Behind the video
Who Has The Best Data? 🤔
Overall Thesis
The speakers discuss proprietary data as a competitive advantage for AI development, highlighting Google's video and email data and Reddit's human discussions. Reddit's arguments and varied perspectives are considered useful for models that help users reason through different situations.
Narratives
GOOGLAlphabetGoogle is described as having an inherent AI data advantage through its video and email data. That advantage is qualified by uncertainty about whether email data can legally be used for model training.
Key Arguments
- Google owns video and email data that could provide an advantage over AI labs that must license data.
Risks acknowledged
- The speakers question whether Google's terms permit email data to be used for model training and acknowledge that they have not checked.
The speakers view Reddit's human conversations as valuable AI training data, endorsing the usefulness of arguments and differing perspectives. This is a favorable assessment of the data rather than a forecast of Reddit's stock or business performance.
Key Arguments
- Reddit discussions contain human arguments and exchanges of differing opinions.
- Different use cases and the process of reasoning toward an answer may be useful beyond supplying a single answer.
- A speaker explicitly agrees that Reddit provides these conversational qualities.
Hedges & Caveats
- Google's right to use email data for model training is questioned, and the relevant terms have not been checked.
- The value of X as an AI training data source is questioned.
- No stock-price targets, investment commitments, or directional forecasts are stated.