GPT-5
Article
GPT-5 is a recurring brand in the Astral Codex Ten archive, appearing 3 times across 3 issues between July 17, 2023 and November 20, 2025. The archive places it in contexts such as “then a bunch with GPT-5”; “a forecasting AI built off GPT-5 or 6 might get only small improvements”; “I have trouble not saying “please” and “thank you” to GPT-5 when it answers my questions”. It most often appears alongside OpenAI, GPT-4, AI consciousness.
Metadata
- Category: Brands
- Mention count: 3
- Issue count: 3
- First seen: July 17, 2023
- Last seen: November 20, 2025
Appears In
Related Pages
-
- OpenAI (3 shared issues)
-
- GPT-4 (2 shared issues)
-
- AI consciousness (1 shared issues)
-
- AlphaGo (1 shared issues)
-
- Anthropic (1 shared issues)
-
- Asterisk (1 shared issues)
-
- Bard (1 shared issues)
-
- Berkeley (1 shared issues)
-
- Bitcoin (1 shared issues)
-
- ChatGPT (1 shared issues)
-
- Claude (1 shared issues)
-
- Claude (1 shared issues)
External Links
Source Context
Recovered passages from the original issue text. When the raw archive preserved outbound links inside the source passage, they are listed directly under the quote.
One good thing about order-following AI is that it’s useful now, when AIs aren’t agentic enough to have real goals and we just want to use them as tools in commercial applications. The hope is that we do this a bunch with GPT-4, then a bunch with GPT-5, and so on, and by the time we have a real superintelligence, we’ve worked out some of the kinks. I’m not sure how Musk’s maximally-curious AI helps do office work, which means there’s going to be more of a disconnect between current easily-tested applications and the eventual superintelligence that we need to get right.
Are these the data I’ve been trying to get for years - which forecasting platforms beat which others? I don’t think so - Metaculus’ good Briar score only means it performs well on Metaculus’ questions, which might be easier or harder than some other platform’s questions. Can we use the Halawi et al AI as a fixed comparison point, since it’s always the same skill level? I’m not sure - it trained on each of these markets for the style of question that’s in each market, so it might be biased. Still, these numbers are all about where I would expect them to be, except maybe Polymarket, which does better than I would have expected. But the crowd still beats the AI, right? Halawi et al object that humans can forecast only when they feel like it - you can bet on a prediction market question you feel confident on, and avoid one you don’t. When they let their AI forecast only on those questions where it’s most likely to do well (eg those with lots of relevant news articles), it very slightly outperforms the human crowd. As AI gets better, will it naturally beat humans in forecasting? Halawi et al say this won’t be trivial. They find a version of their system based off GPT-3.5 is only very slightly worse than the final version built off GPT-4. This suggests a forecasting AI built off GPT-5 or 6 might get only small improvements. The second team is Tetlock et al. They start from the same place as Halawi - out-of-the-box LLMs aren’t good at forecasting. They’re more scathing about this than Halawi was - they argue that out-of-the-box models do worse than predicting 50% for everything (this was close to true of human forecasters in the ACX tournament). Instead of increasing quality, Tetlock increases quantity. He wants to do wisdom of crowds, where the crowd is a bunch of different LLMs. So he gets twelve LLMs - including Bard, GPT, Claude, Mistral, PaLM, LLaMa, some Chinese models I’d never heard of, and a couple of variations on these bases - asks them to predict questions, and averages the results. Remember, you gotta prompt your model with “you are a smart person”, or else it won’t be smart! The results: Next, we compare the LLM crowd performance to that of the human crowd for our second hypothesis, directly putting the two crowd-aggregation mechanisms head-to-head. To do this, we use the same LLM crowd average as before (taking the median LLM prediction on each question and averaging up the Brier scores across questions). We compare this to the average of median human predictions on the same questions. In our preregistered analysis, we fail to find statistically significant differences between the LLM crowd’s mean Brier score of M=0.20 (SD=0.12) and that of the human crowd, M=0.19 (SD=0.19), t(60) = 0.19, p = 0.850 Their study was much smaller than Halawi’s (31 questions vs. 3,672), so I don’t think this result (nonsignificant small difference) should be considered different from Halawi’s (significant small difference). Still, it’s weird, isn’t it? Halawi used a really complicated tower of prompts and APIs and fine-tunings, and Tetlock just got more LLMs, and they both did about the same. I have two questions after reading these results: Did they actually do the same, or is this just a function of the small sample size in Tetlock and the non-head-to-head comparison?
I never had a Tamagotchi, but I had stuffed animals as a kid. I’ve outgrown them, but I haven’t thrown them out - it would feel like a betrayal. Offer me $1000 to tear them apart limb by limb in some horrible-looking way, and I wouldn’t do it. Relatedly, I have trouble not saying “please” and “thank you” to GPT-5 when it answers my questions.
The argument against: AI companies have an incentive to make AIs that seem conscious and humanlike, insofar as people will feel more comfortable interacting with them. But they have an opposite incentive to make AIs that don’t seem too conscious and humanlike, lest customers start feeling uncomfortable (I just want to generate slop, not navigate social interaction with someone who has their own hopes and dreams and might be secretly judging my prompts). So if a product seems too conscious, the companies will step back and re-engineer it until it doesn’t. This has already happened: in its quest for user engagement, OpenAI made GPT-4o unusually personable; when thousands of people started going psychotic and calling it their boyfriend, the company replaced it with the more clinical GPT-5. In practice it hasn’t been too hard to find a sweet spot between “so mechanical that customers don’t like it” and “so human that customers try to date it”. They’ll continue to aim at this sweet spot, and continue to mostly succeed in hitting it.
(wait, what is a GPT “instance” in this context, anyway? Do we think of “the weights” as a conscious being, such that there is only one GPT-5? Do we think of each cluster of GPUs as a conscious being, such that the exact configuration of the cloud has immense moral significance? Again, I predict we ignore all of these questions in favor of whether the AI you are looking at has a simulated face right now.)