Yes
", "time": "2024-03-20T05:06:57Z"}, {"author": "Jean Queralt", "text": "Yes, we can.
", "time": "2024-03-20T05:07:00Z"}, {"author": "Benson Muite", "text": "Tes
", "time": "2024-03-20T05:07:02Z"}, {"author": "Tara Tarakiyee", "text": "I can take notes
", "time": "2024-03-20T05:09:02Z"}, {"author": "Stephen McQuistin", "text": "thank you :-)
", "time": "2024-03-20T05:09:39Z"}, {"author": "Jean Queralt", "text": "Apologies, who was the person on the mic now?
", "time": "2024-03-20T05:31:23Z"}, {"author": "Stephen McQuistin", "text": "the first question was from Alexander Railean
", "time": "2024-03-20T05:32:32Z"}, {"author": "Jean Queralt", "text": "Thx.
", "time": "2024-03-20T05:32:41Z"}, {"author": "Alexandr Railean", "text": "Another thing that would be useful is to ask \"are there any mutually contradicting statements in this document?\", this would help us spot inconsistencies within a document, as well as between RFCs.
", "time": "2024-03-20T05:34:30Z"}, {"author": "Jean Queralt", "text": "Alexandr +1
", "time": "2024-03-20T05:34:54Z"}, {"author": "Stephen McQuistin", "text": "picking out the RFC2119 keyword constraints would be interesting for that
", "time": "2024-03-20T05:36:03Z"}, {"author": "Nick Sullivan", "text": "Finding contradictions in different texts is unfortunately not a strength of current LLMs, in my experience.
", "time": "2024-03-20T05:39:36Z"}, {"author": "Benson Muite", "text": "Probably some work on use of LLMs and AI in analyzing legal texts could be leveraged to find ways to search for contradictions.
", "time": "2024-03-20T05:40:42Z"}, {"author": "Alexandr Railean", "text": "Nick, is any RFCGPT-related source code public? If one were to try and rebuild it using a fully open stack, would they have to start from scratch or could they leverage some of your current code? If yes, what is the license?
", "time": "2024-03-20T05:42:38Z"}, {"author": "Tara Tarakiyee", "text": "I think the moment you start expecting LLMs to be analytical you start running into lots of pitfalls
", "time": "2024-03-20T05:44:44Z"}, {"author": "Ignacio Castro", "text": "wrt contradictions: at the moment and due to halluzinations, answers that can be manually verified easily are a good fit for LLMS, but those that require a very expert understanding to be verified are more challenging. Current LLMs are good for tedious tasks thatcan be easily verified
", "time": "2024-03-20T05:45:06Z"}, {"author": "Chris Box", "text": "For some reason https://dnsrf.org/research/isto--internet-standards-observatory/tracker/index.html isn't showing any results. Probably user error.
", "time": "2024-03-20T05:45:26Z"}, {"author": "Alexandr Railean", "text": "In their current form - yes, but these systems will evolve and get better. A trick I apply is to ask it for a specific citation about a claim it makes, and then check it myself.
", "time": "2024-03-20T05:45:43Z"}, {"author": "Ignacio Castro", "text": "+1 I think a big question is what will be the role of LLMs in the IETF (and other places) in 5 years
", "time": "2024-03-20T05:46:39Z"}, {"author": "Chris Box", "text": "Just says zero contributions.
", "time": "2024-03-20T05:47:34Z"}, {"author": "Ignacio Castro", "text": "yes, same for ITU
", "time": "2024-03-20T05:48:21Z"}, {"author": "Dhruv Dhody", "text": "If these are mainly internet drafts calling them under standards tracker is incorrect!
", "time": "2024-03-20T05:49:22Z"}, {"author": "Tara Tarakiyee", "text": "All this LLM bandwagoning definitely won't help solve the IETF's carbon footprint issue.
", "time": "2024-03-20T05:52:30Z"}, {"author": "Alexandr Railean", "text": "Tara Tarakiyee said:
\n\n\nAll this LLM bandwagoning definitely won't help solve the IETF's carbon footprint issue.
