Greetings Everyone,
I am Monroe Gordon, founder of Arti’s Connectome. We are a new AI research company looking to start over with AI. We want to develop new AI systems directly inspired by the brain and imbued with artistic design. We also want our AI systems to be trustworthy, reliable, and completely interpretable.
We look forward to sharing our insights, research, and discoveries!
Thank you!
Monroe Gordon
Arti’s Connectome Founder
@ArtisConnectome
What does “completely interpretable” mean?
Specifically, I am referring to any artificial neural network. Artificial neural networks are considered “black boxes,” because no one actually understands what the output of neurons in the deeper hidden layers are representing or mean. This makes it very difficult to know for sure if the neural network is learning proper features and representations. We have tools to help us try to figure out what the outputs of the hidden layers may mean, but none of it is fully accurate. Plus, the neurons could respond to several different features, but there is no way for us to find out every thing it responds to.
We want to create neural networks based on biology that respond to features we have designed it to respond to. This means that we will be creating and continuously evolving a large feature library that will cover various domains, so that we and anyone else that uses our AI systems will know what the neurons’ outputs represent. This is what will make our AI “completely interpretable,” and it will make it easier to understand and correct any errors that the AI may make.
Monroe Gordon
Arti’s Connectome Founder
@ArtisConnectome
Interesting. Thanks for your reply.