Showing posts with label KIPO. Show all posts
Showing posts with label KIPO. Show all posts

Saturday, September 6, 2025

Patentability of LLM Prompts: Overcoming Abstract Idea Rejections

 

Can a Simple Command to an AI Be Patented? This article provides an in-depth analysis of how LLM prompt techniques can transcend mere ‘ideas’ to be recognized as concrete ‘technical inventions,’ exploring key strategies and legal standards across different countries.

It seems that almost no one around us thinks of protecting prompt techniques or prompts that instruct LLM models with patents. At first, I was also skeptical, wondering, ‘Can a simple command to a computer be patented?’ However, as I delved deeper into this topic, I came to the conclusion that it is entirely possible if certain conditions are met. This article is a summary of the thought process I went through, and please bear in mind that it may not yet be an academically established view. ๐Ÿ˜Š

 

๐Ÿค” Prompts Aren’t Patentable Because They’re Just ‘Human Thoughts,’ Right?

The first hurdle that comes to mind for many is the principle that ‘human mental processes’ are not patentable subject matter. In fact, the argument that “a prompt is fundamentally human involvement, and technology involving such human mental activity is not patentable” is one of the strongest reasons for rejection in patent examination. This standard has been particularly firm since the U.S. Supreme Court’s Alice Corp. v. CLS Bank decision. It means that merely implementing something on a computer that a person could do in their head is not enough to get a patent.

According to this logic, the act of instructing an AI through a prompt is ultimately an expression of human thought, so one could easily conclude that it cannot be patented. However, this argument is half right and half wrong. And this is precisely where our patent strategy begins.

๐Ÿ’ก Good to Know!
What patent law takes issue with as ‘human intervention’ is not the act of giving a command to a system itself. It refers to cases where the core idea of the invention remains at the level of a mental step that can be practically performed in the human mind. Therefore, the key is to prove that our prompt technology transcends this boundary.

 

๐Ÿ“Š A Shift in Perspective: From ‘Command’ to ‘Computer Control Technology’

The first step to unlocking the patentability of prompt technology is to change our perspective. We need to redefine our technology not as ‘a message sent from a human to an AI,’ but as ’a technology that controls the internal computational processes of a complex computer system (LLM) through structured data to solve a technical problem and achieve concrete performance improvements.’

If you take a close look at the algorithm of China’s DeepSeek-R1, you can see that it implements various prompt techniques as they are.

Think about it. The process of assigning a specific expert role to an LLM with billions of parameters, injecting complex library dependency information as context, and combining numerous constraints to control the generation of optimal code is clearly in a realm that ‘cannot practically be performed in the human mind.’ This is a crucial standard for recognizing patent eligibility in the guidelines and case law of the U.S. Patent and Trademark Office (USPTO).

 

๐ŸŒ A Comparative Look at Key Examination Standards of Major Patent Offices

The patentability of prompt technology is not assessed uniformly across all countries. If you are considering international filing, it is crucial to understand the subtle differences in perspective among major patent offices.

1. USPTO (United States Patent and Trademark Office) – Emphasis on the Abstract Idea Exception

The USPTO strictly applies the Alice/Mayo two-step test, which originated from Supreme Court case law. Instructions or general linguistic expressions that merely replace human thought processes can be dismissed as “abstract ideas.” However, if it can be demonstrated that the prompt is linked to a concrete technical implementation (e.g., improving model accuracy, optimizing specific hardware operations), there is a chance of it being recognized as patent-eligible subject matter.

2. EPO (European Patent Office) – Focus on Technical Effect

The EPO assesses based on “technical character” and “technical effect.” Simply presenting data input or linguistic rules is considered to lack inventive step, but if the prompt structure serves as a means to solve a technical problem (e.g., improving computational efficiency, optimizing memory usage, enhancing interaction with a specific device), it can be recognized as patent-eligible.

3. KIPO (Korean Intellectual Property Office) – Emphasis on Substantive Requirements for Software Inventions

KIPO places importance on the traditional requirement of “a creation of a technical idea utilizing the laws of nature.” Therefore, a prompt as a mere sentence or linguistic rule is not considered a technical idea, but if it is shown to be combined with a specific algorithm, hardware, or system to produce a concrete technical result, it can be recognized as an invention. In Korean practice, presenting a concrete system structure or processing flow is particularly persuasive.