\n
Hardware will get more efficient, enabling use-cases that are impossible (or not sustainable) today.
", "time": "2024-03-20T05:53:51Z"}, {"author": "Ignacio Castro", "text": "wrt Jonathan question about manual work. We have been doing research analysing mallist and drafts and yes, lots of manual work to map peiople to their multiple emails, affiliations and merge mutitple names of people and affiliations that hae been introduced with variations or typos...
", "time": "2024-03-20T06:00:40Z"}, {"author": "Ignacio Castro", "text": "(maybe 40% of the work is on data cleaning in fact https://sodestream.github.io/category/publications.html)
", "time": "2024-03-20T06:02:22Z"}, {"author": "Tara Tarakiyee", "text": "mic please
", "time": "2024-03-20T06:13:03Z"}, {"author": "Stephen McQuistin", "text": "the comment from Lars (not at the mic) was that none of the outliers are IETF drafts
", "time": "2024-03-20T06:13:31Z"}, {"author": "Alexandr Railean", "text": "Even if development happens on Github, the documents have to be submitted to the datatracker anyway, no?
", "time": "2024-03-20T06:16:44Z"}, {"author": "Chris Box", "text": "Yes, but individual commits are likely to be batched into a later draft update, so lower frequency of draft issuance.
", "time": "2024-03-20T06:18:11Z"}, {"author": "Gorry Fairhurst", "text": "Well, in the WGs I read, there's a lot of PRs in github and the draft from some of those activities is pushed when the editors/authors feel its time to update the WG. That is likely a pattern that will show up.
", "time": "2024-03-20T06:18:39Z"}, {"author": "Gorry Fairhurst", "text": "interesting that the framing here is that delay == bad, or at least something that can/needs to be fixed
", "time": "2024-03-20T06:19:24Z"}, {"author": "Ignacio Castro", "text": "(here is the original blogpost https://blog.apnic.net/2020/08/25/opinion-the-making-of-an-rfc-in-todays-ietf/)
", "time": "2024-03-20T06:19:56Z"}, {"author": "Ignacio Castro", "text": "(and here is more related analysis on RFC time to publish, # drafts, # mails/participant... https://sodestream.github.io/publications/characterising-ietf-rfc-deployment.pdf)
", "time": "2024-03-20T06:21:05Z"}, {"author": "Tara Tarakiyee", "text": "Name of speaker at the mic?
", "time": "2024-03-20T06:21:50Z"}, {"author": "Stephen McQuistin", "text": "Colin Perkins
", "time": "2024-03-20T06:21:56Z"}, {"author": "Tara Tarakiyee", "text": "Thanks
", "time": "2024-03-20T06:22:01Z"}, {"author": "Chris Box", "text": "Ouch - implies it'd be a good idea to submit your initial -00 now if you'd like it to be an RFC in 2032.
", "time": "2024-03-20T06:25:30Z"}, {"author": "Jonathan Hoyland", "text": "So what the main takeaway is, is that there is still hope for my 5 year old draft?
", "time": "2024-03-20T06:27:05Z"}, {"author": "Ignacio Castro", "text": "it would be interesting to see what whether time-to-publish is lower for drafts in new areas with no backwards compatibility issues: ie, is it the process that has become worst or simply it is difficult to make a contribution in a crowded space
", "time": "2024-03-20T06:27:13Z"}, {"author": "Chris Box", "text": "Definitely still hope.
", "time": "2024-03-20T06:28:18Z"}, {"author": "Gorry Fairhurst", "text": "Security/privacy implications are becoming importanta cross a wide range of WGs ... that's actually good.
", "time": "2024-03-20T06:29:03Z"}, {"author": "Tara Tarakiyee", "text": "Imagine if the IETF took the time to think more about the impact of their RFCs on society or if more people were able to participate in the discussions. =]
", "time": "2024-03-20T06:33:13Z"}]