Can generative AI solve computer science’s unsolved problem?

Can generative AI solve computer science’s unsolved problem? is profiled by BTW Media because published evidence links it to internet infrastructure, governance, operational dependencies, or market visibility.

• Researchers leverage OpenAI’s GPT-4 to delve into the longstanding question, using a Socratic Method to engage the AI in nuanced discussions.
• The study suggests that large language models like GPT-4 can uncover novel insights, offering prospects for significant discoveries in various fields.
• Researchers aim to demonstrate that P does not equal NP by guiding GPT-4 through multiple iterations, employing personas and intricate prompts to explore the mathematics behind the conjecture.

Also read: What is the purpose of a supercomputer?

Researchers utilise OpenAI’s GPT-4 to delve into the P vs. NP debate, suggesting AI’s potential to uncover groundbreaking discoveries.

What role can AI play in resolving the P vs. NP dilemma?

Is P equal to NP? Posed nearly 50 years ago, this question delves deep into the capabilities of computers, yet despite decades of scrutiny, it remains unanswered. Now, generative AI joins the quest.

In their study titled “Large Language Model for Science: A Study on P vs. NP,” lead author Qingxiu Dong and colleagues harness OpenAI‘s GPT-4 large language model. Using what they term the Socratic Method, they engage GPT-4 in multiple chat interactions.

Also read: Samsung integrates Google’s generative AI in S24 series

How might large language models shape future scientific inquiry?

Dong et al. assert that their work demonstrates how large language models can uncover fresh insights, potentially leading to scientific breakthroughs — a concept they dub “LLMs for Science.”

Across 97 prompt iterations, the authors guide GPT-4 through detailed inquiries into the intricacies of P = NP, prefacing each prompt with a contextual statement to guide GPT-4’s responses. Employing personas, such as “wise philosopher” or “mathematician skilled in probability theory,” they coax GPT-4 to adopt specific roles.

Their tactic involves leading GPT-4 to disprove the equality of P and NP. They do this by initially assuming equality, presenting an example, and then revealing its flaws — a method known as proof by contradiction.