This Saturday I attended the Humanity+ conference in London which was packed full of incredible talks about the impact of technology on biology and what it might mean for medicine and human longevity (topics you might remember me covering when I wrote a series of posts on Kurzweil’s Singularity theory).
The second presentation was a standout from an instant impact point of view though. Professor Kevin Warwick of Reading University has created a robot with a biological brain that learns. We only saw videos, but they were incredible to watch. The robot was built in 2008, so this is not new news, just new to me.
Prof Warwick and his team explain it all in the video below, but in a nutshell they:
- grew a brain out of rat neurons – the neurons self-divided and multiplied
- identified neural pathways with random testing
- strengthened the pathways they found by repeat stimulus
- connected the brain to electrodes which took input from sensors and controlled an electric motor
- rigged it so the input from the sensors went into the pathways they had strengthened and the output from the pathway triggered the robot motor to stop/start/turn
- put the robot in a tray where the sensors told it to turn when it got near a wall (equivalent to teaching a child to respond to a voice command ‘left’)
- watched on as the robot got better at avoiding the walls itself – to start with it always respond to the sensors by turning, but as it practised more the neural pathways got stronger and the robot got better
Watch one of the videos below, it is amazing.
The first video is 3.5mins and has more explanation. If you are short on time the second video goes straight to footage of the robot in action and is only 58s.
The biological brain in this robot had 100k neurons. The next project is to build a brain with 30m. That is still some way short of the 100bn neurons in a human brain, but will be a hell of a step forward.