# OpenAI Claims It Just Did The Impossible

## Executive summary

The video discusses the Exponential Time Hypothesis (ETH), a foundational concept in theoretical computer science stating that certain problems (like 3-SAT) are inherently difficult to solve efficiently. The speaker highlights a major claim from OpenAI, which suggests they may have found a method to defeat this exponential barrier for the 3-Satisfiability problem (3-SAT). If verified, this breakthrough would overturn a core assumption of theoretical computer science and could impact optimization problems ranging from logistics routing to resource assignment.

## Key takeaways

- The Exponential Time Hypothesis (ETH): ETH posits that certain problems are 'cursed,' meaning that as the problem size grows, the computational time required to solve them increases exponentially, making them practically unsolvable for large inputs.
- The 3-Satisfiability Problem (3-SAT): 3-SAT is used as a canonical example of a problem believed to be exponentially difficult. The difficulty increases rapidly with the problem size.
- OpenAI's Claimed Breakthrough: OpenAI published a paper claiming to defeat the exponential barrier for 3-SAT, suggesting a sub-exponential bound. The speaker notes that while the proof is 'increasingly convincing,' expert review is not 100% complete.

## Technical details

- 3-Satisfiability Problem (3-SAT): A computational problem used to demonstrate the difficulty of finding solutions. The complexity increases exponentially with the problem size.
- Karp Reduction: A technique where various complex, real-world problems (e.g., efficient routing, assigning radio frequencies, sensor placement) can be mathematically 'molded' or transformed into an instance of the 3-SAT problem. Solving 3-SAT efficiently would potentially solve all these related problems.
- Sub-exponential Bound: The paper claims to provide an algorithm that achieves a sub-exponential bound for 3-SAT, which is significantly better than the previously assumed exponential time complexity.

## Practical implications

- Optimization problems (e.g., logistics, scheduling, resource allocation) that can be reduced to 3-SAT could potentially be solved much faster than previously thought.
- The breakthrough could fundamentally change the theoretical limits of computation for complex combinatorial problems.
- The ability to find not only a solution but also to count the exact number of solutions is a major algorithmic advance.

## Topics

Theoretical Computer Science, Computational Complexity, Satisfiability Problem (SAT), Optimization Algorithms, Artificial Intelligence (AI), OpenAI Research Paper, Lambda GPU Cloud

Source: https://www.youtube.com/watch?v=CWdTBLsbaQs