GPT-4o: incredible and terrifying
From my archive, written 13 May 2024. These are my impressions at the time.
A new ChatGPT model is released: GPT 4o, at first I had no idea what this was as I only saw it being mentioned in a few TikTok videos, but looking further into it, the first main thing it adds is the ability to be able to directly communicate with the AI, the best way I can describe it is a sort of FaceTime. This ‘FaceTime’ session allows for the AI to directly respond to your environment, facial expressions and in a way quicker time, so quick that it sounds like you’re talking to an actual person, you can also interrupt it, ten years ago this would’ve looked like complete fiction, its incredible, imagine how much this could help blind people for example - asking where objects are around the house or assistance with any other smaller tasks that would otherwise require a person.
It’s also smarter in this model through better understanding of questions by using context more accurately, improved logic and better maths and coding skills. Its pretty interesting, I have access to it since I have a shared subscription with a friend. The next day, I looked more into this 4o update since I thought there had to be more to it than what I’d mentioned and there is; more specifically, it can do things like generate a person in different environments but actually keep the same person, you can generate your idea of your model and actually keep it this time instead of it regenerating a new person every time which is very useful.
Another thing I found fascinating was the ability to generate images from different angles, it gives the example of generating the OpenAI logo from multiple angles whilst the image looks consistent, it also looks like they’ve made text generation in photos a lot better. It’s pretty fascinating stuff and the best/worst part is that Google aren’t even close to catching up, Bing is doing good but I have no idea why Google are Apple have had their fingers up their ass for so long, its been maybe three years since the actual release of GPT and now OpenAI have introduced this incredible live AI FaceTime and what are meant to be the two biggest players have barely done AI, sure Google released Gemini which is fine but for a trillion dollar company, they should be doing way more, Apple are really disappointing too, especially since they have the chance to get rid of that absolutely useless Siri function that only old people use and even then barely so, if they added a new rebranded AI to phones, it’d be incredible.
Every few months something crazy comes out with AI, something better than the last and I really do think we’re in the early stages of something absolutely insane and groundbreaking, I think we’ll see amazing and terrible consequences from this technology that could solve cancer but could also mislead and terrify the general population, the most dangerous weapon if it comes into the wrong hands. And why do I say this, maybe it sounds paranoid but look at these past few years alone; 30th November 2022 marked the release of GPT 3.5 which upon its release was groundbreaking, for the first time people could elaboration on questions rather than looking on Google for pre-made answers, get very specific data that was otherwise hard to find and generate essays, paragraphs, poems, anything about anything that existed online.
Before this model, we had DALL-E, released January 2021, the first ever higher quality AI image generation. Not even two years after the 3.5 model, we not only have a much more complex version of the initial model with amazing problem solving, coding and mathematical skills but we have near life-like image and soon, video generation along with music and now these instant responses from an AI that can see and react to your environment. Two years or less and this is how far we’ve come, its incredible but also terrifying because I wonder constantly where it will end and what kind of world we’ll be living in where people can’t access any piece of information without doubting its validity, how will we distinguish the inevitably life-like AI?
It may seem that it’ll never get to this point but it already has; numerous times now I’ve seen online ads for supplements which’ll show Joe Rogan or someone else discussing it, Facebook is littered with fake AI images which are indistinguishable to old people, scam calls with loved ones voices, these voice cloning, deepfakes and fake content, how would you even begin to explain this to someone 100 years ago and 100 years in the future, what other near incomprehensible examples of this will exist that I wouldn’t even begin to understand?
Everyone expected benefits and don’t get me wrong, there are insane benefits for using this with work but its equally destroying the creative sectors - something people thought would always remain safe when this AI was first launching, but we’ve slowly seem it chip into music, movie effects, ghost writers, photoshops and more, I can only hope we see some other drastic benefits, more specifically I hope in the fields of science, healthcare and space, these machines can sort through hundreds of years of collected information and make connections that would take us decades to see.