AI: Logic Design and Verification
Description
There has been much discussion on the use of Artificial Intelligence (AI) in many fields. This study is a first look at the use of AI in the field of digital electronics, primarily generating VHDL designs as well as self-checking test benches. A series of experiments were conducted, giving the AI design and verification tasks for standard logic constructions as well as modest-sized applications. The AI’s performance was analyzed in detail.
Detailed example
The error or anomaly rate was relatively high for the AI. Indeed, it would be higher than a young engineer. Some of the anomalies were minor and could be corrected by the human or be directing the AI. Other anomalies were moderate to major; a considerable amount of human time would be necessary to fix the problems; in some cases the AI output would have to be replaced by the human.
AI / analytics pattern
Reinforcement Learning: AI trained through trial and error using rewards and penalties to optimize decision-making policies.
Automation level / stage
a) Pre-deployment – The use case is in a development or acquisition status.
Expected benefit
The key benefit of using AI for these applications is the huge speed advantage for generating designs and test benches, when compared to the human. This is for NASA and organizations that do similar engineering; not for the general public. In the course of this study, some cases that would take a human a number of hours to complete, would be done in less than a minute by the AI.
Controls / human review
ATO: Not reported; PIA: Not published