OpenAI is exploring the concept of “PhD-level AI,” a term that refers to AI models capable of performing tasks requiring doctoral-level expertise, such as conducting advanced research, debugging complex code, and analyzing large datasets.
According to a report from “The Information”, OpenAI is planning to launch several specialized AI “agent” products, including a $20,000 per month tier aimed at supporting PhD-level research, alongside lower-cost agents for high-income knowledge workers and software developers. Although OpenAI has not confirmed these pricing details, the company has previously mentioned the potential of PhD-level AI capabilities.
PhD-Level AI: OpenAI’s Advanced Models and Their High-Priced Potential
The term “PhD-level AI” stems from models that excel in benchmark tests, where OpenAI’s o1 series models, for example, showed performance comparable to human PhD students in tasks like coding, math, and science. Additionally, the company’s Deep Research tool, designed to generate research papers, performed well in comprehensive tests. More recently, OpenAI introduced its o3 models, which feature a “private chain of thought” approach, mimicking the iterative problem-solving process of human researchers.
These advanced models have achieved impressive scores on various benchmarks, including a near-perfect performance on the 2024 American Invitational Mathematics Exam and groundbreaking results on the ARC-AGI visual reasoning benchmark. For $20,000 a month, OpenAI suggests customers would be purchasing significant “thinking time” for the AI to work on difficult problems. This high price point reflects the model’s ability to tackle complex tasks, which could be applied in fields like medical research or climate modelling.
PhD-Level AI: High Costs, Performance Concerns, and the Debate Over Value
However, the high costs, especially compared to more affordable AI services like ChatGPT Plus, raise questions about whether the performance justifies the price. While the models show strong benchmark performance, they still face challenges with “confabulations”—plausible but incorrect information. This issue is particularly concerning in research contexts, where accuracy is paramount. The pricing also draws comparisons to the cost of hiring actual PhD students, with some critics pointing out that many top PhD students are not paid as much as the proposed AI subscription fees.
Ultimately, while these AI models may offer significant computational power, questions remain about whether they can replicate the creative thinking and original research that define true doctoral-level work. As technology improves, however, costs may decrease, making these advanced systems more accessible.