Key Comparison Summary

Patent Office Key Requirement
USPTO (U.S.) Emphasis on ‘concrete technical implementation’ to avoid the abstract idea exception
EPO (Europe) Proof of ‘technical effect’ is key; simple data manipulation is insufficient
KIPO (Korea) Must be a technical idea using laws of nature + emphasis on systemic/structural implementation
⚠️ Implications for International Filing
The same “LLM prompt” technology could be at risk of being dismissed as an “abstract business method” in the United States, a “non-technical linguistic rule” in Europe, and a “mere idea” in Korea. Therefore, when considering international filing, a strategy that clearly articulates the ‘concrete system architecture’ and ‘measurable technical effects’ throughout the specification is essential as a common denominator.

 

๐Ÿงฎ A Practical Guide to Drafting Patent Claims (Detailed)

So, how should you draft patent claims to avoid the ‘human intervention’ attack and clearly establish that it is a ‘technical invention’? Let’s take a closer look at four key strategies.

1. Set the subject as the ‘computer (processor),’ not the ‘person.’

This is the most crucial step in shifting the focus of the invention from the ‘user’s mental activity’ to the ‘machine’s technical operation.’ It must be specified that all steps of the claim are performed by computer hardware (processor, memory, etc.).

  • Bad ๐Ÿ‘Ž: A method where a user specifies a persona to an LLM and generates code.
  • Good ๐Ÿ‘: A step where a processor, upon receiving a user’s input, assigns a professional persona for a specific programming language to the LLM.

2. Specify the prompt as ‘structured data.’

Instead of abstract expressions like ‘natural language prompt,’ you need to clarify that it is a concrete data structure processed by the computer. This shows that the invention is not just a simple idea.

  • Bad ๐Ÿ‘Ž: A step of providing a natural language prompt to the LLM.
  • Good ๐Ÿ‘: A step of generating and providing to the LLM a machine-readable context schema that includes library names and version constraints.

3. Claim ‘system performance improvement,’ not the result.

Instead of subjective results like ‘good code,’ you must specify objective and measurable effects that substantially improve the computer’s functionality. This is the core of ‘technical effect.’

  • Bad ๐Ÿ‘Ž: A step of generating optimized code.
  • Good ๐Ÿ‘: A step of controlling the LLM’s token generation probability through the schema to generate optimized code that reduces code compatibility errors and saves GPU memory usage.

4. Clarify the ‘automation’ process.

It should be specified that all processes after the initial input (data structuring, LLM control, result generation, etc.) are performed Automatically by the system without further human judgment, demonstrating that it is a reproducible technical process.

 

๐Ÿ“œ Reinforced Claim Example

By integrating all the strategies described above, you can construct a reinforced patent claim as follows.

[Claim] A computer-implemented method for generating optimized code, comprising:

  1. (a) parsing, by a processor, a user’s natural language input to generate a persona identifier defining an expert role for a specific programming language;
  2. (b) generating, by the processor, by referencing said input and an external code repository, structured context data including library names, version constraints, and hardware memory usage limits;
  3. (c) generating, by the processor, a control prompt including said persona identifier and structured context data and transmitting it to an LLM, thereby automatically controlling the internal token generation process of the LLM;
  4. (d) receiving, from said controlled LLM, optimized code that satisfies said constraints and has a compilation error rate below a predefined threshold and reduced GPU memory usage.

→ This example, instead of focusing on a simple result, greatly increases the chances of patent registration by clarifying system-level measurable technical effects such as ‘reduced compilation error rate’ and ‘reduced GPU memory usage.’

 

Frequently Asked Questions ❓

Q: Can a simple prompt like "write a poem about a cat" be patented?
A: No, that in itself is just an idea and would be difficult to patent. The subject of a patent would be a technical method or system that uses a prompt with a specific data structure (e.g., a schema defining poetic devices, rhyme schemes) to control an LLM to generate a poem, resulting in less computational resource usage or more accurate generation of a specific style of poetry.
Q: What are some specific ‘technical effects’ of prompt technology?
A: Typical examples include reduced compilation error rates in code generation, savings in computational resources like GPU and memory, shorter response generation times, and improved output accuracy for specific data formats (JSON, XML, etc.). The important thing is that these effects must be measurable and reproducible.
Q: Do I need to draft claims differently for each country when filing internationally?
A: Yes, while the core strategy is the same, it is advantageous to tailor the emphasis to the points that each patent office values. For example, in a U.S. (USPTO) specification, you would emphasize the ‘concrete improvement of computer functionality,’ in Europe (EPO), the ‘technical effect through solving a technical problem,’ and in Korea (KIPO), the ‘concreteness of the system configuration and processing flow.’

