I think if this technology goes wrong, it can go quite wrong, and we want to be vocal about it. People should be happy that we’re a little bit scared of this. I think people should be a little bit scared, a little bit of you personally. If I said I were not, you should either not trust me or be very unhappy. I mean this job.
In this video, Sam Altman provides a significant update on GPT 5 and other upcoming models that OpenAI is developing. I’ll detail what he discussed in detail, and if there’s one thing that’s certain, it’s that GPT 5 is a possible revolution. Let’s get started.
During a recent interview, OpenAI CEO Sam Altman made a noteworthy appearance and revealed several thrilling advancements, one of which is the highly anticipated GPT 5. The question of whether GPT 5 is real or merely a concept has been answered since GPT 4’s release. The buzz surrounding GPT 5 has been fueled by rumors of its imminent arrival, a testament to the public’s keen interest in OpenAI’s continued progress and their consistent outperformance with each GPT iteration.
Consequently, the prospect of GPT 5 has generated excitement across various industries. Despite some claims suggesting GPT 5 might be currently undergoing training, the idea of a December 2023 release is merely wishful thinking at this point. During a recent MIT event, Sam Altman provided some insights into the GPT 5 situation. According to him, GPT 5 is not currently undergoing training and won’t be for quite some time. Altman clarified that GPT 5 exists as a concept, but OpenAI is not in a rush to develop it.
However, we can interpret his statements more deeply. Although GPT 5 has not actively been trained at the moment, OpenAI might be in the process of deciding the model’s underlying architecture or even collecting data for its training. While the public may not see any immediate work on GPT 5, there could be ongoing preparations behind the scenes. It’s worth noting that GPT 4 finished its training in August 2022 but was only released slightly over three months ago in March 2023. This indicates that the development and release timelines may not always align perfectly.
We are hinting at the possibility that GPT 5 might have already completed its training or be very close to doing so. However, OpenAI could be deliberately waiting for the right moment to introduce it to the public. Although I cannot confirm this as a fact, it remains a plausible scenario.
Just a few months ago, prominent tech figures signed an open letter urging a pause in training systems more intelligent than GPT 4. Signatories included Elon Musk and Steve Wozniak, co-founder of Apple. Naturally, if GPT 5 were to be released, it would likely attract considerable attention and scrutiny. As a result, OpenAI is proceeding with caution.
Regarding Sam Altman’s response to the letter, he suggested that it lacked technical nuance and failed to specify precisely what needed to be stopped. Nevertheless, the decision on this matter rests with the tech leaders.
So, considering the indications that GPT 5 is likely under development and Sam Altman’s remarks about its potential to surpass its predecessors, what key advancements and features can we anticipate from GPT 5 in comparison to GPT 4 and earlier versions?
It is now apparent that GPT 5 is highly likely to be in development. However, a release in 2023 seems doubtful, and even a release in 2024 appears unlikely. Altman mentioned focusing on GPT 4 to make improvements and fulfill the promised modality. In his remarks, Sam Altman revealed that GPT 5 would be expected to surpass its predecessors.
We knew GPT 4 was good, and we wanted to release it because we thought it was good. We had already had GPT 3.5 at that point for many months, so we’ve gotten used to something much better. We released ChatGPT with GPT 3.5.
He discussed the progression of the models and what can be anticipated from future versions. Interestingly, in 2019, GPT was not widely known, but with the launch of GPT 2, OpenAI’s advancements continued with GPT 3, which paved the way for the revolutionary GPT 3.5. And it was from this foundation that the groundbreaking GPT 4 was built.
According to Sam, he recognized the significance of GPT 4, but even he didn’t anticipate the level of success it would achieve today. Apparently, Sam Altman does not trust ChatGPT. He said during the event, ‘I trust answers generated by ChatGPT less than anybody else on Earth.’
Altman also talked about India’s position when it comes to AI. He said that it is amazing to watch India embrace AI and that OpenAI plans on building AI startups in India. The OpenAI CEO talking about India’s position when it comes to AI said that the country is at the center of AI innovation. Altman also talked about regulating artificial intelligence and said that it should be done in such a way that important innovation doesn’t get hampered. AI will be the most transformative technology, he said, adding that it is important to put guardrails on open-source projects because they have been evolving exponentially. It will be very difficult for governments, policymakers, and regulators to regulate it, but it is important to ensure innovation is not stopped in the long run by over-regulating.
Today, GPT had a staggering 1 million users in just five days. Building on the success of GPT 3.5, OpenAI launched GPT 4, which blew GPT out of the water. GPT 3.5 was a game-changer, with many companies rushing to implement the model in their own services. But GPT 4 showed human-level performance on most of these academic and professional exams.
Following this, we can expect GPT 5 to be a beast. We’ve already seen what GPT 3.5 has done and what GPT 4 is doing. It’s safe to say we’re in for exciting times ahead with GPT 5. But this also brings us to the issue of emerging capabilities in AI.
Here’s an example: these are two different models, GPT and then a different model by Google. And there’s no difference in the models; they just increase in parameter size. That is, they just get bigger. What are the parameters? It’s just like the number of weights in a matrix. So, it’s just the size; you’re just increasing the scale of a thing. And what you see here is that you ask these AIs to do arithmetic, and they can’t do it. And at some point, boom, they just gain the ability to do arithmetic. Nobody understands why it occurs; it simply does.
In 2018, GPT had no theory of mind, and by 2019, it barely displayed any. However, in 2020, it began developing a strategy level comparable to that of a four-year-old. By January 2022, it had progressed to the level of a seven-year-old, and by November of the same year, it had almost reached the level of a nine-year-old.
Now, here’s the truly unsettling part: we only discovered this newfound capability of AI last month, despite being in existence for two years. It had gone unnoticed until February. It’s important to note that AI’s intelligence and reasoning abilities continue to evolve, regardless of whether we are aware of it or not. This is both fascinating and concerning.
So, as a matter of fact, don’t forget to like, share, and subscribe for more intriguing AI updates and discussions.
