When is GPT4 coming out, and what is the fastest way to get early access? Please provide step-by-step instructions. I heard it will be “out next week”. Is this a beta?
I have heard they have been rushing work on it in collaboration with Bing/Microsoft, which is realistic given this came around the same time that Microsoft increased their stake in OpenAI LP to 49% and got an agreement that promises Microsoft 100% of all profits up until a certain ‘payoff’ milestone is reached.
Given that Microsoft’s original stake was to quite some degree not bought with cash, but with ‘tokens’ for runtime on Azure, and that this upgraded deal had even further links with Azure, I’m absolutely certain that a large part of Microsoft’s side of the deal was a LOT of runtime on Azure to train and test the models. Naturally, this would radically speed up the development time for GPT4 since almost all of the main cost and investment of any LLM is in the multi-million dollar runtimes for each single build or test run (many of which have to be ‘thrown back’ and started over).
It really really is worth noting the many, many statements by the head of OpenAI though that people are going to be massively disappointed by what GPT4 actually is, in terms of increased potential, and that it will not live up to the hype. Yes, it is trained on a far, far larger set of data, and that will (hopefully) make it more granular, better at the fine detail of patterns, just like a higher resolution image has more points of data, but the level of detail is not the sole function or attribute of any creation. Once above a point of acceptable clarity, having twice the resolution doesn’t necessarily make the result twice as good. There’s a lot of stuff made for old TV standard 240 resolution that is miles and miles better than some of the modern stuff coming out in 4k today. Same principle applies.
3 Likes
Andreas Braun, the CTO of Microsoft Germany, couldn’t contain his enthusiasm and leaked some information about the upcoming release.
GPT 4.0 is expected to take things to a whole new level. According to Braun, GPT 4.0 will be “multimodal,” meaning it will be capable of generating not only text but also videos and other types of media. The possibilities are endless!
Braun also promised that GPT 4.0 will be a game changer, on par with the iPhone’s moment. This is huge news, as we may be witnessing a turning point in the AI industry.
2 Likes
I caught that too - about the multi-modal output. In a way, GPT was already a more ‘general purpose’ AI than many chatbots, and OpenAI have stressed their focus is on AGI (Artificial General Intelligence) rather than just AI (which is often built around one specific use or purpose, such as many of the different AI systems that have been a part of Google’s spam detection, query rewriting, and the ability to adapt to sudden changes in meaning or context to a given search term.
GPT has for a long time, perhaps all along, been focused on its ability to understand the prompt, the user input, whatever its purpose. It could generate code samples, even fix code snippets, edit text, create narratives, and far more. So in a way I always thought of it as rather multi-modal anyway, just not in the multi-media sense.
I think the ‘game changer’ turning point already happened though. It happened with the release of ChatGPT as a publicly accessible and usable thing when DeepMind and other AI producers were limiting the exposure of AI and the access to it. OpenAI broke ranks with all those who were concerned about the security aspects, or long-term consequences, and for those reasons had held back until either studies were done on those effects, or until they were a lot more certain about how to prevent incidents of misuse.
Releasing ChatGPT was like opening Pandora’s Box - whatever happened afterward, there was no going back. The very definition of a turning point. The world changed, and would never be able to unopen that box. It is now going to take some years for us to fully understand what has been unleashed, and what the long term consequences are.
With ChatGPT, suddenly relatively small groups of activists, such as those Russians who are said to have influenced and interfereed with both the US Presidential Elections, and with the Brexit campaigns, back when they were limited to who could type and use Google translate, may be able to use prompts to produce at least 100 times as much effect per hour. Or worse, pedophile rings can have GPT play the role of children to catfish other children. And we have no reliable detection systems and even less reliable legislation. Can a human even be charged for what an AI writes? What will be the loopholes and flaws in current regulations, and what horrors will have to occur for us to find out?
If GPT4 is as multi-modal as suggested, able to create video as easily as GPT3 created text, what deepfake video footage will become not just possible, but commonplace?
Don’t misunderstand me. I have always adored the idea of AI, and I am genuinely excited by all the positive possibilities. But history has always taught us that a lot of abuse and misuse, sometimes over centuries, has to occur first. These are most definitely “interesting times” of the kind that old Chinese curse suggests. 
2 Likes
I agree @Ammon that ChatGPT release was the turning point.
It was also the “Iphone Moment” per original quote by the Nvidia CEO just Feb 14, but got little press in Germany.
Yet another good quote stolen/repurpose by a German manager, I’ve witnessed that for decades and just last week one other promoted an Einstein quote as his to 1000s of followers.
Anyways, Pandorra’s box is open and cannot be unopened. It happened.
Now we get
- APIs, already happened
- Lower prices, already happend at 1/10th for ChatGPT
- Faster speed
- More output modals, like video
Mass communication will disappear and everyone will have it’s “own netflix”, but that’s all just scaling up what we see here.
The Iphone moment was November 2022.
We’re now in the “stellar adoption growth” phase already.
1 Like
I´m really hyped about the future. As you both said, Pandoras Box is open and the humanity is in a new role now to embrace and use this technology.
Sure also people will use the technology for bad purposes, but what is different then with all the big technologies before: Now everybody with a phone has the chance to use it and form his own dreams in every kind of way.
2 Likes
It’s here
I am excited to be part of this movement! I can’t wait til the live stream.
1 Like
@aiprm-christophc I don’t see that GPT4 option in the dropdown yet, but I red that they are converting every “plus” users to see that.
I’m on Plus and it’s just shown up in my options for New Chat. I’m on here now because I can’t find any info on its supposed new funtionality…
I recommend to read the originally linked posts, I just spent an hour reading thru all the material and examples.
Thanks Christoph - at the moment I’m trying to work out why the image-recognition isn’t working. It can’t see the image if I post the link and I don’t see a special image-button. I see there is a livestream shortly so I’ll watch that.
1 Like
When you read their announcement you know that image recognition isn’t avail to public yet.
1 Like
Hmm, I missed that bit, sorry. Is it on here? I find this interface so hard to navigate, noisy.
I posted the links to it twice above now
Yes, I did skim them (not a top-level geek!) - and now see it in a small sentence. If there’s not to be howls of protest from thousands like me, I suggest it’s made very clear up front!