|
|
|
|
|
|
 |
 |
|
Savant AI: The Computable Human
|
|
|
|
|
|
| Organization: | SavantX, Inc. |
| Organization: | SavantX, Inc. |
|
|
|
|
|
|
Wolfram Technology Conference 2015
|
|
|
|
|
|
Champaign, Illinois USA
|
|
|
|
|
|
Savant is a weak AI system specializing in helping users Search, Discover and Analyze very large data sets. Savant is not semantically based, but rather views human-generated data as a type of encoded sensor output. It is currently in use at a large energy production plant where it applies its generalized relationship analysis capabilities towards maintaining the safe and efficient operation of the plant. Savant uses custom algorithms for its massively parallel workflow, is almost entirely Mathematica-based and can readily assimilate large amounts of human and machine-generated-data spanning a vast range of data types including: TXT, PDF, TIF, JPG, XLS, DOC, PST, MSG, EML, BAK, SQL, and more. Automated OCR is just one of Savant’s features—helping modernize entire legacy datasets as well as grant clear-text transparency to images both new and old. Savant is a WL package that is isometric in nature: all (28 to 200+) Wolfram kernels run the same Mathematica code… configuring themselves to the highest priority task at hand and using continuous and discreet manufacturing methods. All this allows the application of AI science to many aspects of its tasks including Classify, energy reduction graph theory, Predict, etc. This presentation will discuss Savant in some detail, including a brief history, the use case, hardware and software architecture and techniques, including performance characteristics. Some of the details of Savant science will be presented. A short demo of Savant will end the presentation.
|
|
|
|
|
|
|
|
|
 |
|
|
| 1445364133.nb (3.9 MB) - Mathematica Notebook |
|
|
|
|
|
|
|
| | | | | |
|
For the newest resources, visit Wolfram Repositories and Archives »
