Resolves YES if GPT-5 is publicly released before Jan 1, 2025; resolves NO otherwise.
Update from March 18: If the next OpenAI flagship model deviates from the GPT-N naming scheme and there will clearly be no future model named GPT-5 that is going to be released from OpenAI (e.g. they completely changed the trajectory of their naming scheme) then I will count it as "GPT-5" regardless.
However, if the next flagship model is not called GPT-5 but still conforms to the OpenAI GPT-N naming scheme (e.g. GPT-4.5) or it is implied that there will eventually be a model called GPT-5 that will be released (i.e. they are going to continue with the same naming scheme) then I will not count it as "GPT-5".
Related questions
@chrisjbillington I found some tweets saying that they said /by/ early next year in the Stargate article behind the paywall. I don't trust those tweets though. I think it's a summary of a summary of a summary.
@chrisjbillington Ive seen summaries of this saying it includes an early next year timeframe for openais next big ai upgrade https://www.theinformation.com/articles/microsoft-and-openai-plot-100-billion-stargate-ai-supercomputer
but I never actually read the article because paywall so if anyone pays for theinformation and can read the source please let us know
I wanted to share my thinking on why I think gpt-5 *won't* be coming out this year.
I would love to get any feedback from people that disagree. It's very possible I may be overlooking some things, especially given the market seems to overwhelmingly think otherwise.
1. GPT-5 is confirmed to be a big step up from GPT-4 (“I expect that the delta between [GPT-]5 and 4 will be the same that was between 4 and 3 and I think it is our job to live a *few years* in the future and remember that the tools we have now are going to kind of suck, looking backward at them, and that’s how we make sure the future is better.” & https://youtu.be/jvqFAi7vkBc?t=4008 -- note the "what you'd expect from a gpt-5")
2. A more capable model = more opportunity for misuse & unintended consequences. Also leading it to take more time to red team as there is a greater "space" of possibilities that need to be convered.
3. Sam & OpenAI both know that to get to AGI & continue to lead they need institutional support (i.e. support from 'elites', government, media, financiers etc.)
4. Therefore, they can't release it before the November election. Even if the odds are low, they cannot under any circumstances risk some PR-disaster like Facebook had post-2016. The risk-reward is not worth it as, among other things, they will lose institutional support, pave the way for regulatory overreach & allow for a less-tainted competitor to take their place
5. Sam has stated (https://youtu.be/jvqFAi7vkBc?t=3874) that the general public, institutions etc. need "time to adapt" & their goal is "not to surprise people". This aligns with OpenAI's strategy with Sora (show the capabilities but don't release it to the public immediately)
5. If they announce & demo it pre-election they will run the risk of having it be overly politcized (AI is of course already politicized, but prior to an election there is a far greater incentive for politicians & the public to overreact)
Therefore, they I think they will unveil & demo it post-election & release it to the general public much later on (like they have with sora)
Which means release = sometime in 2025, not before
@elf All they really need now is to release a model that outperforms the competition in the next few months. GPT-4.5 would do that without potentially introducing any new political or safety risks.
That segment in the Fridman interview where Altman talks about the non-zero chance of being assassinated speaks to #4. Personal risk mitigation must be a driving factor for decisions made by an org like OpenAI and right now I can't think of a greater risk accelerator than crossing the doomer and political news cycle streams.
All they really need now is to release a model that outperforms the competition in the next few months. GPT-4.5 would do that without potentially introducing any new political or safety risks.
🎯
Exactly, and I think GPT4.5 is exactly what the BusinessInsider article is referring to that’s targeting a Summer release.
Interestingly, ArsTechnica received this response from OpenAI:
imo, this seems to further strengthen our argument.
Will be interesting to see how it all plays out over the coming months!
GPT-5 or GPT-4's successor supposedly will come out this summer.
Source: https://www.businessinsider.com/openai-launch-better-gpt-5-chatbot-2024-3
tldr:
- OpenAI is expected to release GPT-5, the next major version of its language model powering ChatGPT, around mid-2024, likely during the summer.
- Early feedback from enterprise customers who have seen demos of GPT-5 indicates it is "materially better" than previous versions, with new capabilities like calling AI agents for autonomous tasks.
- GPT-5 is still in training and will undergo safety testing and "red teaming" before release, which could delay the launch timeline.
@T38f3
This sounds like GPT 4.5, not GPT 5.
Sam on the Lex episode a few days ago said he "had no idea" when GPT-5 would be releasing.
But he did say there are "many different things" people will see over the coming months that are not GPT-5 & mid-2024 sounds right in line with one of those non-gpt5 "things".
Timestamp: https://youtu.be/jvqFAi7vkBc?t=4011
I may be misunderstanding as I haven’t seen the full Lex episode yet, but does this confirm GPT-5 will be coming just not this year? @VictorLJZ
@elf Oh man I guess I will have to listen to Lex 😒. Gizmodo summarized this conversation in a way that makes this market seem much more promising.
https://gizmodo.com/sam-altman-also-thinks-chatgpt-kinda-sucks-openai-1851345956
Ok some clarifications on the resolution criteria: If the next OpenAI flagship model deviates from the GPT-N naming scheme and there will clearly be no future model named GPT-5 that is going to be released from OpenAI (e.g. they completely changed the trajectory of their naming scheme) then I will count it as "GPT-5" regardless.
However, if the next flagship model is not called GPT-5 but still conforms to the OpenAI GPT-N naming scheme (e.g. GPT-4.5) or it is implied that there will eventually be a model called GPT-5 that will be released (i.e. they are going to continue with the same naming scheme) then I will not count it as "GPT-5".
@VictorLJZ So if this year they release GPT Alfa, but not GPT-5 this resolves YES? Seems like a diversion from your previous clarification year ago:
not the new version of GPT, GPT-5 specifically. if they call it 4.5 or whatever it will not count for this market.
@Weezing if they release GPT Alfa and will not be releasing any model called GPT-5 in the future then I will count it as GPT-5; if they release GPT-4.5 I will not count it as GPT-5.
@VictorLJZ What if they plan to release gpt5 , but it is not publicly known at the time of gpt alfa release?
Do they have to say they will stop naming models in the gpt-N format for the yes resolution?
@Weezing I will likely postpone this market resolution until we have definitive details regarding their naming decisions
@VictorLJZ I wish this had been in the description. So GPT 4.5 or GPT 4 max or something like that would count as well?