Day 2 of Shipmas: Announcements for Developers Today

[vb]

Yesterday was all about ChatGPT, but today, it’s time to bring some fresh updates to the Developer Community!

Join us for community fun and discussions about today’s announcements and presentations during the live event.

Here is the link to the stream:

12 Days of OpenAI: Day 2

From the description:

Join Mark Chen, SVP of OpenAI Research, Justin Reese, Computational Researcher in Environmental Genomics and Systems Biology, Berkeley Lab, and some team members from OpenAI as they demo and discuss Reinforcement Fine-Tuning.

Please note that commenting on YouTube will not be available. Feel free to share your impressions here instead.

The stream will go live at 2024-12-06T18:00:00Z (the time should automatically adjust to your device’s time zone).

Shout-out to @grandell1234 for sharing extra info

[platypus]

So…what’s in the rumour mill today? “Computer use”? o1 API? Sora?

[vb]

My bet is on improved usage tiers.
But that wouldn’t be enough for a whole event, would it?

[grandell1234]

There are two roads they might take with the Sora release:

• Release it near the end to build up suspense.
• Release it near the start to build up excitement for everything else.

---

Computer use with better API access / user tiers would be cool too see today.

[grandell1234]

The description says:

Join Mark Chen, SVP of OpenAI Research, Justin Reese, Computational Researcher in Environmental Genomics and Systems Biology, Berkeley Lab, and some team members from OpenAI as they demo and discuss Reinforcement Fine-Tuning.

It looks like there are going to be some fine-tuning updates.

Here is the stream link: https://www.youtube.com/watch?v=fMJMhBFa_Gc

[platypus]

Wow now that is super interesting! Our very own o1s!

[merefield]

yes, could do with some refinement, for sure!

[PaulBellow]

Reinforcement fine-tuning…

[PaulBellow]

They should have pulled a Zuck and put a bottle of BBQ sauce on the shelf behind them to get it to go viral lol…

[mitchell_d00]

-drinks sauce - mmm this is what us humans… I mean we like right?

[PaulBellow]

Smaller, faster, cheaper is good!

1,100 sources for dataset isn’t that many!

[mitchell_d00]

@PaulBellow

Smaller, faster, cheaper is good

So is BBQ sauce lol

[Torres2k]

I don’t think that any of this really matters for developers until they give the api o1 acces. Why would any of this matter if you can’t connect this to your ide via the API?

[PaulBellow]

Welcome to the community, @Torres2k !

I don’t know for certain, but I’m sure o1 will be in the API sooner rather than later. It’s a good revenue source they won’t leave on the table.

[mitchell_d00]

It really don’t, it is OpenAI ChatGPT set of basic tools. Not every user is a developer obviously, but giving users advanced tools at hello bridges that gap.

[PaulBellow]

Props to OpenAI for making fine-tuning so relatively easy…

[platypus]

Yes, having experimented RL about 7 years ago - it is NOT easy to get right!

[Torres2k]

I thought that it would come out to day since Altman said that they would release something for developers. But, I can’t really see any developer benefiting from this properly without the API .

[Torres2k]

But don’t get me wrong, the RL is really impressive.

[PaulBellow]

@Torres2k

Altman said that they would release something for developers

Ten more days of Shipmas to go!

I hear you, though, in that o1 API would’ve been cool to have first. I’m sure they spent a lot of time thinking about how to release everything and in what order.

@Torres2k

But don’t get me wrong, the RL is really impressive.

No problem! How did you find us?

ETA: Okay, it looks like Reinforcement is coming early next year…