Welcome to a special“ what did I miss this summer? ” edition of Hashtag Trending. I’m your host, Jim Love. Let’s get into it.
I started out following one story about what might have been, finally, an announcement of GPT 5 from OpenAI.
I ended up going down the rabbit hole and reflecting on the summer’s stories about AI which, fairly or not, focused on OpenAI and the hope for that next big development.
There were two blockbuster stories this summer. CrowdStrike brought the Windows world to its knees on the cybersecurity front and Open AI dominated the chatter on the AI front.
In fact OpenAI has mastered the art of keeping our attention. You would think that with all the drama of firing their CEO and then bringing him back, they would have used up all their drama points, but they have found a way to keep themselves in the forefront of the news.
One of their tactics was tried and true.
First, keep yourself in the leadership by announcing new projects without a clear date. In fairness, that’s been a strategy for software companies for decades.
But that strategy has a drawback, if you announce too early, by the time you finally do release the new feature, it’s yesterday’s news. And sometimes, people would compare what was promised and go, “is that all there is?”
Steve Jobs changed all that. Apple became the most secretive of organizations. All you knew is that they would announce something great. And because it was all a secret, you were seeing it for the first time. Jobs was a master of suspense – no matter how sensational the release was, he kept back “one more thing…”
But OpenAI has master both aspects of this. They keep the rumour mill going with more leaks than a sieve in a rainstorm…but there are so many different versions that you never know what to expect.
Logically, we have all been looking for GPT5 – that big next step in AI.
But instead of GPT5, Altman and company released GPT 4o (4 oh) with the o being short for omni. To be fair, it showed us an incredible conversational AI that could not only converse, but which appeared to be able to understand our emotions. A truly multi-modal AI and if anything has ever finally buried the famous Turing test, the inability to distinguish between a human conversation and one with a machine – this did it.
And this made us all forget that OpenAI had unveiled the most incredible photo realistic version of a movie generation software, Sora – what ever happened with that? Rumours are that OpenAI is busy working with Hollywood studios.
Yes, the announcement of 4 oh kept us from realizing we really hadn’t seen Sora available to the rest of us?
So how do you keep that buzz alive?
There have been a ton of rumours over the summer. There was Q Star that was rumoured to be a new height in machine intelligence. Was this finally GPT 5?
There was Altman’s lesson on how intelligence would evolve. AI would first have the intelligence of a child, albeit a child that has read half the internet, but it could only reason, if that were even the right term, at the level of a grade school child – easily fooled.
Later versions would be that of a smart high school student. Was version 4o a demonstration of that level of intelligence? Maybe. But most likely not. To get to that level, you have to be able to solve problems of reasoning.
And this has to be really frustrating for Google – because Google has provably achieved this level of intelligence. Their AI was able to score at the level to win a silver medal in a math competition. I’m Canadian, so anything more than Bronze is incredible. But for those who say, “why not gold” – it solved 4 out of 6 tough math problems which would put it up there with the brightest math whiz in your high school. Or it you want the stats, only 9% of the total population are even rated as “proficient at math.”
So this was a big deal – yet it got almost no attention.
The folks at Google have to be absolutely crazed on this one – they can beat OpenAI at their own game and still get nowhere in terms of recognition.
Why?
We were all following the carrot of OpenAI’s next revelation. I told you they were good at this.
There was Q* and then we started to hear about Strawberry, yes, a release called Strawberry. Strawberry was apparently this new logical AI. And other rumours said that it was going to solve the data problem. With models needing more and more data and OpenAI running into lawsuits for using copyrighted data from novels and news sites, it was the saviour of AI. But that release that we never saw was replaced by Orion.
Supposedly, Orion had been shown to the US government and the pentagon – a rumour made more credible by the appointment of Paul M. Nakasone, a retired U.S. Army General and former director of the National Security Agency (NSA), to OpenAI’s board of directors in June 2024.
Orion hit the rumour mills big time. There was even one blogger who found a picture of the Orion constellation used in the background of one OpenAI presentation. At this point, we were going into the land that the older part of our audience might have experienced when the rumours spread that Paul McCartney of the Beatles had died and was replaced by an Ontario OPP officer.
So where is GPT 5 in all of this? The answer is – nowhere to be seen.
But this week, we finally got an official announcement – sort of.
News went out that Tadao Nagasaki, the CEO of OpenAI Japan, might have announced a release date for GPT-5 in a presentation at the KDDI summit in Japan, a large scale business event organized by a major telecommunications company in Japan.
Except he didn’t announce GPT 5.
He did provide some intriguing information about OpenAI’s next major language model, which he referred to as “GPT-Next”.
At the KDDI Summit 2024, Nagasaki revealed that OpenAI is developing a new AI model called GPT-Next, which is expected to be significantly more powerful than its predecessors.
Specifically, he stated that GPT-Next will be “100 times more powerful than GPT-4”.Key points from Nagasaki’s presentation:
• GPT-Next is slated for release in 2024, according to a timeline shown in his presentation.
• The model will reportedly use a smaller version of OpenAI’s secretive “Project Strawberry”.
• The performance boost is attributed to improvements in the model’s architecture rather than just increased computing power.
It’s important to note that GPT-Next appears to be distinct from GPT-5. While there has been speculation about GPT-5, OpenAI has not officially announced its development or release date.
In fact, OpenAI CEO Sam Altman stated last November that there were still many things to figure out before they would make a model called GPT-5.
The exact relationship between GPT-Next and potential future models like GPT-5 remains unclear. However, Nagasaki’s announcement suggests that OpenAI is making significant strides in AI development and plans to release a highly capable new model in the near future.
So that’s what you missed over the summer.
A lot of stuff that didn’t get released. And a lot that did get released that was overshadowed by the rumours.
A truly great accomplishment from Google than got no press at all. And some huge moves forward in deepfakes and photo realistic and behaviour realistic AI. We did a demo last week using an AI generated host that replaced me. It was one of our most highly rated daily episodes.
A list of other AI developments that have advanced what is available and useful for companies. The list goes on and on.
But still no date for GPT 5, but the promise of even more updates
Where does this leave all of us.
For those of us in industry, we know this one thing. The pace of this evolution of AI is incredible. There are developments out today that we still have not yet absorbed into our business processes and systems. And no matter what you call it, there is a model coming which is potentially 100X what GPT 4 is – and even that is not what GPT 5 will be.
Buckle up.
We’ll try to help you sort out the fact from fiction and the hype from the “have to start using this.”
In that light, we’re going to rerun the practical AI episode we ran over the labour day weekend. My rational is that’s it’s a great episode featuring two executives from financial services and medical firms, one of the brightest technical experts in open systems and system architecture than I know and a remarkable young woman who is working at the overlap of cybersecurity and AI.
Our stats tell us that our podcast distributer had some difficulties and didn’t release this to the majority of podcast audience. We’ll give them a pass on this one – but to anyone who uses a cloud provider who you can only reach with a service request and you get back answers saying, “no problem on our side” – you’ll identify with how I feel.
