As The Wall Street Journal (WSJ) reported late Friday (Dec. 20), there may not be enough data in the world to make the project, known as GPT-5 and code-named Orion, smart enough.
GPT-5 has been in the works for more than 18 months, the report said, and is designed to be a major upgrade for the technology behind OpenAI’s ChatGPT. Sources told WSJ that Microsoft — OpenAI’s biggest investor — had expected to see GPT-5 by midyear.
OpenAI has held at least two large training sessions, each of which involved months of crunching vast amounts of data to make Orion smarter. With each test run, sources close to the project told WSJ, new problems cropped up and the software fell short of the results for which researchers had hoped.
In the best case scenario, the sources said, Orion outperforms OpenAI’s current offerings, but hasn’t progressed enough to justify the massive cost of running the new model. Six months of training run can cost roughly half a billion, WSJ said, based on public and private estimates of various aspects of the training.
PYMNTS has contacted OpenAI for comment but has not yet gotten a reply.
Advertisement: Scroll to Continue
WSJ wrote that since ChatGPT’s launch in 2022, AI companies have promised to unveil technology that could improve and become part of all human life. Analysts project that big tech companies could spend $1 trillion on AI projects in the years to come.
That’s put a lot of expectations on OpenAI, the report said. The company was valued at $157 billion in October, a figure predicated on CEO Sam Altman’s prediction that GPT-5 will mark a “significant leap forward” in a range of subjects and tasks.
PYMNTS wrote earlier this year about the pending release of GPT-5, arguing that it could mark a pivotal moment in AI development, or just be another smaller upgrade. With expectations of significant advancements, experts across various sectors are examining how this next-generation model could affect industries from healthcare to finance.
“For commerce, the implications of a more advanced LLM [large language model] like GPT-5 are vast,” Cache Merrill, the founder and CTO of Zibtek, an AI-based software company, told PYMNTS.
“We could see significant improvements in customer service bots, offering more coherent and contextually appropriate interactions without human intervention. In digital marketing, content generation could become more sophisticated and tailored, enhancing engagement strategies.
“Furthermore, enhanced LLMs could streamline operations such as contract analysis, risk assessment and more by quickly processing and analyzing large volumes of text-based data with a high degree of accuracy,” Merrill added.