In conclusion, there is a clear path to protecting AI prompts with patents. However, it requires a strategic approach that goes beyond the idea of ‘what to ask’ and clearly demonstrates ‘how to technically control and improve a computer system.’ I hope this article provides a small clue to turning your innovative ideas into powerful intellectual property. If you have any more questions, feel free to ask in the comments~ ๐Ÿ˜Š

‘์ถ”์ƒ์  ์•„์ด๋””์–ด’ ๊ณต๊ฒฉ์„ ํ”ผํ•˜๋Š” ํ”„๋กฌํ”„ํŠธ ํŠนํ—ˆ ์ „๋žต

 

AI์—๊ฒŒ ๋‚ด๋ฆฌ๋Š” ๋‹จ์ˆœํ•œ ๋ช…๋ น, ํŠนํ—ˆ๊ฐ€ ๋  ์ˆ˜ ์žˆ์„๊นŒ? LLM ํ”„๋กฌํ”„ํŠธ ๊ธฐ์ˆ ์ด ๋‹จ์ˆœํ•œ ‘์•„์ด๋””์–ด’๋ฅผ ๋„˜์–ด ์–ด๋–ป๊ฒŒ ๊ตฌ์ฒด์ ์ธ ‘๊ธฐ์ˆ ์  ๋ฐœ๋ช…’์œผ๋กœ ์ธ์ •๋ฐ›์„ ์ˆ˜ ์žˆ๋Š”์ง€, ๊ทธ ํ•ต์‹ฌ ์ „๋žต๊ณผ ๊ตญ๊ฐ€๋ณ„ ๋ฒ•์  ๊ธฐ์ค€์„ ์‹ฌ์ธต์ ์œผ๋กœ ๋ถ„์„ํ•ฉ๋‹ˆ๋‹ค.

์šฐ๋ฆฌ ์ฃผ๋ณ€์— LLM ๋ชจ๋ธ์—๊ฒŒ ์ง€์‹œํ•˜๋Š” ํ”„๋กฌํ”„ํŠธ ๊ธฐ๋ฒ•์ด๋‚˜ ๊ทธ ํ”„๋กฌํ”„ํŠธ๋ฅผ ํŠนํ—ˆ๋กœ ๋ณดํ˜ธ๋ฐ›์„ ์ƒ๊ฐ์„ ํ•˜๋Š” ์‚ฌ๋žŒ์€ ๊ฑฐ์˜ ์—†๋Š” ๊ฒƒ ๊ฐ™์Šต๋‹ˆ๋‹ค. ์ €๋„ ์ฒ˜์Œ์—๋Š” ‘๋‹จ์ˆœํžˆ ์ปดํ“จํ„ฐ์— ๋‚ด๋ฆฌ๋Š” ๋ช…๋ น์ธ๋ฐ ์ด๊ฒŒ ํŠนํ—ˆ๊ฐ€ ๋ ๊นŒ?’ ํ•˜๋Š” ์˜๊ตฌ์‹ฌ์ด ๋“ค์—ˆ์ฃ . ํ•˜์ง€๋งŒ ์ด ์ฃผ์ œ์— ๋Œ€ํ•ด ๊นŠ์ด ํŒŒ๊ณ ๋“ค๋ฉด์„œ, ํŠน์ • ์กฐ๊ฑด์„ ๋งŒ์กฑํ•  ๊ฒฝ์šฐ ์ถฉ๋ถ„ํžˆ ๊ฐ€๋Šฅํ•˜๋‹ค๋Š” ๊ฒฐ๋ก ์— ์ด๋ฅด๊ฒŒ ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ์ด ๊ธ€์€ ์ œ๊ฐ€ ๋„๋‹ฌํ•œ ์ผ๋ จ์˜ ์‚ฌ๊ณ  ๊ณผ์ •์„ ์ •๋ฆฌํ•œ ๊ฒƒ์ด๋ฉฐ, ์•„์ง ํ•™์ˆ ์ ์œผ๋กœ ํ™•๋ฆฝ๋œ ๊ฒฌํ•ด๋Š” ์•„๋‹ ์ˆ˜ ์žˆ๋‹ค๋Š” ์ ์„ ๊ฐ์•ˆํ•˜๊ณ  ์ฝ์–ด์ฃผ์‹œ๋ฉด ์ข‹๊ฒ ์Šต๋‹ˆ๋‹ค. ๐Ÿ˜Š

