GPT-4 is an advanced large language model (LLM) released by OpenAI in 2023. It was published on Tuesday 14 March.
Read the OpenAI article on the release here.
- The use of GPT-4 is similar to ChatGPT
- GPT-4 will be able to follow instructions more closely
- GPT-4 can process images (not produce them)
Where can GPT-4 be used?
According to OpenAI, GPT-4 will first be made available through ChatGPT Plus. The monthly fee is around 25 dollars.
What is the difference between ChatGPT and GPT-4?
OpenAI has previously introduced two main models to the public: GPT-3 and ChatGPT.
- GPT-3 is a text generator
- ChatGPT is a chatbot
GPT-4 is, at least according to the first experiments, a new evolution of ChatGPT, but not a complete revolution. Of course, the fact that AI can produce text for different purposes better than most humans is revolutionary enough.
What did I expect from GPT-4?
The jumps in past models have been so shocking that expectations for the GPT-4 were a little too high.
My guess was that GPT-4 would be something between GPT-3 and ChatGPT: a dialogue-based AI tool capable of producing content in text, image and voice. Multimodal output was not yet introduced in GPT-4.
These are the ones I was waiting for:
- Faster response generation
- Taking a broader context into account
- Multimodality: sound, image and moving image
Multimodality is included in the new model, but in such a way that it can also be fed in as an image. From a generative AI perspective, this may not be as shocking as being able to get a response with, say, animation, but it still opens up a whole new range of applications.
What can GPT-4 do?
GPT-4 continues ChatGPT’s world conquest. It writes code, articles, translates and writes poetry.
At least in my experience, generative AI is at its best in tasks where you have some background, but not full mastery.
Here are some examples where I have used it myself:
- Writing code snippets
- Writing functions in Google Sheets
- Ideating on the just little bit outside your expertise
When was GPT-4 released?
As originally reported by the German Heise magazine, German Microsoft CTO Andreas Braun said at the event that GPT-4 will be released on on the week starting 13th of March. And that’s the way it went down.
It was also speculated that the announcement would take place at Microsoft’s “Reinventing Productivity with AI” event on 16th March. This did not happen, but the announcement was made on Tuesday 14 March.
Where can you find out more about the potential of artificial intelligence?
If you want to experience the possibilities of AI, I recommend my Aum Golly series of AI poetry books: the first part written by GPT-3 and the second by ChatGPT.
Leave a Reply