What OpenAI just announced

Hi!

With Christmas taking over at this time of year, you might think that AI news would slow as companies start to wind down, saving their big reveals to grab our attention in the New Year. But nope!

Prepare for a stream of OpenAI news as they've just launched their 12 Days of OpenAI, promising new updates, reveals, and demos over the coming fortnight.

Day 1 kicked off with a video from CEO Sam Altman and demos from some of the OpenAI research scientists. So what's been revealed so far?

OpenAI o1 is here

The OpenAI team has been busy behind the scenes incorporating user feedback on the o1-preview model released in September. The full version of o1 is now available to Plus users and outperforms o1-preview across several benchmarks.

One of the highlights of o1-preview was its improved reasoning - the model was designed to take time to think before responding. This resulted in better performance and more accurate responses but came with a slower response time across all types of queries.

With the full o1 release, improvements have been made in both speed and accuracy. Simpler questions now receive quicker responses, while more complex problems still benefit from the model's thoughtful, deliberate approach to problem-solving.

As an added bonus, the full o1 version includes multi-modal input, supporting both text and image inputs. This expanded capability allows for more versatile and comprehensive interactions.

ChatGPT Pro

A new tier has been added to the ChatGPT lineup with the introduction of ChatGPT Pro. With unlimited access to various OpenAI models, it's been designed to tackle tougher questions and challenges that demand advanced computational intelligence.

With the hefty price tag of $200 per month, it certainly isn't for everyone. But for researchers, engineers, and professionals who rely on cutting-edge technology daily, it could be a game-changer.

o1 pro mode 

The Pro tier also comes with a special o1 pro mode - the most intelligent model yet. Designed to tackle complex problems in challenging areas, it outperforms both o1-preview and o1 in various math, coding and science tests. 

There’s a real focus on reliability with this offering. o1 pro mode has been evaluated on a stricter standard than previous versions - needing to solve questions consistently across multiple attempts before it is deemed to have successfully solved a problem. It takes longer to respond but should deliver more reliable and comprehensive answers as a result. 

Pro grants

OpenAI also announced Pro grants, offering free access to 10 medical researchers at leading US institutions. The initiative aims to provide advanced AI tools to researchers working on critical problems, with plans to expand the program to other regions and research domains. 

While the long-term impact remains to be seen, these grants could potentially accelerate research by providing access to high-compute AI models at no cost to the researchers.

Reinforcement fine-tuning

The day 2 reveal gave us a preview of a model customization option that's expected early next year - reinforcement fine-tuning.

The value of this new feature lies in the ability for users to tailor an o1 model to their precise needs and fine-tune responses using their own data sets. 

By grading the model’s responses and providing reference answers, users are able to refine the model’s performance and expand its expertise.

This new capability will be particularly useful for anyone carrying out work in niche areas requiring specialist knowledge - think medical researchers, legal professionals, and developers who need AI that can understand the intricate nuances of their specific field. 

With 10 more days to go it’s going to be interesting to see what else is on the cards. Expectations are high that the long-awaited text-to-video tool, Sora, will feature but who knows what else OpenAI has up its sleeve. The next announcement is due on Monday so watch this space…

OK, speak soon!

Inge
Top Apps Founder

P.S. Check out the Top Apps AI courses here.

There’s a reason 400,000 professionals read this daily.

Join The AI Report, trusted by 400,000+ professionals at Google, Microsoft, and OpenAI. Get daily insights, tools, and strategies to master practical AI skills that drive results.

Sign up now for free and work smarter, not harder.