 

๐Ÿค” ํ”„๋กฌํ”„ํŠธ, ๋‹จ์ˆœํ•œ ‘์‚ฌ๋žŒ์˜ ์ƒ๊ฐ’์ด๋ผ ํŠนํ—ˆ๊ฐ€ ์•ˆ ๋œ๋‹ค๊ณ ์š”?

๋งŽ์€ ๋ถ„๋“ค์ด ๊ฐ€์žฅ ๋จผ์ € ๋– ์˜ฌ๋ฆฌ๋Š” ์žฅ๋ฒฝ์€ ๋ฐ”๋กœ ‘์ธ๊ฐ„์˜ ์ •์‹ ์  ํ™œ๋™’์€ ํŠนํ—ˆ ๋Œ€์ƒ์ด ์•„๋‹ˆ๋ผ๋Š” ์›์น™์ผ ๊ฒ๋‹ˆ๋‹ค. ์‹ค์ œ๋กœ “ํ”„๋กฌํ”„ํŠธ๋Š” ๊ทผ๋ณธ์ ์œผ๋กœ ์ธ๊ฐ„์˜ ๊ฐœ์ž…์ด๋ฉฐ, ์ด๋Ÿฌํ•œ ์ธ๊ฐ„์˜ ์ •์‹ ์  ํ™œ๋™์ด ๊ฐœ์ž…๋˜๋Š” ๊ธฐ์ˆ ์€ ํŠนํ—ˆ ๋Œ€์ƒ์ด ์•„๋‹ˆ๋‹ค”๋ผ๋Š” ์ฃผ์žฅ์€ ํŠนํ—ˆ ์‹ฌ์‚ฌ์—์„œ ๊ฐ€์žฅ ๊ฐ•๋ ฅํ•œ ๊ฑฐ์ ˆ ์ด์œ  ์ค‘ ํ•˜๋‚˜์ž…๋‹ˆ๋‹ค. ์ด๋Š” ํŠนํžˆ ๋ฏธ๊ตญ ์—ฐ๋ฐฉ๋Œ€๋ฒ•์›์˜ ์•จ๋ฆฌ์Šค ํŒ๋ก€(Alice Corp. v. CLS Bank) ์ดํ›„ ๋”์šฑ ํ™•๊ณ ํ•ด์ง„ ๊ธฐ์ค€์ด์ฃ . ์‚ฌ๋žŒ์ด ๋จธ๋ฆฟ์†์œผ๋กœ ์ƒ๊ฐํ•ด์„œ ํ•  ์ˆ˜ ์žˆ๋Š” ์ผ์„ ๋‹จ์ˆœํžˆ ์ปดํ“จํ„ฐ๋กœ ๊ตฌํ˜„ํ•œ ๊ฒƒ๋งŒ์œผ๋กœ๋Š” ํŠนํ—ˆ๋ฅผ ๋ฐ›์„ ์ˆ˜ ์—†๋‹ค๋Š” ์˜๋ฏธ์ž…๋‹ˆ๋‹ค.

