Matt Desmarais' Raspberry Pi-Powered Logic Analyzer Looks for Flaws in Your Arguments, Not Circuits

Maker Matt Desmarais has built a Raspberry Pi-powered device that can help you hone your arguments — by keeping a digital ear open for logical fallacies.

"[The] AI Logic Analyzer [is] a Raspberry Pi-based project to consume audio and analyze it for logical fallacies," Desmarais explains of his creation. "[It uses a] Raspberry Pi, USB sound card (mic[rophone] or HDMI audio extractor), [Pimoroni] Rainbow HAT, [and a] touchscreen. It will listen for speech, [and] when speech stops it will analyze what was said with Whisper and GPT 3.5 Turbo."

Whisper is OpenAI's open source neural network for automatic speech recognition, released by the company last year with a model built from 680,000 hours of multilingual and multitask data collected from around the web. In addition to high-accuracy speech-to-text transcription, it also offers the ability to automatically translate transcribed speech into English — though the latter feature doesn't form a core part of Desmarais' creation.

GPT 3.5, meanwhile, is one of OpenAI's controversial large language models (LLMs) — designed to accept natural language input and respond accordingly. It's a powerful tool, but one which comes with no guarantees of accuracy or truth — and in Desmarais' case which is being cajoled into analyzing the recognized speech and attempting to locate logical fallacies, ranging from the classic ad hominem attack and equivocation to appeals to emotion and authority.

When a fallacy is detected in the transcribed speech, the Pimoroni Rainbow HAT lights up red while an LED segment display offers a count of the number of fallacies GPT 3.5 thinks it found — with an explanation printed to the touchscreen display for those looking for exactly where their speech went wrong.

The Python source code for the project has been published as a GitHub Gist under an unspecified open source license, with more information available on Desmarais' Hackaday.io page.