GPT4, Midjourney v5 and the devualation of knowledge
It's happened, finally GPT-4 has arrived, and everyone is freaking out.
There was a lot of talk about the release of the new generation model from OpenAI, and now it's here.
The new model is multimodal, which means it can accept images as input and reason about them:
Another characteristic is that it accepts more context. According to OpenAI, "GPT-4 is capable of handling over 25,000 words of text, allowing for use cases like long-form content creation, extended conversations, and document search and analysis."
This is amazing, as one issue with the previous model was its limited context capacity. Now, with this improvement, we can use it for more advanced tasks.
The third improvement is its enhanced reasoning and fewer errors in its responses.
I saw on Twitter that it can build basic games in minutes:

So I tried myself and I can build one myself very quickly.
This is incredibly exciting, and a multitude of possibilities have opened up:




This is another game, really cool:
This raises various questions, such as what will constitute unique knowledge in the future, and what will be considered low-value knowledge?

In other news, Midjourney v5 was released today and it looks awesome:


I tried it myself, and the quality of the images is very impressive. To be honest, I have only tried Stable Diffusion and DALL-E-2 and never Midjourney, so perhaps for this reason, the quality of the images seemed out of this world to me. In contrast to Stable Diffusion Midjourney v5 creates very amazing images with less difficulty.
So, yes, this has been a wild week, and it reminds me of the time when Stabble Difusion and ChatGPT were released in the latter part of last year.