Neural Horizons AI

Opinion & Predictions

This Didn't Happen in Five Years. It's Been 76 Years in the Making.

AI feels new because the last five years were loud. The record says otherwise: this is a 76-year arc that finally hit its steep part. Here's the verified history, the scientists who called it decades ago — physicists, chemists, and computer scientists among them — where I personally stand, and my predictions for the next two years. Agree with me or see it differently; either way, I'd love for this to start a conversation rather than end one.

Lisa Warren, Neural Horizons AI 17 July 2026 9 min read

The Atom of Intelligence — 76 Years in Motion

Three orbits, three different levels of independence from human control — Agentic AI, Autonomous AI, and AGI — circling 76 years of compounding capability. The strand in the corner is the DNA underneath all of it: the biological inspiration the field keeps returning to. Click any orbit, the nucleus, or the DNA strand below the diagram to see what it means.

THE DNA OF AI

An illustrative visualization, not a literal dataset — a way to picture how far each idea is from human oversight. The chips below are the dated, verifiable record.

Click any orbit, the nucleus, or the DNA strand above to see what it means:

1950 · Turing Test proposed 1956 · Dartmouth Conference, term "AI" coined 1974–80 · First AI Winter 1997 · Deep Blue beats Kasparov 2012 · Deep learning breakthrough (AlexNet) 2017 · Transformer architecture published 2024 · Nobel Prizes in Physics & Chemistry go to AI pioneers Nov 2022 · ChatGPT goes public 2026 · Agentic AI in production

Seventy-six years lie between Alan Turing asking "Can machines think?" and today. AI has been built, defunded, doubted, and rebuilt at least twice before most of us ever typed a prompt. What happened in the last five years wasn't the beginning of the story — it was the moment compute, data, and architecture converged and decades of groundwork compressed into a run of very loud years. That's not a shorter story. It's a longer fuse — and in 2024, the Nobel committees themselves confirmed it, awarding the Prize in Physics to the inventors of neural network learning and the Prize in Chemistry to the team behind AlphaFold.

To make that concrete: the three orbits in the diagram are Agentic AI (plans and acts with a human checkpointing it), Autonomous AI (acts continuously, with no checkpoint in between), and AGI (still theoretical — the dashed orbit, because nothing has earned a solid line yet). The DNA strand in the corner is the reminder that none of this was invented from scratch: the field keeps borrowing from biology, from Hopfield's 1982 model of neurons straight through to today's architectures.

How an Agent Actually Thinks

"Agentic" isn't magic — it's a specific, learnable architecture. Strip away the hype and every agentic system runs the same basic cycle. Watch it move through the loop below.

Click any circle above, or a tag below, to see what it means:

This is the same core "plan → act → observe" pattern used across OpenAI, Anthropic, Google, and Microsoft's agent frameworks in 2026 — the human checkpoint is a design choice, not a limitation.

Why the CHECK step is non-negotiable

An agent is only as good — and as fair, and as safe — as the data it was trained on and the checkpoints built around it. That's the entire argument for keeping a human in this loop: not because the underlying math is untrustworthy, but because nobody has yet built a system that weighs consequences the way a person does. It's the same principle behind everything in "Where I Stand," below.

In Their Own Words

Physicists, chemists, and computer scientists — verified quotes spanning nearly a century, plus where mine sits alongside them.

Albert Einstein

Physicist Nobel, Physics 1921

"Imagination is more important than knowledge. Knowledge is limited. Imagination encircles the world."

Saturday Evening Post, 1929

Alan Turing

Mathematician, founder of computer science

"I believe that at the end of the century... one will be able to speak of machines thinking without expecting to be contradicted."

"Computing Machinery and Intelligence," 1950

John Hopfield

Physicist Nobel, Physics 2024

"As a physicist, I'm very concerned about something that is not controlled, something that I don't understand enough to know what limits might be imposed on this technology."

Remarks following the Nobel Prize announcement, October 2024

Geoffrey Hinton

"Godfather of AI," Turing Award 2018 Nobel, Physics 2024

"I'm just a scientist who suddenly realized that these things are getting smarter than us. We should worry seriously about how we stop these things getting control over us."

On leaving Google, May 2023

Fei-Fei Li

Creator of ImageNet, Co-Director, Stanford HAI

"The mission of AI is not to replace humans, but to understand and empower the human mind."

Stanford Human-Centered AI Institute

John Jumper

Director, Google DeepMind — AlphaFold Nobel, Chemistry 2024

"Public data were essential to the development of AlphaFold — exactly what enables our machine learning models to generalise well across such a huge range of proteins."

On AlphaFold's data foundations

Demis Hassabis

CEO, Google DeepMind, Knighted 2024 Nobel, Chemistry 2024

"[This] cannot be compared to standard technological breakthroughs... the magnitude of this technology's impact will be unprecedented — perhaps 10x the Industrial Revolution, at 10x the speed."

July 2026

David Baker

Biochemist, Institute for Protein Design, UW Nobel, Chemistry 2024

"The goal of de novo protein design is to make new proteins that can solve modern-day problems."

Nobel Prize lecture, December 2024

Lisa Warren

Neural Horizons AI, 2026

To be clear: I'm not placing myself alongside the scientists above. Their work is what I find genuinely fascinating — this is just where my own thinking lands in response to it.

