Alex UsedToGoToVenice

@mattsheffield@mastodon.social por_que_no_los_dos.gif

@fraggle @jonny did same with Copilot; same results This has been sending me all day; implication that these things are *never* "looking" at images is ... Not great given the account of medical care and etcetc being out into them

@hannah qualtrics floated this (and maybe went through with it idk) right after chatgpt first went live What hilarious is that this agentic things is plausibly *worse* since you're having simulations fill out the survey instead instead of using an llm to generate a bazillion responses. The responses will still be just as bad, but at a huge markup of energy costs

@jonny@neuromatch.social [laugh_cry.gif]

@jonny [nods] yes, a strong prerequisite for the deeper ways

@ingorohlfing I think it's still feasible; boils down to being a time-stamped and public lab notebook Reg 1: "we don't know the exact shape of the data yet, but we think it's gonna have X, Y, and Z. We're gonna do [Plan] if true" Reg 2: "The data had X and Y but not Z; need change [Plan]" I think this applies to the registered report case, too. Diff fields are gonna have diff norms for how much you oughta know about your data columns a priori, but within that context can make judgments fine

@ingorohlfing like, go Bob Ross with this shit; be a little creative with these tools Doing model development? Do a two-stage registration where stage 1 is "here're the general principals I'm gonna follow to build the model"; stage 2 is "here's how that went + deviations; here's the model and how we're gonna do confirmatory/validity testing"; final report is how *that* went

@satrevik@fediscience.org fair about B not really being a stronger test per se - I think the longer I'm a program officer the more I'm doubling down on the need for sensitivity analysis / real hesitant about putting much weight on any single NHST, pre-specofors or not. Hence "I think B is a bit stronger of a signal for [reasons], but it wasn't an a priori called shot and A limits the strength of certainty for [other reasons]" is how i'd deal with it

@satrevik@fediscience.org Report both; lend more weight in discussion/interpretation to B But this likely isn't a clear-cut finding if A and B diverge Edit: I might have responded to the old wording, but am not sure!

@dstephenlindsay @siminevazire

@Orb2069 @anthracite @cstross ah, yes, reminding me of the paraphilia module in my abnormal psych course in undergrad The pictorial record of the consequences of a vacuum fetish being projected onto a 25ft tall screen will forever stick with me Sadly

@oberman @caal are you going to have an old man standing up and yelling out made up numbers or an Elsevier VP going after an audience member for having a normal reaction to her baldface lies or gormless awe at coal-burning plagiarism machines the true markers of open science

@jonny what was the second layer

@JudithterSchure The Employer doesn't do open reviews for awards applications. We provide feedback to applicants at several stages of the process, but otherwise the raw materials are kept confidential There's a qualitative difference in the level of legal scrutiny that goes on with awards for tens of millions of USD vs. manuscript publication. I'm ambivalent about the closed nature of the process, but net of everything I get it now that I'm on the other side of things

@phpete@mastodon.coffee @cstross@wandering.shop @Phosphenes@mastodon.social I work with so many of them People who genuinely believe sales pitches and hype; eagerly and proudly post copilot slop into teams Really truly feel like companies have our best interests at heart

@phpete@mastodon.coffee @cstross@wandering.shop @Phosphenes@mastodon.social 🥲🤜🏻🤛🏻🥲

@cmconseils@mastodon.social #JustTwengeThings