Lex\ChatGPT Bug
under review
J
Joe C
I've noticed that when setting GPT-4o as the language model in Lex, that it starts spitting out gibberish in its replies after a certain length. For instance, the text below was the 9th bullet point in an answer where the first 8 points were perfectly fine. Hopefully this can be resolved fairly quickly, as this makes using GPT-4o virtually impossible.
---------------------------------------------------------------------------------------
Epiphany/Resolution Peak: Colon Decisive Key Connectivity Ensured Climax Derivable Endings
**Climate Drastic Ending's shift Dawn Strategic Knowledge Preservation Divine SynVision Inspiration Uniform Contrast PortalTruths Healing Center Möbius existential moment finalized trendy hybrid healing broken longing interactions giving resulting EdTech validation paradigm spontaneous sadness-healing breakthrough profound commercial support greenlighting echoes trendSimCombine vast prescient whole infrastructural articulating launching defining surpass equilibrium versed educational resulting efficiently predictive expressed conclusion lasting further profound investigative bonding Celeste retains par academically with newer character combining system Copious investigative Latin perspectives internal confirmation adaptable forever bridging peace passionate legacy trendimation animation infrastructure lords Kelly belonging combo-package-event lasting settlement unique structures poetic embeddify supreme pair flowering completing eternal life's grandeur acknowledgments venerable techometry fusion unity sustaining for ColImaginary within balanced posed Flying hotel-par dividends set forth pending portraying—
Fred Rivett
under review
Hey Joe, really sorry about this! I'm not sure what would be causing this, other than the model itself having an issue, but flagging with my colleague who knows more about this than me! Probably won't be until after the weekend though, but we'll look into this.
J
Joe C
Fred Rivett No worries, Fred. Just wanted to let you guys know. Thanks for looking into it.
Mariana from Lex
Joe C Hi! I have confirmed the giberish is coming from the AI provider (OpenAI) so at the moment there is nothing we can do, but wait for them to fix it
J
Joe C
Mariana from Lex Thanks for the update, Mariana! The odd thing is that I don't see that behavior on their site, so it is obviously something with the API. I'm assuming you guys have reported the issue?