"Turing asked if machines could think. Seventy-six years and two Nobel Prizes later, the only question that matters is whether we let them make us smaller or make us stronger. I'm building for the second answer — ethics and augmentation from day one, not replacement dressed up as efficiency. The industries pretending they can sit this one out aren't being careful. They're running out of runway."

Where I Stand

I'm not neutral on this, and I don't think leaders should pretend to be. I'm embracing AI — with morals and ethics built in from the start, not bolted on after the first scandal. My goal has never been to use this technology to make people smaller. It's to make my people work smarter: faster research, fewer repetitive hours, more room for the judgment calls that still need a human being. Augmentation, not replacement. That's the whole bet.

This is moving fast — genuinely fast, faster than most leadership teams are built to track, and I think it's worth saying so plainly. I also believe a meaningful number of industries are choosing not to look closely, not because they've weighed the evidence and decided to wait, but because looking closely would mean admitting how much catching up they have to do. I'd rather be early and occasionally wrong than careful and definitely irrelevant. This isn't about hype — it's about preparing for a future that's arriving whether we're ready or not, in a way that's better for the people inside it, not just the balance sheet.

My Predictions for 2026–2028

Strong opinions, held loosely enough to be proven wrong — but not written softly enough to be ignored. I'd welcome hearing where you see it differently.

Bold call

A named Fortune 500 company will have a public, material incident caused by an under-supervised autonomous agent by 2027.

Not a hypothetical, not an anonymized case study — a real company, a real filing, a real headline. The technology is outrunning the checkpoints.

Bold call

The real skills gap by 2028 won't be "can you code" — it'll be "can you direct an agent." And it will pay more than a degree.

The premium is shifting from producing the work yourself to knowing how to specify, supervise, and correct AI that produces it for you.

Bold call

At least two more G20 governments pass binding autonomous-AI law with real enforcement teeth by the end of 2027.

The Responsible AI Defense Act is the first domino, not the last. Once one government legislates autonomy specifically, the rest follow faster than people expect.

Bold call

Companies that use AI to quietly shrink headcount will underperform the ones that reinvest the productivity gains into their people — and it will show up in the data by 2028.

Efficiency without reinvestment is a short-term trade. The compounding advantage goes to whoever builds capability, not just cuts cost.

Bold call

30–40% of mid-sized companies in construction, logistics, legal, and healthcare admin will still have no functioning AI governance policy by 2028 — despite running AI in production.

That's not caution. That's gambling and calling it patience. The industries most convinced this doesn't apply to them are the ones most exposed when it does.

Bold call

Agentic AI security incidents will keep outpacing governance maturity for at least 24 more months — and we'll see something bigger than anything disclosed in 2026.

Attackers are iterating faster than compliance teams are staffing up. That gap doesn't close on its own; it closes because someone forces it to.

I'd genuinely welcome a different point of view here. If you see it differently, I want to hear it — let's keep this conversation going.

Sources & Citations

Albert Einstein, interview with G.S. Viereck, "What Life Means to Einstein," Saturday Evening Post, 26 Oct 1929.
Alan Turing, "Computing Machinery and Intelligence," Mind, 1950.
Dartmouth Summer Research Project on Artificial Intelligence, 1956.
The Nobel Prize in Physics 2024 — John J. Hopfield and Geoffrey E. Hinton, "for foundational discoveries and inventions that enable machine learning with artificial neural networks." NobelPrize.org.
John Hopfield, remarks on AI risk following the Nobel Prize announcement, October 2024 (reported by AFP/France 24, Pressenza).
Geoffrey Hinton, interviews with CNN and The New York Times on leaving Google, May 2023.
The Nobel Prize in Chemistry 2024 — David Baker, Demis Hassabis, John M. Jumper, for computational protein design and structure prediction (AlphaFold). NobelPrize.org.
John Jumper, Google DeepMind, on AlphaFold and public data — Nobel Prize Chemistry 2024 press materials.
Demis Hassabis, "A Framework for Frontier AI and the Dawning of a New Age," July 2026.
Fei-Fei Li, on human-centered AI — Stanford HAI / McKinsey Author Talks.
David Baker, Nobel Prize in Chemistry lecture, "The Coming of Age of De Novo Protein Design," December 2024, NobelPrize.org.
Agent loop architecture ("plan, act, observe") — consistent pattern across major 2026 agent framework documentation (OpenAI, Anthropic, Google, Microsoft).
Lighthill Report (1974) — first AI winter; IBM Deep Blue match record (1997); Krizhevsky/Sutskever/Hinton AlexNet (2012); Vaswani et al., "Attention Is All You Need" (2017); OpenAI ChatGPT public launch (Nov 2022).
Predictions in this piece are the author's own forward-looking opinion, not verified fact — offered to invite discussion.

Photo credits (Wikimedia Commons): Einstein (1921 press portrait, public domain) · Turing (Elliott & Fry, 1951, public domain mark) · Hopfield, Hinton, Jumper, Hassabis & Baker (2024 Nobel Week portraits, WikiPortraits project, CC BY-SA 4.0) · Fei-Fei Li (AI for Good 2017, ITU Pictures, CC BY-SA 2.0). Images load directly from Wikimedia's servers and are credited to their respective photographers per license terms.