Reply to thread

Message

<blockquote data-quote="MNblockhead" data-source="post: 9700054" data-attributes="member: 6796661">No, the summaries are useful. But Gemini also tries to determine &quot;action items&quot; or &quot;next steps&quot; based on the conversation. So if you say &quot;I'll talk to greg to get the cost estimates for the new server&quot;, Gemini will add a &quot;Get new-server estimate from Greg&quot; (or something similar) to the &quot;Next Steps&quot; section. I'm not sure how accurate or useful it would be in my typical game session. For example, over a several hour session there could be a number or planning discussions just for a specific encounter. If it pulls &quot;next steps&quot; out of those conversations, it won't be useful. I don't know if it is trained on gaming sessions enough to distinguish planned actions for an encounter from more useful action items like &quot;I'll calculate XP between sessions&quot; or &quot;we'll divide the treasure next session.&quot;I haven't had an opportunity to test, but I'm assuming that the LLM is trained mostly on business meetings. I could see there being a lot of &quot;false&quot; action items that it could pull from a long gaming session.But I expect that the summaries of what occurred during the session would be quite useful.</blockquote>

[QUOTE="MNblockhead, post: 9700054, member: 6796661"] No, the summaries are useful. But Gemini also tries to determine "action items" or "next steps" based on the conversation. So if you say "I'll talk to greg to get the cost estimates for the new server", Gemini will add a "Get new-server estimate from Greg" (or something similar) to the "Next Steps" section. I'm not sure how accurate or useful it would be in my typical game session. For example, over a several hour session there could be a number or planning discussions just for a specific encounter. If it pulls "next steps" out of those conversations, it won't be useful. I don't know if it is trained on gaming sessions enough to distinguish planned actions for an encounter from more useful action items like "I'll calculate XP between sessions" or "we'll divide the treasure next session." I haven't had an opportunity to test, but I'm assuming that the LLM is trained mostly on business meetings. I could see there being a lot of "false" action items that it could pull from a long gaming session. But I expect that the summaries of what occurred during the session would be quite useful. [/QUOTE]

Verification