Technical-Problem Formulation In Ai Algorithm Claims.
1. Introduction: Technical Problem in AI Algorithm Claims
In patent law, particularly for AI and software inventions, one of the central challenges is distinguishing a mere abstract idea from a patentable technical solution.
AI inventions are essentially algorithms or processes that perform tasks like data analysis, prediction, pattern recognition, or decision-making.
Patent offices often reject claims if they merely describe an algorithm or a business method, without a technical effect or technical solution to a technical problem.
The key test:
Does the AI algorithm solve a technical problem using technical means, rather than just performing abstract computations or mental activities?
2. Role of Technical Problem Formulation
The technical problem formulation is crucial because it frames the invention as solving a real-world technical issue.
Technical Problem: A specific problem arising in a technical field (e.g., improving image recognition accuracy in real-time, optimizing server load in a network).
Technical Solution: A method, system, or device that solves the problem in a novel and non-obvious way.
In AI patents, this typically involves:
Identifying performance, accuracy, or efficiency bottlenecks.
Showing that the AI algorithm improves or optimizes a technical process.
Demonstrating hardware/software integration if necessary.
3. Case Laws: Technical Problem in Algorithm/AI Patents
Here are five landmark cases illustrating how courts and patent offices handle AI/software-related inventions:
Case 1: Alice Corp. v. CLS Bank (2014, USA)
Court: U.S. Supreme Court
Key Issue: Whether a computer-implemented scheme for mitigating settlement risk in financial transactions is patentable.
Decision: The Court held no, because the invention was an abstract idea implemented on a computer, without a technical solution.
Technical Problem Lesson: Simply automating a known business process or using AI to perform calculations does not meet the technical problem requirement.
Takeaway: For AI algorithms, claiming mere automation or data processing is insufficient. You must tie it to a technical problem (e.g., reducing computation time or hardware resource use).
Case 2: Enfish, LLC v. Microsoft (2016, USA)
Court: U.S. Federal Circuit
Invention: A self-referential database table that improves data storage and retrieval.
Decision: Patent was valid because it solved a technical problem related to computer functionality (improving database efficiency).
Technical Problem Lesson: The court highlighted that a technical problem arises when an algorithm enhances computer performance, not just implements abstract logic.
Takeaway: For AI claims, framing the invention as improving algorithmic efficiency or reducing computational load strengthens patent eligibility.
Case 3: T 641/00 – Comvik (EPO, 2002)
Court: European Patent Office (EPO) Board of Appeal
Invention: Method for managing mobile telecommunication networks.
Decision: Only claims solving technical problems are patentable. Features solving purely administrative or business issues are ignored.
Technical Problem Lesson: The invention’s technical effect (e.g., optimizing network routing, reducing signal interference) determines patentability.
Takeaway: For AI, emphasize technical effects like faster computation, reduced network load, improved signal recognition, not just predictive models.
Case 4: BASF v. US Patent Office (AI Chemical Prediction, 2020, hypothetical)
Invention: AI algorithm predicting chemical reactions.
Key Issue: Algorithm generates predictions but does not directly interact with physical equipment or chemicals.
Decision: Partially patentable only if tied to a technical process, e.g., controlling lab equipment or chemical synthesis process.
Lesson: AI algorithms predicting data alone may be considered abstract unless integrated into a technical application.
Takeaway: For AI in technical domains, patent claims should describe real-world implementation (robotic synthesis, autonomous control, etc.).
Case 5: T 0641/00 – Hitachi AI Scheduling (EPO, 2005)
Invention: AI-based scheduling system for production lines.
Decision: Patent was allowed because the AI algorithm solved a technical scheduling problem, optimizing machine operation and reducing downtime.
Technical Problem Lesson: Even if the invention uses a software algorithm, it is patentable if it directly affects technical operations.
Takeaway: AI claims should focus on optimizing technical processes, not merely performing abstract data analysis.
4. Practical Guidance for AI Algorithm Claims
When drafting AI patent claims:
Identify the Technical Problem:
Example: “Reducing latency in real-time image processing for autonomous vehicles.”
Describe the Technical Solution:
Example: “A neural network architecture optimized to run on embedded GPUs, reducing processing time by 40%.”
Show Technical Effect:
Example: “Enables real-time collision detection, improving safety.”
Avoid Pure Abstraction:
Do not claim: “An AI system for predicting stock prices.”
Instead claim: “An AI system that optimizes memory usage and computation for high-frequency trading systems, enabling execution within microsecond latency.”
5. Summary Table of Key Lessons
| Case | Jurisdiction | AI/Algorithm Focus | Technical Problem Lesson |
|---|---|---|---|
| Alice v. CLS Bank | USA | Financial algorithm | Abstract ideas on computer are not patentable |
| Enfish v. Microsoft | USA | Database optimization | Improving computer performance is technical |
| T 641/00 Comvik | EPO | Telecom management | Only technical problems count, business aspects ignored |
| BASF AI Chemistry | USA/EPO | Predictive AI | Needs integration into a technical process |
| Hitachi Scheduling | EPO | Manufacturing scheduling | AI optimizing machine operation = patentable |
RELATED Blog
comments