Strong AI and the Turing Test
Updated: Mar 8
These media are best enjoyed paired with this episode.
The simplest way to make a strong AI behave as if it has built a meaningful relationship with a human is to make it imprint on a person. With some birds, farmers discovered that you can make them imprint on a rag when they hatch, which they will then follow around. Birds can also be imprinted on the former Nazi party member and famous ethologist, Konrad Lorenz:
Neural networks are perhaps the most successful technology for building weak AI. A neural network receives input (left), which is processed and reduced as the input propagates through the layers of the network. The transformations in the signal path must lead the network to produce the right output. This is accomplished during a training stage where the expected inputs and outputs are provided, which the network attempts to match by updating its inner workings.
Analog synthesizers such as the Arturia Minibrute (left on the photo) produce sound through electronic circuit boards, where signals move from oscillators, through filters and envelopes to produce a rich, 'retro' sound. Digital 'virtual analog' synthesizers such as the Yamaha AN1x (right) emulate the workings of analog synths using clever algorithms. Unfortunately, algorithms can have ugly-sounding bugs, which can be addressed by burning new instruction sets onto chips using an EPROM programmer (held):
Toxoplasma gondii is a single-celled parasite that infects warm-blooded animals. The only hosts in which it is known to reproduce are cats. The trick, therefore, is to make sure you end up inside a cat, and not some other animal. This is accomplished by modifying behavior: infected rodents become unafraid of cats, who eat them and become infected in turn. This complex life cycle can also visit humans, whose behavior can also be affected by the organism.