On Thu, 11 Jul 2024 at 11:34, Josh Boyer <jwbo...@fedoraproject.org> wrote: > > On Wed, Jul 10, 2024 at 12:46 PM Daniel P. Berrangé <berra...@redhat.com> > wrote: > > > > On Wed, Jul 10, 2024 at 09:36:35AM -0400, Stephen Gallagher wrote: > > > I'm guessing Josh's response was an attempt to use AI to generate a > > > summary. In the future, please label it as such, so people will read > > > it with a critical eye. It has some flaws which I will address inline: > > > > If content is labelled as written by AI I'd skip reading it entirely. > > Reading "with a critical eye" implies we have to either guess (or > > research) which parts are accurate and which are wrong, neither option > > being desirable. > > > > I'd ask for human written summaries, or none at all, in prefence to > > misleading AI summaries. > > Human in the loop is of course important. The intention would be to > have FESCo review any AI generated content for accuracy before sending > it out. From a consumer perspective, you get content that has been > validated by a human and you should be able to trust it like you would > any other email sent. > > If FESCo doesn't want to use AI at all, I don't care. I just want > better summaries than what zodbot is currently providing (which aren't > reviewed either afaics...) >
The 'summaries' are just what the person running the meeting is collecting during the meeting while trying to also keep the meeting running AND be a participant of the meeting. The person doing this is a volunteer and generally has zero training in note taking, running meetings or a dozen other things which are expected. If better notes are expected, then better training or having an agreed upon professional whose sole job is this task is generally needed. I don't know if the AI summary came from the general notes of the meeting or if it was built off the actual conversations between people. If it was built off of https://meetbot.fedoraproject.org/meeting_matrix_fedoraproject-org/2024-07-09/fesco.2024-07-09-17.00.log.txt like a professional transcriber (or whatever the role is called these days), then the next step is to send it around to the participants for 'approval'. This is why various meetings start off with a 'approval of the previous minutes' at the start of a meeting so that people who need to make corrections can do so and corrections can be voted on as needed. If LLMs become this 'professional' body, then there will probably need to be an open way to know their 'credentials' to make sure that summaries are fair. What was the tool and dataset used What was the 'meeting' data fed to the tool What was the query (or queries) used What were the retraining queries used to change 'summaries' that were found to be wrong. -- Stephen Smoogen, Red Hat Automotive Let us be kind to one another, for most of us are fighting a hard battle. -- Ian MacClaren -- _______________________________________________ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
