“How wonderful that we have met with a paradox. Now we have some hope of making progress.”
Niels Bohr
William Blake’s earlier version of Bohr’s quote was, “Without contraries is no progression.” But there is another line, from William Blake, that has been rattling around in my head while thinking about one of the most interesting challenges currently being posed to artificial intelligence:
“There is no other God, than that God who is the intellectual fountain of humanity.”
It feels like an appropriately Blakean way of saying something rather important. Perhaps the greatest breakthroughs do not come from intelligence simply knowing more than everyone else. Perhaps they come when intelligence is able to see something differently. To escape psychological inertia, and to make cognitive leaps.
Demis Hassabis has suggested what might be called an Einstein Test for AI. Can a machine make the kind of creative leap that transformed our understanding of the universe in the hands of Albert Einstein? It is a fascinating question. But it prompts another. Before asking whether AI can become another Einstein, might it be worth asking whether we can understand, at least in part, how Einstein-like leaps happen? And if we could identify some of the cognitive transformations involved, could we teach an AI to use them deliberately? That question leads to a rather intriguing thought experiment.
The Pre-TRIZ Einstein Experiment
Imagine that we could freeze history in 1905. Einstein has just published his Special Theory of Relativity. The world now has a radically new understanding of space and time. The speed of light is invariant. Space and time are no longer the absolute background assumed by Newtonian physics. But General Relativity does not yet exist. Einstein will spend the next decade trying to solve the problem of gravity.
Now imagine giving the scientists of 1905 something that did not yet exist. Not knowledge from the future. Not curved spacetime. Not the equivalence principle. Not Einstein’s field equations. Instead, imagine giving them a systematic map of the inventive transformations that humanity had already used successfully before 1905.
That map has a name today. TRIZ.
A map of inventive thought
TRIZ emerged from the work of the Soviet engineer, inventor and researcher Genrich Altshuller. He began with a deceptively simple question: ‘do successful inventions have anything in common? His research eventually suggested that, despite the enormous diversity of technologies and industries, inventors repeatedly used a relatively small number of recurring conceptual moves. These became the famous 40 Inventive Principles.
Altshuller started his research in the wake of the Second World War, and so clearly the 40 Principles as we know them today did not exist in 1905. But the inventions from which such Principles could be extracted already did. Humanity had already discovered the moves. It simply had not yet organised them into a systematic library. So our experiment is not quite as absurd as it first sounds. We are not giving 1905 the answers from the future. We are giving 1905 a better understanding of the patterns contained in its own past.
But first we need a problem. And TRIZ begins with one of the things Bohr found so exciting. A contradiction.
Gravity must act at a distance… but not too quickly
By 1905, Newtonian gravity remained spectacularly successful.
It explained the motion of planets. It described falling bodies. For more than two centuries, it had provided one of humanity’s most powerful descriptions of the physical universe. But Newton’s gravity carried an uncomfortable assumption. Gravity appeared to act across distance instantaneously. Move a massive object and, in the Newtonian picture, its gravitational influence is reflected everywhere immediately.
Then came Einstein’s Special Relativity. One of its central implications was that physical signals and influences could not propagate faster than the speed of light.
And suddenly we have a contradiction.
We want gravity to retain its successful ability to influence objects across enormous distances… but we must not allow physical influence to travel instantaneously.
Or, more provocatively:
Gravity must act at a distance… but it must not act at a distance too quickly.
This is precisely the kind of situation that TRIZ finds interesting. Neither side of the contradiction is obviously wrong. Newtonian gravity works. Special Relativity works.
The difficulty is that the two successful descriptions cannot simply be placed together without creating a problem.
That is where progress often begins.
Step One: Freeze history
So let us establish the rules. The year is 1905.
We are allowed everything that was scientifically known at that point.
We are allowed every invention made by humanity before that date.
We are allowed the accumulated evidence of successful problem solving contained in the history of technology.
But we are not allowed to cheat.
No curved spacetime.
No General Relativity.
No Einstein field equations.
No knowledge of what Einstein will discover during the next ten years.
No answer may be smuggled backwards through history.
Instead, we have only a contradiction.
How can gravity influence across distance without violating the relativistic constraints imposed by the speed of light?
Step Two: Ask who had already solved similar problems
This is where the experiment becomes interesting.
Translated crudely into TRIZ terminology, we have a conflict involving distance and speed.
The modern Contradiction Matrix is the result of accumulated research into which Inventive Principles have historically been used to resolve particular kinds of conflict.
For the distance-versus-speed contradiction, the most highly ranked recommendations include three particularly intriguing Principles:
13 – The Other Way Round
17 – Another Dimension
14 – Spheroidality/Curvature
The Matrix itself obviously did not exist in 1905. But here is the important question, ‘had humanity already used these cognitive moves to solve problems before Einstein needed them? The answer is plainly yes.
Turn the problem around
Consider the moving sidewalk. The conventional way of helping someone travel a distance faster might be to make the person move faster. But the moving sidewalk reverses the relationship. Keep the traveller relatively still. Move the world beneath them instead. That is Principle 13: The Other Way Round. Change what moves and what remains stationary. Reverse the conventional relationship between system and environment.
Leave the existing dimension
Consider the spiral staircase. The problem is one of distance, but the available footprint is limited. The obvious solution – extend the route – consumes more space and therefore requires either more time or more speed to get from bottom to top. The inventive solution is to leave the plane. ‘Use another dimension’. The spiral staircase allows substantial travel distance to be accommodated within a compact footprint. That is Principle 17. If the problem cannot be solved within the existing dimensional arrangement, change the arrangement.