์ด ๋…ผ๋ฆฌ์— ๋”ฐ๋ฅด๋ฉด, ์šฐ๋ฆฌ๊ฐ€ ํ”„๋กฌํ”„ํŠธ๋ฅผ ํ†ตํ•ด AI์—๊ฒŒ ๋ฌด์–ธ๊ฐ€๋ฅผ ์ง€์‹œํ•˜๋Š” ํ–‰์œ„๋Š” ๊ฒฐ๊ตญ ์‚ฌ๋žŒ์˜ ๋จธ๋ฆฟ์† ์ƒ๊ฐ์„ ํ‘œํ˜„ํ•œ ๊ฒƒ์ด๋‹ˆ, ํŠนํ—ˆ๊ฐ€ ๋  ์ˆ˜ ์—†๋‹ค๋Š” ๊ฒฐ๋ก ์— ์‰ฝ๊ฒŒ ๋‹ค๋‹ค๋ฅผ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ํ•˜์ง€๋งŒ ์ด ์ฃผ์žฅ์€ ์ ˆ๋ฐ˜์€ ๋งž๊ณ  ์ ˆ๋ฐ˜์€ ํ‹€๋ฆฝ๋‹ˆ๋‹ค. ๋ฐ”๋กœ ์—ฌ๊ธฐ์„œ๋ถ€ํ„ฐ ์šฐ๋ฆฌ์˜ ํŠนํ—ˆ ํ™•๋ณด ์ „๋žต์ด ์‹œ์ž‘๋ฉ๋‹ˆ๋‹ค.

๐Ÿ’ก ์•Œ์•„๋‘์„ธ์š”!
ํŠนํ—ˆ๋ฒ•์ด ๋ฌธ์ œ ์‚ผ๋Š” ‘์ธ๊ฐ„์˜ ๊ฐœ์ž…’์€ ์‹œ์Šคํ…œ์— ๋ช…๋ น์„ ๋‚ด๋ฆฌ๋Š” ํ–‰์œ„ ์ž์ฒด๊ฐ€ ์•„๋‹™๋‹ˆ๋‹ค. ๋ฐœ๋ช…์˜ ํ•ต์‹ฌ ์•„์ด๋””์–ด๊ฐ€ ์ธ๊ฐ„์˜ ๋จธ๋ฆฟ์†์—์„œ ์‹ค์งˆ์ ์œผ๋กœ ์ˆ˜ํ–‰๋  ์ˆ˜ ์žˆ๋Š” ์ •์‹ ์  ๋‹จ๊ณ„์— ๋จธ๋ฌด๋Š” ๊ฒฝ์šฐ๋ฅผ ์˜๋ฏธํ•ฉ๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์šฐ๋ฆฌ์˜ ํ”„๋กฌํ”„ํŠธ ๊ธฐ์ˆ ์ด ์ด ๊ฒฝ๊ณ„๋ฅผ ๋„˜์–ด์„œ๋Š” ๊ฒƒ์ž„์„ ์ฆ๋ช…ํ•˜๋Š” ๊ฒƒ์ด ํ•ต์‹ฌ์ž…๋‹ˆ๋‹ค.

 

๐Ÿ“Š ๊ด€์ ์˜ ์ „ํ™˜: ‘๋ช…๋ น์–ด’์—์„œ ‘์ปดํ“จํ„ฐ ์ œ์–ด ๊ธฐ์ˆ ’๋กœ

ํ”„๋กฌํ”„ํŠธ ๊ธฐ์ˆ ์˜ ํŠนํ—ˆ ๊ฐ€๋Šฅ์„ฑ์„ ์—ด๊ธฐ ์œ„ํ•œ ์ฒซ๊ฑธ์Œ์€ ๊ด€์ ์„ ๋ฐ”๊พธ๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ์˜ ๊ธฐ์ˆ ์„ ‘์ธ๊ฐ„์ด AI์—๊ฒŒ ๋ณด๋‚ด๋Š” ๋ฉ”์‹œ์ง€’๊ฐ€ ์•„๋‹ˆ๋ผ, ’๋ณต์žกํ•œ ์ปดํ“จํ„ฐ ์‹œ์Šคํ…œ(LLM)์˜ ๋‚ด๋ถ€ ์—ฐ์‚ฐ ๊ณผ์ •์„ ํŠน์ • ๊ตฌ์กฐ์˜ ๋ฐ์ดํ„ฐ๋ฅผ ํ†ตํ•ด ์ œ์–ดํ•˜์—ฌ, ๊ธฐ์ˆ ์  ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ณ  ๊ตฌ์ฒด์ ์ธ ์„ฑ๋Šฅ ํ–ฅ์ƒ์„ ์ด๋Œ์–ด๋‚ด๋Š” ๊ธฐ์ˆ ’๋กœ ์žฌ์ •์˜ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.

