A few moments ago, Greg Brockman, President, and Co-Founder of OpenAI presented a developer demo showcasing GPT-4 and some of its capabilities/limitations. OpenAI has been building this technology for over two years, tuning and rebuilding the entire framework several times.

Fault Barrier

We have all already understood that GPT4 will work better, with its 100 trillion parameters against the 165 billion of GPT3, but finally, we have some working examples!

During the demo, Brockman built a Discord Bot that could receive text and an image as input to chatGPT APIs. In the demo, GPT-4 was able to describe an image and even understand what was funny about a comic drawing.

That’s right, one of the most significant advantages of GPT-4 is its increased ability to consume existing content and understand text and images.

What is really great and covers up all his limitations, is that now can handle 32k tokens(words) in input as context. This is a huge improvement. Consider that Chat GPT-3 can handle up to 2048 tokens(words) in input as context.

While the training data for GPT-4 is still from 2021, the ability to provide content means that it can continuously improve and learn from more recent data.

This is crucial, as the model needs to stay up to date with the latest trends and information. For example, he did the integration of the discord chatbot with the new version of the API (which has been released recently) by providing details about the integration as a context input.

This means that you can give the AI a huge context about what you need to do. Greg also fed the AI his tax and payment history and asked chatGPT to do his taxes. The result was Amazing.

With GPT-4 inching closer to its release, it’ll be fascinating to witness the ways in which it’s utilized and the novel applications it’ll inspire. Brockman’s plea to collaborate with these advanced AI models to achieve greater heights is an alluring concept.

The possibilities for GPT-4 and other AI models to advance and reshape numerous industries and fields are immense, and it’s thrilling to think about what the future holds.

All things considered, GPT-4’s proficiency in comprehending both language and images is a noteworthy breakthrough in the AI sector. It’ll pave the way for new uses and possibilities, and it could transform the way we interact with technology. As GPT-4 approaches its release date, we can look forward to even more captivating advancements in this area.