I've been a journalist for more than two decades and have covered all kinds of events and announcements, but a tech company embracing the '12 days of Christmas' as a theme is a new one for me.
Between now and December 20 OpenAI will release at least one new feature, model or concept every weekday. In CEO Sam Altman's own words, the company will "have a livestream with a launch or demo, some big ones and some stocking stuffers." I suspect more stocking filler than a major feature, but time will tell.
So far we've seen the launch of the o1 reasoning model and a new o1 Pro model behind a new $200 per month ChatGPT Pro subscription. Finally, we got something for research scientists on Friday in the form of reinforcement fine-tuning of models.
With some huge announcements still to come over the next two weeks — rumored to include the launch of AI video generator Sora after months of anticipation — we have much to cover, so we're doing it live.
12 Days of OpenAI: Announced so far
- Day 1 saw a rather awkward roundtable live session where Altman and his team announced that the o1 reasoning model is now fully released and no longer in public preview.
- On top of that, there's a new Pro More for advanced scientific research and complex mathematical problem solving that you can get for $200 a month (this also comes with unlimited o1 use and unlimited Advanced Voice).
- Day 2 saw another roundtable, but this time without Altman. The focus was on the power behind the models. It saw the launch of reinforcement fine-tuning for AI models on complex datasets.
12 Days of OpenAI: Predictions
Of course, every announcement can't be a blockbuster. OpenAI themselves say that you can expect new things, "big and small." Let's go through a quick list of what we expect heading into the 12 Days of OpenAI.
- Sora soars: After its apparent leak towards the end of last week, we're most confident that OpenAI's video generator will be one of the blockbuster announcements of the show.
- Advanced search gets more advanced: We've seen rumors and leaks about Voice Engine being able to recreate human voices with just a 15-second recording. This would be a wild announcement if it happens, but I'm only 50/50 on it actually seeing a launch this year. Instead, I'd look more to small but significant tweaks like getting access to live search or being able to look through a camera (such as your webcam or phone snapper).
- Canvas improvements: Finally, Canvas has become one of our favorite writing tools that fuses the assistance of generative AI with your workload in a way that feels slick and actually helpful. We predict some updates coming that we saw subtly hinted towards — like stacking different elements within a chat.
- ChatGPT-tweaks: And finally, turning to ChatGPT itself, we're anticipating some improvements to the capability of OpenAI's bread and butter chatbot — including video analysis.
English (US) ·