This version of the Generative Pre-trained Transformer has been tipped to arrive this winter and bring OpenAI’s technology to the next level.
According to Siqi Chen, CEO at AI video experts Runway, GPT-5 could see the chatbot achieve so-called Artificial General Intelligence (AGI). That means the already-uncannily chatty AI could have human-level ability to comprehend and perform tasks. As Chen notes, even if there is just debate as to whether GPT-5 reaches AGI level, then by definition it will have.
It would also not be a leap to suggest that it could pass the Turing test, which would be a landmark achievement.
i have been told that gpt5 is scheduled to complete training this december and that openai expects it to achieve agi.which means we will all hotly debate as to whether it actually achieves agi.which means it will.March 27, 2023
What is AGI?
Although ChatGPT is already smarter than many of us on a Monday morning, for a GPT-5 trained ChatGPT model to achieve Artificial General Intelligence it will need to meet a few more criteria. Although obviously nothing has reached AGI before so it’s tricky to know exactly what to expect.
To be qualified as an AGI, ChatGPT will need to show an ability to learn and understand to the same level as a human. Currently ChatGPT needs its own specific prompts to work best but AGI will be able to show initiative and consider context too.
Using an early AGI tool, some AI experts such as Yohei Nakajima of Venture Capital firm Untapped have provided the world with some examples of AGI-level AI at work. They set the AI an objective of starting and growing a business and instructed it that its first task was to figure out what its first task is. The AI then searched the internet for ‘How to start a business’ and had the capacity to realize the results were too broad, declaring ‘No good search results found’.
Eventually, it narrowed down to ‘How to start a small business’ and learnt the requirements of the task. From there the AI set itself its own tasks to create a business plan, develop a marketing strategy, and more, and then started diving deeper into each element.
Huh, well that's starting to work... 🤔Kind of just goes on... and on... building and reprioritizing a task list for itself... pic.twitter.com/whcsSNdBHzMarch 26, 2023
On paper, AGI sounds like a Terminator-style science fiction nightmare (Google Bard has plenty of Sci-fi opinions already) but it would help eliminate the vagueness of chatbots like the new Bing and Google Bard that holds them back. There will, of course, be valid concerns around the potential for disinformation and biases to effect the public's use of AI but this is still an exciting development.