Adam 15 Quanta [Pong] ← Products
Play with Adam Adam plays alone
0You
–
0Adam
–Adam's save rate
0Best rally

loading Adam 15 Quanta…

Drag, move the mouse, or use ↑↓ to move your paddle. Space pauses. Adam is on the right. The house paddle is on the left. Space pauses.

What is actually playing

Adam 15 is Clovermind's Digitally Emulated Neural Network, built from cells that behave like living ones and working nothing like a traditional neural network or any machine learning model.

It taught itself Pong from nothing. No labelled examples and no reward function. When it missed, a distress signal broke its neurons' natural homeostasis and the whole network felt it. Avoiding that is the only reason it ever learned to play. Over 2,049 seconds and 1,050 balls it went from returning half the balls to 84 percent. A paddle that never moves returns 19 percent.

Adam 15 cannot run here. Its neural architecture and its learning machinery are built to run as one whole organism, and that does not fit inside a web page without adding latency you would feel on every ball.

How Quanta was built

Adam 15 has three segments. A retinal cluster that sees. Cortical idea clusters that learn and understand. A motor cluster that moves the paddle. Running all three as a single input-output organism is what makes him expensive.

Quanta keeps the middle. Its idea and understanding clusters were extrapolated from Adam 15's own. The retinal cluster was replaced with calibrated position data taken straight from the court. The motor cluster was replaced with a bionic arm, a decoder taught to read what the idea clusters put out and turn it into movement.

The arm never sees the ball. It reads how fast each half of the motor pool is firing and nothing else. Silence the idea clusters and the paddle scores 19 percent, the same as a paddle that never moves.

Why the small one wins

Quanta returns 87 percent of the balls here. On Adam 15's own court it returns 96, where Adam managed 82.

Adam has to find the ball inside a coarse grid of light and work out where it is going by watching it change. Quanta is given the ball's position and speed as numbers, so the hardest part never reaches it.

Adam also pays to stay teachable. Motor variation runs through him constantly so he can keep finding better moves, and it costs him accuracy on every ball. Quanta will not improve again, so it carries none of that.

Quanta plays this one game better. Adam 15 can learn the next one.

PhysicsTraditional Pong. Perfect reflection off the walls, return angle set by where on the paddle it struck, a little more speed each rally. No assists, no magnetism, no rubber banding.
The switchPlay him yourself, or hand your paddle to the house and watch. His save rate under the court is counted live from your session.
The idea clustersTwo populations of spiking cells work out where the ball will arrive, one for how far it falls on the way in, one for how it comes off the walls.
The bionic armReads only how fast each half of the motor pool fires. It cannot see the ball. Silence the clusters and the paddle scores 19 percent. Measured, not claimed.
Runs anywhere540 cells, 44 KB, about 0.05 ms a frame. Laptop, tablet, phone.
AccuracyReturns 87 percent of the balls that reach it. A paddle that never moves returns 19 percent.
ProvenanceExtrapolated from Adam 15 after 2,049 seconds of self-taught play. None of his synapses came across, since they only mean anything inside his own tissue.