Passing the Turing Test has been considered a major milestone in the development of Artificial Intelligence (AI). Alan Turing proposed his famous “imitation game” back in 1950, asking ‘judges’ to chat with two hidden individuals and figure out who’s human and who’s the AI.
So, how convincing has AI become?
Well, according to a preprint study by Cameron Jones and Benjamin Bergen at UC San Diego’s Department of Cognitive Science, 264 human judges were fooled into thinking GPT-4.5 was human, 73% of the time.
Easy peasy, does that mean we’ve crossed the finish line?
Not quite. Sure, imitating a typical human is impressive, but hardly a reason for the sci-fi panic about robots taking over the world. Where does AI go from here?
The Turing Test was just a warm-up. Next is Artificial General Intelligence (AGI), where machines think as well as most humans. And if things get wild enough, we might even see Artificial Super Intelligence (ASI) — where AI blows past human brainpower, discovering new knowledge faster than we can say, “What just happened?”
So, how would AI stack up to a not-so-generic human?
Here’s a challenge: What if we built an AI version of one of history’s biggest brainiacs, Isaac Newton — but with a catch? Instead of giving digital Newton all the answers from today, we’d only let it have the exact knowledge young Isaac had up to his graduation day at Cambridge back in 1665.
No Einstein, no quantum physics, and certainly no Googling for answers — just 17th-century texts and some classic old-school curiosity.
But why Newton and not a more modern discoverer?
Glad you asked. Newton lived way before mass media and the internet. That means our AI’s training data will be dramatically smaller, orders of magnitude less than the enormous datasets used to train today’s large language models (LLMs). This leaner dataset isn’t just historically accurate; it’s also quicker and cheaper when training the AI.
Armed with just the essentials, our digital Newton will try to recreate the groundbreaking discoveries the real Newton made during the two incredible years after earning his bachelor’s; things like the laws of motion, calculus, universal gravitation, and figuring out why white light breaks into a rainbow of colors.
To build this retro digital Newton, we’d load it up with exactly what young Isaac had studied:
- Ancient Classics: Works by Aristotle, Euclid, and other Greek and Roman thinkers that formed the cornerstone of Newton’s early education.
- Medieval Scholarship: Mathematics and optics from Islamic scholars like al-Khwarizmi and Avicenna, which deeply influenced European intellectual circles.
- Renaissance Breakthroughs: Game-changing findings by Copernicus, Kepler, and Galileo, who transformed our understanding of astronomy and motion.
- Newton’s Own Early Notes: Newton’s personal writings, filled with early insights and formative questions about mathematics, optics, and physics.
Then the fun begins. We’d challenge our digital Isaac with puzzles similar to what the real Newton tackled post-graduation, asking things like:
- “You’ve observed apples fall, pendulums swing, and objects collide; what universal principles can you derive from these motions?”
- “Planets trace elliptical orbits around the sun; how could you explain their motion mathematically?”
- “Playing with sunlight and prisms; what’s your theory about the nature of colors and light itself?”
Here’s where things get exciting: our digital Newton won’t just absorb knowledge passively, it will actively pursue it, just like the real scientist himself. Digital Newton will pose hypotheses and design experiments that could’ve been reasonably conducted during the real Newton’s lifetime.
We’ll run (simulate) these experiments; no fancy particle accelerators or quantum computing, just simple, Newton-era gear; and feed the results back to our AI. With each experiment, our digital genius will refine its understanding, inching forward on its own journey of discovery.
Could our digital Newton replicate his namesake’s astonishing breakthroughs? Would it discover gravity, invent calculus, and decode the secrets of color? Or would it stumble at some hidden hurdle, revealing unexpected limits of AI creativity?
Whatever happens, the Newtonian Challenge opens fascinating questions. Sure, AI might ace the Turing Test, but can it innovate from scratch and build upon limited historical knowledge? Or will it always need hints from human intuition and real-life experience to break new ground?
By creating and testing our digital Isaac, we might find answers about how creative AI can be, and how close we might get to ASI. And hey, if our digital Newton pulls this off, we might just be standing at the edge of a brand-new era, where machines not only learn from past geniuses but pioneer paths of discovery themselves.
Who’s ready to see if AI Newton can rewrite history?
Thanks for reading. Feel free to repost, comment or ask a question. Here’s another article you might find worthwhile: This Virtuous Cycle Will Lead To Artificial Super Intelligence