarkable discovery: AI models started exhibiting behaviours beyond their explicit programming.

In 2019, OpenAI found their models unexpectedly generating original content. This unplanned phenomenon marked the birth of generative AI.

By 2022, Anthropic's research revealed these capabilities emerged suddenly at critical points of scale, rather than gradually. These emergent behaviours include:

• Few-shot learning: Learning new tasks from just a few examples.

• Basic reasoning: Solving logic problems and performing abstract thinking without explicit training.

• Compositional generalisation: Combining unrelated concepts into new, creative ideas.

In late 2022, OpenAI introduced ChatGPT, explicitly harnessing these magical new capabilities.

Looking forward

Generative AI is genuine magic: even leading researchers at OpenAI, Anthropic, Google DeepMind, or Meta can't fully explain it.

It's also transformative: adoption is outpacing any previous tech innovation, including the internet. I firmly believe that leaders who aren't actively exploring and applying generative AI now will find themselves struggling to keep up by the end of 2025. Neither fear nor ignorance will provide protection.

When you're ready, here's how I can help:

Andrew Walker
Consulting to for-purpose CEOs to deliver more impact with existing teams and systems - by freeing humans up from admin.
https://www.linkedin.com/in/andrew-walker-the-impatient-futurist/

Did someone forward this email to you? Want your own subscription? Head over here and sign yourself right up!

Back issues available here.