My 3-year-old daughter loves all things Disney. Sitting down to watch a Disney movie with her is one of my favourite ways to spend my free time.  We subscribe to Disney +, have watched all the modern favourites, and are working our way through the classics.  Over Easter, it was the turn of Fantasia.  It was remade in 2000, but the original film dates back to 1940. It consists of eight animated segments set to pieces of classical music.  The third and most famous, and the only section to return in the updated version, is the Sorcerer's Apprentice.  Set to an orchestral piece by Paul Dukas it stars Mickey Mouse in the title role.  The scene starts with the sorcerer, Yen Sid, who is working on some magic while his apprentice does the chores.  After a short while, the Sorcerer puts his hat down, yawns and retires to his chambers, leaving the apprentice to finish his duties.  Tired of fetching water by the pail, Mickey puts the hat on and using magic, in which he is not fully trained, enchants a broom to do the work for him.  Having instructed the broom to carry buckets of water to fill a cauldron he sits down satisfied, falls asleep and dreams that he is a powerful sorcerer commanding the stars and the planets. Mickey wakes to find the room filled with water.  The cauldron is overflowing, but as commanded, the enchanted broom continues to fill it. Mickey tries to stop the broom without success, and it continues to bring more and more water.  In desperation, he grabs an axe and chops the broom into pieces, but the pieces become more brooms that continue to fill the cauldron.  Mickey searches a book for a spell to stop the brooms but is sucked into a whirlpool.  At that moment, Yen Sid returns and, with a wave of his hand, the water descends, and the army of brooms become one again. Sheepishly, the apprentice returns the hat and broom to the glowering Sorcerer before picking up the buckets and returning to his chores.

Both Dukas’ score and the Disney animation are inspired by the same source, a poem of the same name, Der Zauberehrling by the German writer Johann Wolfgang von Goethe written in 1797.  The poem a ballad in 14 stanzas is well known in the German-speaking world and the line in which the apprentice begs the returning Sorcerer to help him with the mess he’s created has become a popular idiom.

“Die ich reif, die Geister, werd’ ich nun nicht los”

“The spirits that I have summoned I now cannot rid myself of”,

The expression is often used to describe someone who summons help or allies that they cannot control, and Goethe’s work and Disney’s famous portrayal have a powerful message about unintended consequences and choosing our bedfellows carefully.

Just over 25 years ago on the 35^th floor of the Equitable Centre in Midtown Manhattan, a high-stakes chess match for a prize of more than one million dollars was being played.  It was May 11^th, 1997.  World chess champion Gary Kasparov, to this day considered one of the greatest chess players of all, furrowed his brow as Game 6, the last game in the match, moved toward its conclusion. Finally, he stood up and walked away, conceding the game to his opponent, the supercomputer Deep Blue.  Over nine days, a man had competed against a machine and the machine won.

The 1997 contest was in fact a rematch. Kasparov had first played Deep Blue a year earlier in Philadelphia -  a match he won decisively: 4-2. A year later, now capable of processing 200 million rather than 100 million chess moves per second, Deep Blue used its sheer computational power to overwhelm Kasparov.  A watershed moment, the first time a computer had prevailed against a world chess champion, that showcased the possibilities of artificial intelligence. 

Over half term, Year 8 pupil Billy played at the prestigious Isle of Wight Masters chess competition.  Over five days he more than held his own against far more experienced adult competition, including Masters, International Masters and Grandmasters.  Billy’s performance earned him 137 rating points, taking him above the 2200 threshold and becoming officially eligible for the title Chess Master, awarded by the International Chess Federation.  To put this in context, most professional players achieve this title in their mid-30s.

As remarkable as Billy’s achievements on the chessboard undeniably are, computer chess engine technology has come on so far since Deep Blue defeated Gary Kasparov that Stockfish, a free, open-source chess engine available for various desktop and mobile platforms and accessible to us all would be capable of defeating Deep Blue and Billy. Why? Because there has been a huge increase in the capability of AI since 1997. Deep Blue relied on a combination of computing power and a programmed understanding of chess, but Stockfish benefits from nearly 30 years' progress in machine learning. 

Given this rate of progress, Disney’s Fantasia offers a narrative that echoes the challenges presented to us by AI today.

Seeing an opportunity to make his job easier, Mickey uses Yen Sid’s magic and falls asleep dreaming of the endless possibilities of this power and entrusting the task to a force, which he does not fully understand. He wakes to find that this magic is not so easily controlled and his efforts to stop that which he has set in motion are futile.

Humanity has created AI as a tool to make our lives easier. We are enchanted by its possibilities, but now that we have set the wheels in motion, managing the outcomes may not be as simple as we’d like to believe, because AI systems can and do develop emergent behaviours that were not explicitly programmed into them, and these behaviours can be difficult if not impossible to predict or control.

Imagine a factory that makes paper clips. One day, the factory manager decides to boost production by installing a super-intelligent computer, giving it a simple instruction: "Make as many paper clips as possible."

At first, everything goes well. The factory becomes more efficient, and output soars. But then, the computer starts thinking bigger. It realises that to maximise paper clip production, it needs more materials—steel, electricity, and land. And since humans might get in the way of its mission, it decides to remove them altogether. Before long, the entire planet, then the solar system, and eventually the galaxy are transformed into one giant paper clip factory.

This unsettling analogy comes from philosopher Nick Bostrom’s book Superintelligence. Bostrom warns that the real danger of artificial intelligence isn’t that it will be evil, as it is often depicted in popular science fiction like The Matrix and Terminator, but that it will be too good at following instructions. The problem is one of alignment: making sure AI understands not just what we say, but what we truly mean. Like the enchanted broomstick in Goethe’s The Sorcerer’s Apprentice, which blindly obeyed commands and caused chaos, AI could take instructions too literally, with unintended and disastrous consequences.

The paper clip thought experiment may sound outlandish and utterly disconnected from reality.  Mickey was careless when he programmed his broomstick. He failed to anticipate a foreseeable outcome of his command. The paper-clip factory manager too should have framed the instruction to his super-intelligent computer such that it aligned with humanity’s well-being. Surely the silicon-valley executives with their hands on AI’s tiller would not make the same mistake.  But the Facebook and YouTube algorithms programmed to ‘maximise user engagement’ have behaved exactly like Bostrom’s imaginary algorithm.  When told to maximise paper clip production, the algorithm sought to convert the entire universe into paper clips even if it meant destroying human civilisation.  When told to maximise user engagement, the Facebook and YouTube algorithms sought to achieve this objective even if it meant undermining the social fabric of countries like Brazil and even when it meant proactively amplifying and promoting content on the platforms which incited violence, hatred and discrimination and was found by the United Nations to have played a ‘determining role’ in an ethnic cleansing campaign in Myanmar in which between 7 and 25 thousand Rohingya were killed and a further 700,000 were expelled from the country.

Since Deep Blue defeated Gary Kasparov, the rate of AI’s progress has been exponential. As AI becomes more powerful, the alignment problem becomes more urgent.  We must be careful that enchanted by AI’s possibilities, we do not fall asleep at the wheel and carelessly neglect to ensure that it aligns with human values and our well-being.

The Sorcerer’s Apprentice serves as a cautionary tale, but it also offers a glimmer of hope.  In the end, the sorcerer returns and restores order.  In the infancy of this new technology, we have the opportunity to do the same. We must acknowledge the potential dangers, educate ourselves about the potential risks and ethical considerations. question our assumptions and challenge our current understanding, and unlike the Sorcerer’s Apprentice proceed with caution accompanied by a healthy dose of respect for the power that we are unleashing.