By now you have heard a lot about the promise and the limits of CG. This presentation is meant to show you how we can realize more of the promise by overcoming some important limitations. The way I’ll demonstrate this is by showing how two core ideas that have been around for a while when implemented well can increase the power of machines to act like super brains. The two ideas are AM and CD based on KC.
• Love story between Kolmogorov Complexity (KC) & Associative Memories (AM).
Associative memory means the ability to associate huge amounts of data and find pattern in real time much faster than human beings can do. At Saffron Technologies we have scaled AM to Big Data. We use CD a way to find meaning in the data and reason likes humans do but much more powerfully and faster. When CD and AM are combined it is like a match made in heaven that realizes the promise of cognitive computing. This perfect match overcomes the inability of most machine learning approaches to be able to add new knowledge on the fly in a consistent way.
• Use cases.
I’m going to share with you 2 applications where we have applied cognitive computing. Where we are helping human beings to make decisions based on data that had already existed but had no meaning until we applied AM.
o Mount Sinai
• Why are KC & AM a match made in heaven?
What draws our lovers AM & CD together are 3 qualities.
o Context matters
o Compression – sparse coding
• What is Kolmogorov Complexity?
o Discern the signal from the noise to make better decisions
o Alice and the judge
• How do use KC ⊕ as absolute measure for information distance between objects
o Cowboy, saddle and movie
• We need context to resolve ambiguity
o Meaning comes from context
o Cognitive distance allows for context
o Associative Memories allow for context since they implement graph
• What are associative memories?
One message, the weights are deterministic. The weights are the strength of the connection between the neurons. These weights are deterministic. We do not need optimization to calculate them but they are baked in.
• How has Saffron Technologies implemented them?
• What are the applications of Cognitive Distance on top of Associative Memories?
• Summary – What is Saffron’s contribution to cognitive computing?