Following the launch of GPT-4o in May, OpenAI has announced that it is now training its "next frontier model," which aims to deliver groundbreaking capabilities in its ongoing pursuit of developing the world’s first Artificial General Intelligence (AGI). AGI, though still hypothetical, refers to a generative compute system capable of performing a wide range of cognitive tasks with human-level precision.
While the journey toward achieving AGI is ongoing, GPT-5 is anticipated to significantly surpass OpenAI's current models. Expected enhancements include superior natural language processing, advanced content generation capabilities, a broader knowledge base, and improved reasoning skills, potentially allowing it to outpace Claude 3.5 Sonnet in benchmark performance.
In a recent interview with Dartmouth Engineering, OpenAI's Chief Technology Officer Mira Murati discussed the evolving capabilities from GPT-4 to GPT-5. "If you look at the trajectory of improvement, systems like GPT-3 represent toddler-level intelligence," Murati explained. "Systems like GPT-4 can be compared to a smart high schooler. In the coming years, we're aiming for Ph.D.-level intelligence in specific tasks. Things are changing and improving quickly." While specific details remain sparse, here’s a summary of what to expect.
Release Timeline for GPT-5
OpenAI has been rapidly progressing in the development of its large language models (LLMs). GPT-4 was introduced on March 14, 2023, just four months after the launch of GPT-3.5 alongside ChatGPT. Although OpenAI has not provided a definitive release date for GPT-5, speculation suggests that it may debut as early as late 2024.
Model Availability
Currently, all three commercially available versions of GPT—3.5, 4, and 4o—can be accessed in the free tier of ChatGPT. Subscribers to ChatGPT Plus benefit from enhanced rate limits while using the latest GPT-4o model, along with access to additional tools, such as the Dall-E image generator. It remains unclear whether GPT-5 will be available for free users upon its eventual launch. Notably, spinoff tools, such as Bing Chat, have already been utilizing the latest models; Bing Chat launched secretly with GPT-4 before its official announcement. A similar scenario could unfold with GPT-5, but time will tell how the rollout occurs.
Will GPT-5 Reach AGI?
It’s unlikely that GPT-5 will achieve AGI. The quest for machines with human-level reasoning has been anticipated since the 1960s, often likened to the dreams of flying cars and cures for cancer. Industry experts have consistently projected AGI to be several years or even decades away. However, following the emergence of ChatGPT in 2022, the generative AI landscape has evolved rapidly, yielding exponential advances over the past few years. Last year, Shane Legg, co-founder and Chief AGI Scientist at Google DeepMind, indicated to Time Magazine a 50% likelihood that AGI will be realized by 2028. Dario Amodei, co-founder and CEO of Anthropic, expressed an even more optimistic view, suggesting that “human-level” AI could arrive within the next two to three years. Meanwhile, OpenAI CEO Sam Altman believes that AGI could potentially be attained within the next five years.
Concerns Surrounding GPT-5
Although development of GPT-5 is well underway, there are growing calls to pause its progress. A petition signed by over a thousand public figures and tech leaders, including notable figures such as Elon Musk, Steve Wozniak, and Andrew Yang, has been circulated, urging a halt to any advancements beyond GPT-4. The focus of this petition is clearly directed at GPT-5, amid rising concerns about the implications of this technology among government authorities and the general public.
Ranking
