Intel introduced four AI Reference Kits. The free kits consist of configurations, libraries and manuals for AI model training. The kits allow data scientists to develop higher quality models in less time, making AI accessible to a larger number of organizations.
CEO Pat Gelsinger used his Intel Vision keynote to emphasize that Intel wants to help companies adopt AI. The strategy makes sense, because AI applications depend on hardware. Intel designs and sells the processors used for AI model training. The more companies train AI models, the greater the demand for Intel’s products.
AI Reference Kits are one of the ways in which Intel supports adoption. Each kit is similar to a construction kit. Say you’re in the market for a kitchen cabinet. You could buy an axe, cut wood, saw planks and glue it all together. Without prior woodworking experience, the result is likely to disappoint. AI applications aren’t very different.
It’s absolutely possible to start from scratch, but a successful project takes time and training. The new AI Reference Kits provide building blocks for data scientists that reduce the required time and expertise.
AI Reference Kits
The four new kits have four different applications. The first kit – utility asset health – is interesting for grid operators. Grid operators typically manage a region’s electricity network. Electricity is transported by means of utility poles. Each pole requires maintenance. The AI reference kit helps develop an AI model that predicts defects in poles.
The second kit – visual quality control – is interesting for pharmaceutical companies. The quality of medicine needs to be reviewed continuously. Quality control is a time-consuming, often manual process. Computer vision technology makes it possible to determine the quality of medicine with cameras and software. The AI Reference Kit helps to develop an AI model that recognizes defects in pills.
The third kit – customer chatbot – is interesting for various companies. The kit helps develop a chatbot. There are several existing open-source kits available, but this model highly efficient and accurate. The final kit – intelligent document indexing – is interesting for various companies as well. The kit helps to develop a model that reads and understands documents.
Intel developed the reference kits in collaboration with Accenture. The first four are just the beginning. Intel will expand the offering over time.