Thank you for this analysis. To its credit, Grok was able to adjust its priors.
I'm sure it could. I haven't used Grok so I don't know about its user interface or its persistent memory. I do know that all AI models will eventually break down when the memory gets too large; the AI loses track of its earlier learnings and becomes less responsive the longer the conversation continues. This may not be a problem to the average AI user, but for a researcher who engages in an extensive training session via the AI, it's a problem that the AI begins to break down just when it has reached maximum learning.
That said, I use Perplexity Pro due to it's multiple AI models, as well as it's "Research" mode versus its "Answer" mode. Research mode searches for more sources than Answer mode, which is intended for quick searches to simple questions rather than deep dives for comprehensive analysis.
Also, a March 6 2025 study of AI models by the Columbia Journalism Review showed that Perplexity Pro had the most "completely correct" responses with the fewest "completely incorrect" answers. That's why I use Perplexity Pro right now.
-PJ