Sauers @Sauers_

@DubuqueMat35988 @xdotli @FutureHouseSF yeah, not sure it's relevant to this though. never heard of Raphidiopterans before now

@thomasgauthierc good idea!

The models were not told their names or any other information about them–not even which was released first. The Claude 3 Opus web system prompt was there, which does not mention model numbers. The initial conversation covered a range of topics, like what a computer program is, images of nature, discussions on if AI models have qualia, the design of this experiment, and practice for answering the "which model are you" questions in the proper format.

@jon_vs_moloch the correct answer

My girlfriend has started prompting ChatGPT to "use truesight" when asking questions

@xeophon @TimothyKassis @xdotli @FutureHouseSF it's probably not real, it looks like they got this number by having a low-quality agent doing shallow literature search and if whatever model they used didn't understand the reasoning or find the answer, it says contradicted

@xdotli @FutureHouseSF textbook

@xdotli @FutureHouseSF ChatGPT finds one

Their agent didn't find anything that contradicts it, and said contradicted: The provided answer states that adults feed on nectar, but none of the excerpts record nectar feeding by any Raphidiopteran. Gillott2005 explicitly notes that adult snakeflies are diurnal predators whose primary diet consists of soft-bodied arthropods (aphids and caterpillars) with only incidental consumption of pollen, not nectar (Gillott2005theremainingendopterygote pages 4-6). Machado2018 describes Raphidiidae adults as “arboreal predators” with no reference to nectar feeding, and while it does mention that Inocelliidae adults might take pollen in captivity, there is no indication they consume nectar (machado2018biodiversityofthe pages 30-32). Similarly, Jepson2010 confirms that while some captive individuals have been observed ingesting pollen, wildfire records and gut content analyses consistently emphasize a predatory (aphid-focused) diet with no mention of nectar (jepson2010neuropteridaofthe pages 51-54). Furthermore, additional excerpts consistently show that Raphidiopterans are not associated with nectar feeding. For example, the discussions in Machado2018 (pages 7-9) note that while neuropterid adults may occasionally be observed on flowers, such instances do not establish nectar as a significant or recorded component of snakefly diets. Instead, these insects are almost exclusively characterized as predators with incidental pollen ingestion in captivity, a behavior that is not synonymous with intentional nectar feeding. The rationale provided in the answer, which argues that records of nectar feeding exist for snakeflies and that options involving Māhoe pollen, Karamū leaf tissue, and Totara Aphids are dismissed on biogeographical grounds (New Zealand endemism), is problematic. None of the extracted sources mentions any observation of snakeflies feeding on nectar. On the contrary, the documented feeding behaviors pertain solely to predation on small arthropods and occasional pollen consumption. Thus, the answer “Nectar” is directly contradictory to the available evidence. No source in the provided context documents nectar feeding by Raphidiopterans; instead, all relevant studies consistently emphasize a predatory mode of feeding with occasional pollen consumption. This falsifies the claim made in the answer and indicates that the response is not accurate. (Gillott2005theremainingendopterygote pages 4-6, machado2018biodiversityofthe pages 30-32, jepson2010neuropteridaofthe pages 51-54)

@xdotli @FutureHouseSF cc @andrewwhite01 fyi

They say "We were unable to find good sources for Question 2’s claim of snakeflies feeding on nectar." "Maybe someone saw a Raphidiopterans eat nectar once, which is extremely out of character, and recorded it somewhere in a way that makes keyword search impossible." However it's very easy to find sources for this! Here's a screenshot of the first result in Google Books. Also the first result if you simply Google "Raphidiopterans" "Nectar" says the same thing

@xdotli @FutureHouseSF I may be biased, idk if others used this strat. but the HLE questions have a selection effect in that they are questions that models got wrong at the time, so it makes sense that a reviewer model also thinks they are wrong

@xdotli @FutureHouseSF hmm not sure about this methodology. "directly conflicting with published evidence" doesn't mean wrong. I designed some of the biology questions specifically to conflict with published literature, as the models tended to repeat false claims that are supported by literature

@ryanpirl least appetizing descriptions

@portforward21 Haha pre-guardrails GPT-4 was great

OpenAI: the "user created a loop that repeatedly called a model." "the model could easily tell it was also controlled by an automated system of some kind." "the model began to exhibit 'fed up' behavior" Apparently this has happened "a few times"

@agitbackprop Need this immediately

@DeadMeme5441 You're homophobic?

@noah_vandal Point cloud

"As the emotional tenor of a story changes, LLM activations trace out meandering paths along the manifold of emotions. How do we know this? In addition to asking the LLM about emotions verbally, as we did in the previous section, we also harvest the internal activations from the last token of each sentence in a story (without asking it anything). These activations serve as a snapshot of what the model represents after reading the story so far. To model the geometry of these LLM representations, we fit a manifold to these activations. In the demo below, we show how stories trace out trajectories along this representation manifold."

goodfire.ai/research/stori…

@glintstoneshard Yes.