Are you guys excited for it? What are you going to test out first?
If you could ask openai one question during the live stream, what would it be?
Are you guys excited for it? What are you going to test out first?
If you could ask openai one question during the live stream, what would it be?
there are also rumors for o4-mini, but nothing confirmed yet (and it seems a bit early, doesn’t it?)
if I could ask openai one question, it would be on in terms of the recent patents and articles on neuromorphic chips. there is some amazing stuff out there that could certainly outperform GPUs!
EDIT: seems like o4-mini will be talked about too!
ok, its live! and I updated the topic’s description!
ooooh gdb is there!!! this must be really big!!!
man, as someone that has coded with him, I just love gdb’s code
they are releasing o4-mini and o3!!! lets goooo!
Very exciting. Another great release by OpenAI.
Can’t wait to try it out. The combination of imagery and text is… Insane. Robotics when?
Interesting to see it write up the “brute-force” or “linear-thought” solution and then simplify/clean it up.
Benchmarks are looking good
Indeed, chagpt nor the api has been updated yet.
I estimated 2 years ago that we’d see robots in 7 years, so maybe 5 years from now? I know OAI has been hiring people in robotics for few months (maybe a year?) and then there is figure1, but who knows! hopefully soon and under 5-10k usd
o4-mini getting 99.5 on AIME 2025 is absolutely bonkers, it’s hard to even understand how amazing this is compared to the best model oai had 1 year ago
ohhhhhhhhh daaang!! CodeX CLI is making a comeback
wait, this can’t be it, right? GitHub - microsoft/Codex-CLI: CLI tool that uses Codex to turn natural language commands into their Bash/ZShell/PowerShell equivalents
Crazzaayyyy
open source as well!
Crazy. We can replicate interfaces from a photo. Oh my LORD the live video input->ASCII was INSANE.
The moat is gone
So many assumed evergreen paradigms are being destroyed.
Here’s the link:
wow, turning the webcam in asci.. in what? 30s? that… I’m out of words
You forget March 2023? Greg’s web site mock-up
To web page by GPT-4
(the Discord script demo never worked step-by-step in a released GPT-4…)
How about today? o4-mini
, medium
We’re talking about the invention of stoves and you’re showing us photos of campfires! OpenAI has been cooking with gas!
Nice
very nice… how do we get around this? seems odd Type: invalid_request_error, Message: 400 Previous
response cannot be used for this organization due to Zero Data Retention..
Please verify your settings and try again.
use /model
to switch the model to something you actually have available (check platform.openai.com/playground, see what models you have there) and make sure you have export OPENAI_API_KEY=sk-...
With a higher understanding of what you report:
See if I am correct.
Here’s what you can consider:
Etc.
Zero Data Retention would implicitly be that: extra mechanisms that rely on retaining data without direct “delete” control would be disallowed.
It takes specific effort to get this policy approved for an organization.
So everything works beyond stateful reuse of response IDs for one particular model, it would seem that OpenAI has some false trigger - either in setting up the API endpoint, or data on your account that only applies to one model.
I would see what self-help you can do: issue a new API key for the project, never entering limits, and test. Create a new project with “all” permissions and test a key from there.