In the 1997 Robin Williams flick Flubber, an absent-minded professor creates a sentient ball of goo with incredible capabilities.
Now, more than 25 years later, scientists have made a surprising discovery that could bring Flubber into the real world.
Researchers from the University of Reading have created a non-living ‘hydrogel brain’ which is capable of playing the video game Pong.
Using a plate of electrodes hooked up to the classic game, the water-based jelly even managed to get 10 per cent better as it practised.
While it might not be quite as bouncy as Robin Williams’ invention, the researchers believe this breakthrough could change the future of artificial intelligence.
The researchers were inspired by an earlier experiment which showed that a plate of brain cells could play Pong when hooked up to electrodes.
Those experiments suggest that ‘something that resembles intelligence’ could be made out of very simple systems.
To take that idea one step further, the researchers investigated whether non-living jelly could learn to play the game too.
To test the gel’s game-playing prowess, the researchers hooked an ‘ionic hydrogel’ up to a single-player version of Pong in which the player had to bounce a ball off a solid wall for as long as possible.
Surprisingly, the hydrogel developed a sort of ‘memory’ which allowed it to improve its performance over time.
Researchers have gotten one step closer to making real-life flubber as they design a jelly capable of playing the video game pong
With practice, they found that the jelly became 10 per cent better at the game and was able to hold longer rallies.
Lead author Dr Vincent Strong of the University of Leeds says: ‘We showed that hydrogels are not only able to play; they can actually get better at it over time.’
However, the researchers are not saying that the jelly is sentient or even that it necessarily ‘learns’ how to play.
Hydrogels, like gelatin or agar, are complex chains of polymers which become jelly-like once you add water.
The only difference between the jelly in this experiment to the one in your kitchen is that the researchers used an ‘electro-active polymer’.
Ionic hydrogels (pictured) like the one used by the researchers are able to develop a form of ‘memory’ by shifting in response to electrical stimulation
The researchers found that a hydrogel could learn to play the classic game Pong (pictured) and even improve by up to 10 per cent with practice
These polymers form a gel that can respond to electrical stimulation thanks to the presence of charged particles, or ions, trapped within its structure.
The kind of ‘memory’ displayed by the hydrogel is entirely down to where those floating ions end up.
To make the jelly play Pong, it was first sandwiched between two plates, each featuring a three-by-three array of small electrodes.
Six of those electrode pairs, making a three-by-two grid, were charged to simulate the movement of the ball across the screen.
The remaining three pairs were then used to represent the back wall where the player moves their paddle.
The researchers placed the jelly in a grid of electrodes (left of picture) which allowed the jelly to simulate the movement of the paddle and ball through electrical stimulation
The researchers were inspired by a previous study which showed that a plate of mouse neurons (pictured) in what they called a ‘DishBrain’ could learn to play Pong
Those six electrodes measured where the ions in the hydrogel were most concentrated and moved the paddle to that location.
When the hydrogel is stimulated, the charged ions move about, dragging water molecules with them and changing the jelly’s shape.
As the ions moved, the point with the highest current would shift on the rear wall as the ball travelled across the screen, allowing the jelly to change the paddle’s position.
At the start, all of the ions are spread out evenly in the gel so the paddle moved somewhat randomly.
But as the ball moved around the court, adding more current, the ions shifted and built up in areas where the ball tended to go.
Dr Strong explains: ‘Over time, as the ball moves, the gel gathers a memory of all motion. And then the paddle moves to accommodate that ball within the simulated environment.
The hydrogel ‘learned’ to play thanks to the movement of charged particles which built up in areas where the ball was simulated to be most often. In this experiment (pictured) the jelly reached peak performance after 20 minutes of practice
‘The ions move in a way that maps a memory of all motion over time, and this “memory” results in improved performance.’
This memory allows the gel to move the paddle into the ball’s path more often, creating longer rallies.
The researchers claim that this basic form of memory is actually quite similar to what happened in the previous brain cell experiments.
Co-author Dr Yoshikatsu Hayashi, also from the University of Reading, says that the basic principle is the same.
In both hydrogels and neurons, charged ions move into a distribution which maps with the loops of motion in the simulated world of Pong.
Dr Hayashi says: ‘In neurons, ions run within the cells; in the gel, they run outside.’
Beyond learning to play a simple game like Pong, the researchers hope to extract the algorithms which allow the hydrogel to acquire memories to form a new simple basis for artificial intelligence
The main difference between the two is that the hydrogel is a slower learning – taking 20 minutes to reach peak performance as opposed to 10 minutes for the brain cells.
In the future, the researchers believe this could provide a novel kind of ‘intelligence’ to form the basis of new AI.
At the moment most AIs are based on the arrangement of neurons in the brain, hence the name ‘neural networks’.
However, the sort of memory exhibited by a hydrogel could provide a simpler basis for intelligent algorithms.
The researchers say their next goal is to extract algorithms from the hydrogel which allow for this unique form of memory.
HOW ARTIFICIAL INTELLIGENCES LEARN USING NEURAL NETWORKS
AI systems rely on artificial neural networks (ANNs), which try to simulate the way the brain works in order to learn.
ANNs can be trained to recognise patterns in information – including speech, text data, or visual images – and are the basis for a large number of the developments in AI over recent years.
Conventional AI uses input to ‘teach’ an algorithm about a particular subject by feeding it massive amounts of information.
AI systems rely on artificial neural networks (ANNs), which try to simulate the way the brain works in order to learn. ANNs can be trained to recognise patterns in information – including speech, text data, or visual images
Practical applications include Google’s language translation services, Facebook’s facial recognition software and Snapchat’s image altering live filters.
The process of inputting this data can be extremely time consuming, and is limited to one type of knowledge.
A new breed of ANNs called Adversarial Neural Networks pits the wits of two AI bots against each other, which allows them to learn from each other.
This approach is designed to speed up the process of learning, as well as refining the output created by AI systems.
