At first we didn’t understand why everyone had misinterpreted what Lyle Dunning had done. All the video conversation were right there in front of everyone. The videos weren’t tampered with, hidden, or obfuscated in any way. Lyle had shared every single video immediately after his conversations with investors. As he said in his deposition, he was straightforward and honest about what he was doing with the investors from day one.
At first we had watched all the videos to understand what had gone on, and we didn’t understand why everyone was accusing Lyle of deception when it came to his representation of the investors. It all seemed like the investment group just had it out for Lyle, and he had accurately represented the investors, what he had told to the investment group, and in his deposition. It wasn’t until we actually read all of the summaries posted alongside each video conversation that we realized what had happened.
Lyle had posted a video of each investor conversation, and then with it he had submitted the transcript to ChatGPT for summarization. This is where everything went wrong. The summarizations did not accurately represent the conversations at all, and at first we had assumed that the LLM had misinterpreted what was being said, but then we realized that they were doctored by Lyle after generating and before submitting. That was the real crime, but next, not one single person in the investment group had ever watched one of the videos—they just took the summaries at face value and assumed that they were an accurate and honest representation of what was stated.
Throughout the investigation Lyle stayed firm that he had accurately represented the investment group—the video was his evidence. Once he was questioned about the summaries posted alongside each video, he just said simply they are autogenerated from the video conversations, and had nothing to do with them. Once we dug into the “autogeneration” of summaries, we saw that indeed they had been altered. Then we asked everyone in the investment group if they had watched any of the videos or had just relied on the summaries. Every one of them replied that they read only the summaries and believed they were accurate summarizations of the video conversations and it never occurred to them to watch the videos at any point, even once they started to suspect that something was off—they simply trusted that the summary representation was accurate.