์ค‘๊ตญ์˜ DeepSeek-R1์˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ฐ€๋งŒํžˆ ๋“ค์—ฌ๋‹ค๋ณด๋ฉด, ๋‹ค์–‘ํ•œ ํ”„๋กฌํ”„ํŠธ ๊ธฐ๋ฒ•์„ ๊ทธ๋Œ€๋กœ ๊ตฌํ˜„ํ•˜๊ณ  ์žˆ๋‹ค๋Š” ์‚ฌ์‹ค์„ ๋ฐœ๊ฒฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

์ƒ๊ฐํ•ด๋ณด์„ธ์š”. ์ˆ˜์‹ญ์–ต ๊ฐœ์˜ ํŒŒ๋ผ๋ฏธํ„ฐ๋ฅผ ๊ฐ€์ง„ LLM์— ํŠน์ • ์ „๋ฌธ๊ฐ€ ์—ญํ• ์„ ๋ถ€์—ฌํ•˜๊ณ , ๋ณต์žกํ•œ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ ์˜์กด์„ฑ ์ •๋ณด๋ฅผ ์ปจํ…์ŠคํŠธ๋กœ ์ฃผ์ž…ํ•˜๋ฉฐ, ์ˆ˜๋งŽ์€ ์ œ์•ฝ ์กฐ๊ฑด์„ ์กฐํ•ฉํ•ด ์ตœ์ ์˜ ์ฝ”๋“œ๋ฅผ ์ƒ์„ฑํ•˜๋„๋ก ์ œ์–ดํ•˜๋Š” ๊ณผ์ •์€ ๋ช…๋ฐฑํžˆ ‘์ธ๊ฐ„์˜ ์ •์‹  ๋Šฅ๋ ฅ์œผ๋กœ๋Š” ์‹ค์งˆ์ ์œผ๋กœ ์ˆ˜ํ–‰ ๋ถˆ๊ฐ€๋Šฅํ•œ(cannot practically be performed in the human mind)’ ์˜์—ญ์ž…๋‹ˆ๋‹ค. ์ด๋Š” ๋ฏธ๊ตญ ํŠนํ—ˆ์ƒํ‘œ์ฒญ(USPTO)์˜ ๊ฐ€์ด๋“œ๋ผ์ธ์ด๋‚˜ ํŒ๋ก€์—์„œ๋„ ํŠนํ—ˆ ์ ๊ฒฉ์„ฑ์„ ์ธ์ •ํ•˜๋Š” ์ค‘์š”ํ•œ ๊ธฐ์ค€์ด ๋ฉ๋‹ˆ๋‹ค.

 

๐ŸŒ ์ฃผ์š”๊ตญ ํŠนํ—ˆ์ฒญ๋ณ„ ํ•ต์‹ฌ ์‹ฌ์‚ฌ ๊ธฐ์ค€ ๋น„๊ต

ํ”„๋กฌํ”„ํŠธ ๊ธฐ์ˆ ์˜ ํŠนํ—ˆ ๊ฐ€๋Šฅ์„ฑ์€ ๋ชจ๋“  ๋‚˜๋ผ์—์„œ ๋™์ผํ•˜๊ฒŒ ํ‰๊ฐ€๋˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค. ๊ตญ์ œ ์ถœ์›์„ ๊ณ ๋ คํ•œ๋‹ค๋ฉด ์ฃผ์š”๊ตญ ํŠนํ—ˆ์ฒญ์˜ ๋ฏธ๋ฌ˜ํ•œ ์‹œ๊ฐ์ฐจ๋ฅผ ์ดํ•ดํ•˜๋Š” ๊ฒƒ์ด ๋งค์šฐ ์ค‘์š”ํ•ฉ๋‹ˆ๋‹ค.

