Generative AI Has Hit A Ceiling, GPT-5 May Not Surpass Its Predecessor: Bill Gates

Nonetheless, Gates also emphasised the short-term potential of AI, particularly in enhancing reliability and comprehensibility through ongoing research.

Gates acknowledged the expense of training LLMs and the ongoing high costs associated with computing power and semiconductor hardware.

Source : Getty

In a recent interview with the German business newspaper Handelsblatt, Microsoft co-founder Bill Gates shared his views on the current state of generative AI and its future prospects. The past year has witnessed a surge of interest and investment in generative AI technologies by numerous companies. Notably, OpenAI's launch of ChatGPT in November 2022 marked a pivotal moment in the evolution of artificial intelligence, following the groundbreaking GPT series.

While the GPT series, or Generative Pre-trained Transformer, has significantly influenced AI development across various industries, Gates expressed reservations about the trajectory of this technology. He suggested that generative AI may have reached a plateau, although he acknowledged the possibility of being mistaken.

Contrary to OpenAI's outlook on GPT-5, Gates believes that current generative AI has hit a ceiling. He characterised the advancement from GPT-2 to GPT-4 as "remarkable."

Gates went on to predict that over the next two to five years, AI software's accuracy will increase substantially while costs decrease, opening doors to the creation of new and reliable applications. Despite anticipating initial stagnation in development, he believes that GPT-4 represents a limit, and he doubts that GPT-5 will surpass its predecessor.

Nonetheless, Gates also emphasised the short-term potential of AI, particularly in enhancing reliability and comprehensibility through ongoing research. He envisioned that developing nations would benefit significantly from AI, pointing to the example of health advice delivered via smartphones.

Regarding the cost and reliability of AI, Gates acknowledged the expense of training large language models (LLMs) and the ongoing high costs associated with computing power and semiconductor hardware. He highlighted that while some AI chips from Nvidia could cost up to $30,000 each, actual usage costs have decreased over time, now estimated at approximately three cents per query.

Gates also stressed the potential of AI in healthcare, particularly in accelerating drug and vaccine development. Despite existing reliability issues, he sees AI as integral to progress in these fields.

Discussing the concept of AI as a "black box," with limited insight into its inner workings, Gates noted that efforts were already underway to unravel its mysteries. As for Artificial General Intelligence (AGI), he expressed uncertainty about its arrival, acknowledging its potential to profoundly impact humanity.

Gates also touched on the role of AI in mitigating climate change, citing improvements in climate models and the potential for new crops to withstand high temperatures. He disclosed his investments in nearly 100 companies working on AI-driven enhancements to the power grid, emphasising the critical role of AI in addressing environmental challenges.