Try curvature
And then there is perhaps the oldest and simplest example of all. The wheel. Dragging a load across a surface creates resistance. The solution is not simply to drag harder. It is to transform linear motion. Introduce rotation. Introduce curvature. That is Principle 14. And it is difficult to imagine a more ancient demonstration of the fact that sometimes an apparently linear problem becomes soluble only when we stop thinking in straight lines.
Three cognitive provocations
Now return to 1905. We still do not know the answer. But we now have three historically demonstrated cognitive transformations. And our 1905-TRIZ allows us to turn them into questions.
Principle 13: The Other Way Round
Change the relationship between observer and phenomenon.
What happens if we reverse the conventional viewpoint?
What does gravity look like from inside the system experiencing it?
What happens if the observer is falling?
This is not yet the equivalence principle.
But it is a route into precisely the kind of thought experiment that made the equivalence principle possible.
Principle 17: Another Dimension
What if the problem cannot be solved within the existing dimensional framework?
Are we describing the universe using the wrong arrangement of dimensions?
Could the difficulty disappear if space and time are treated differently?
Again, no answer has been supplied.
But an entirely new territory has been opened for exploration.
Principle 14: Curvature
And then comes the most provocative transformation.
What if we stop assuming flatness?
Must the geometry within which gravity operates be flat?
What happens if curvature is introduced?
And then the question that, in 1905, would have sounded extraordinary: What if space itself can be curved?
We have not rediscovered General Relativity. Einstein would still have had an enormous amount of work to do. But something important has happened. We have entered the conceptual territory in which General Relativity becomes thinkable.
And perhaps that is the real point of the Einstein Test.
The purpose is not to produce Einstein
The Pre-TRIZ Einstein Experiment cannot prove that TRIZ would have produced General Relativity. That would be an absurd claim. Einstein brought extraordinary scientific judgement, mathematical capability, persistence and imagination to the problem. He also had to distinguish between a provocative thought experiment and a description of physical reality. TRIZ cannot tell us which of our ideas nature will permit. But it can help with an earlier and equally important problem. Where should we look?
The experiment might therefore support a much more modest – but potentially profound – proposition… Systematic inventive heuristics may be capable of generating the kinds of conceptual provocations from which revolutionary scientific insights emerge.
That is rather different from saying that creativity can be reduced to an algorithm. It says that some of the transformations used in creative leaps may be recognisable, repeatable and deliberately invoked. And that brings us back to AI.
The question before the Einstein Test
If Hassabis’ Einstein Test asks: Can AI make a genuinely novel conceptual leap?
Perhaps there is a preceding question.
Can we identify the cognitive transformations that made Einstein capable of asking questions nobody else had thought to ask—and teach an AI to use them deliberately?
TRIZ represents one attempt to answer precisely that question.
Altshuller did not study Einstein. He studied invention. His remarkable proposition was that the accumulated history of human ingenuity contained recurring patterns. Successful inventors did not all think identically. But their solutions repeatedly involved similar transformations. Separate. Invert. Combine. Move into another dimension. Introduce curvature. Use available resources. Change the relationship between the system and its environment. And many others. The provocative possibility is that these are not merely engineering tricks. They are rather elements of a much more general library of cognitive moves. Moves that can be applied whenever a successful model encounters a contradiction it cannot resolve.
Which leads to the much bigger question, Why wait for another Einstein?
The Einstein Test asks, in effect, can AI produce one Einstein-level breakthrough? But if the cognitive transformations underlying major breakthroughs can be identified from the accumulated history of invention, why wait? Why wait for another Einstein to emerge? Why not systematically apply those transformations to the unresolved contradictions that science can already see?
After all, science is full of them. Successful theories that do not comfortably fit together. Measurements that refuse to behave as expected. Models that explain almost everything except the inconvenient thing that they cannot explain at all. The existence of such contradictions is usually treated as evidence of a problem. Bohr saw something else. “How wonderful that we have met with a paradox. Now we have some hope of making progress.” Perhaps every major unresolved contradiction in science should be treated in the same way. Not as a monument to ignorance. But as an invitation. An invitation to search.
The bigger destination
Here, perhaps, is where the Pre-TRIZ Einstein Experiment becomes something more than a historical curiosity.
Even before Einstein knew the answer, the accumulated history of human invention already contained cognitive transformations capable of pointing towards it.
If that proposition is even partly true, its implications are enormous. Today we possess not only the inventive history available in 1905. We have more than a century of additional science and technology. We have millions of documented inventive cases. We have systematic knowledge extracted from those cases. And we increasingly have AI capable of maintaining, searching and connecting conceptual spaces far larger than any individual human can hold in their head. Perhaps the most useful role for AI is therefore not to sit in isolation waiting for a mysterious moment of artificial genius. Perhaps it is to become a partner in a systematic process.
Identify the contradiction.
Search the accumulated history of successful transformations.
Find analogous conflicts.
Generate cognitive provocations.
Combine them.
Explore the resulting conceptual territory.
And then allow human scientists to bring the things AI does not possess… judgement, purpose, context, intuition and, perhaps above all, the ability to recognise what is genuinely important.
The question is no longer simply, Can AI become Einstein?
The more interesting question is, Can AI help thousands of scientists explore the conceptual territory from which the next Einsteins will emerge?
And perhaps there is an even bigger possibility. If Altshuller was right that humanity’s history of invention contains a discoverable library of recurring cognitive transformations, then that library should not be confined to engineering or business or IT problems. It should be pointed at the contradictions that are still holding science back. The great unanswered questions. The theories that work beautifully… until they collide with another theory that also works beautifully. The paradoxes.
Because perhaps Bohr was right.
Perhaps the moment we encounter a contradiction is not the moment progress stops.
Perhaps it is the moment when progress finally has somewhere to begin.
And if we can systematically identify the cognitive moves that have helped humanity escape contradictions in the past, perhaps we do not have to wait for another Einstein to arrive.
Perhaps we can start looking now.