1. USPTO (๋ฏธ๊ตญ ํŠนํ—ˆ์ฒญ) – ์ถ”์ƒ์  ์•„์ด๋””์–ด ์˜ˆ์™ธ ๊ฐ•์กฐ

๋ฏธ๊ตญ ํŠนํ—ˆ์ฒญ์€ ๋Œ€๋ฒ•์› ํŒ๋ก€์—์„œ ๋น„๋กฏ๋œ Alice/Mayo 2๋‹จ๊ณ„ ํ…Œ์ŠคํŠธ๋ฅผ ์—„๊ฒฉํžˆ ์ ์šฉํ•ฉ๋‹ˆ๋‹ค. ๋‹จ์ˆœํžˆ ์ธ๊ฐ„์˜ ์‚ฌ๊ณ  ๊ณผ์ •์„ ๋Œ€์ฒดํ•˜๋Š” ์ง€์‹œ๋‚˜ ์ผ๋ฐ˜์  ์–ธ์–ด ํ‘œํ˜„์€ “์ถ”์ƒ์  ์•„์ด๋””์–ด”๋กœ ๋ณด์•„ ๋ฐฐ์ œ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ํ”„๋กฌํ”„ํŠธ๊ฐ€ ๊ตฌ์ฒด์ ์ธ ๊ธฐ์ˆ ์  ๊ตฌํ˜„(์˜ˆ: ๋ชจ๋ธ ์ •ํ™•๋„ ๊ฐœ์„ , ํŠน์ • ํ•˜๋“œ์›จ์–ด ์—ฐ์‚ฐ ์ตœ์ ํ™”)์— ์—ฐ๊ฒฐ๋˜์–ด ์žˆ์Œ์„ ์ž…์ฆํ•˜๋ฉด ํŠนํ—ˆ ์ ๊ฒฉ์„ฑ์„ ์ธ์ •๋ฐ›์„ ์—ฌ์ง€๊ฐ€ ์žˆ์Šต๋‹ˆ๋‹ค.

2. EPO (์œ ๋Ÿฝ ํŠนํ—ˆ์ฒญ) – ๊ธฐ์ˆ ์  ํšจ๊ณผ(technical effect) ์ค‘์‹ฌ

์œ ๋Ÿฝ ํŠนํ—ˆ์ฒญ์€ ”๊ธฐ์ˆ ์  ์„ฑ๊ฒฉ” ๋ฐ “๊ธฐ์ˆ ์  ํšจ๊ณผ”๋ฅผ ๊ธฐ์ค€์œผ๋กœ ํŒ๋‹จํ•ฉ๋‹ˆ๋‹ค. ๋‹จ์ˆœํžˆ ๋ฐ์ดํ„ฐ ์ž…๋ ฅ์ด๋‚˜ ์–ธ์–ด ๊ทœ์น™์„ ์ œ์‹œํ•˜๋Š” ์ˆ˜์ค€์€ ๋ฐœ๋ช…์„ฑ์ด ์—†๋‹ค๊ณ  ๋ณด์ง€๋งŒ, ํ”„๋กฌํ”„ํŠธ ๊ตฌ์กฐ๊ฐ€ ๊ธฐ์ˆ ์  ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๋Š” ์ˆ˜๋‹จ(์˜ˆ: ์—ฐ์‚ฐ ํšจ์œจ ๊ฐœ์„ , ๋ฉ”๋ชจ๋ฆฌ ์‚ฌ์šฉ ์ตœ์ ํ™”, ํŠน์ • ์žฅ์น˜์™€์˜ ์ƒํ˜ธ์ž‘์šฉ ๊ฐ•ํ™”)์œผ๋กœ ๊ธฐ๋Šฅํ•œ๋‹ค๋ฉด ํŠนํ—ˆ ์ ๊ฒฉ์„ฑ์„ ์ธ์ •ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

3. KIPO (ํ•œ๊ตญ ํŠนํ—ˆ์ฒญ) – ์†Œํ”„ํŠธ์›จ์–ด ๋ฐœ๋ช…์˜ ์‹ค์ฒด์  ์š”๊ฑด ๊ฐ•์กฐ

ํ•œ๊ตญ ํŠนํ—ˆ์ฒญ์€ “์ž์—ฐ๋ฒ•์น™์„ ์ด์šฉํ•œ ๊ธฐ์ˆ ์  ์‚ฌ์ƒ์˜ ์ฐฝ์ž‘”์ด๋ผ๋Š” ์ „ํ†ต์  ์š”๊ฑด์„ ์ค‘์‹œํ•ฉ๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ๋‹จ์ˆœํžˆ ๋ฌธ์žฅ์ด๋‚˜ ์–ธ์–ด ๊ทœ์น™์œผ๋กœ์„œ์˜ ํ”„๋กฌํ”„ํŠธ๋Š” ๊ธฐ์ˆ ์  ์‚ฌ์ƒ์ด ์•„๋‹ˆ๋ผ๊ณ  ๋ณด์ง€๋งŒ, ํŠน์ • ์•Œ๊ณ ๋ฆฌ์ฆ˜·ํ•˜๋“œ์›จ์–ด·์‹œ์Šคํ…œ๊ณผ ๊ฒฐํ•ฉํ•˜์—ฌ ๊ตฌ์ฒด์ ์ธ ๊ธฐ์ˆ ์  ๊ฒฐ๊ณผ๋ฅผ ๋„์ถœํ•จ์„ ์ž…์ฆํ•˜๋ฉด ๋ฐœ๋ช…์œผ๋กœ ์ธ์ •๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ํŠนํžˆ ํ•œ๊ตญ ์‹ค๋ฌด์—์„œ๋Š” ๊ตฌ์ฒด์  ์‹œ์Šคํ…œ ๊ตฌ์กฐ๋‚˜ ์ฒ˜๋ฆฌ ํ๋ฆ„์„ ํ•จ๊ป˜ ์ œ์‹œํ•ด์•ผ ์„ค๋“๋ ฅ์ด ํฝ๋‹ˆ๋‹ค.

ํ•ต์‹ฌ ๋น„๊ต ์š”์•ฝ

ํŠนํ—ˆ์ฒญ ํ•ต์‹ฌ ์š”๊ฑด
USPTO (๋ฏธ๊ตญ) ์ถ”์ƒ์  ์•„์ด๋””์–ด ์˜ˆ์™ธ๋ฅผ ํ”ผํ•˜๊ธฐ ์œ„ํ•œ ‘๊ตฌ์ฒด์  ๊ธฐ์ˆ  ๊ตฌํ˜„’ ๊ฐ•์กฐ
EPO (์œ ๋Ÿฝ) ‘๊ธฐ์ˆ ์  ํšจ๊ณผ(technical effect)’ ์ž…์ฆ์ด ํ•ต์‹ฌ. ๋‹จ์ˆœ ๋ฐ์ดํ„ฐ ์กฐ์ž‘์€ ๋ถˆ๊ฐ€
KIPO (ํ•œ๊ตญ) ์ž์—ฐ๋ฒ•์น™์„ ์ด์šฉํ•œ ๊ธฐ์ˆ ์  ์‚ฌ์ƒ + ์‹œ์Šคํ…œ/๊ตฌ์กฐ์  ๊ตฌํ˜„ ๊ฐ•์กฐ
⚠️ ๊ตญ์ œ์ถœ์› ์‹œ์‚ฌ์ 
๋™์ผํ•œ “LLM ํ”„๋กฌํ”„ํŠธ” ๊ธฐ์ˆ ์ด๋ผ๋„ ๋ฏธ๊ตญ์—์„œ๋Š” “์ถ”์ƒ์  ๋น„์ฆˆ๋‹ˆ์Šค ๋ฐฉ๋ฒ•”์œผ๋กœ, ์œ ๋Ÿฝ์—์„œ๋Š” “๋น„๊ธฐ์ˆ ์  ์–ธ์–ด ๊ทœ์น™”์œผ๋กœ, ํ•œ๊ตญ์—์„œ๋Š” “๋‹จ์ˆœ ์•„์ด๋””์–ด”๋กœ ๋ฐฐ์ œ๋  ์œ„ํ—˜์ด ๊ฐ๊ฐ ์กด์žฌํ•ฉ๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ๊ตญ์ œ์ถœ์›์„ ๊ณ ๋ คํ•œ๋‹ค๋ฉด, ๊ณตํ†ต๋ถ„๋ชจ์ธ ‘๊ตฌ์ฒด์ ์ธ ์‹œ์Šคํ…œ ์•„ํ‚คํ…์ฒ˜’์™€ ‘์ธก์ • ๊ฐ€๋Šฅํ•œ ๊ธฐ์ˆ ์  ํšจ๊ณผ’๋ฅผ ๋ช…์„ธ์„œ ์ „๋ฐ˜์— ๋ช…ํ™•ํžˆ ๋“œ๋Ÿฌ๋‚ด๋Š” ์ „๋žต์ด ํ•„์ˆ˜์ ์ž…๋‹ˆ๋‹ค.

 

๐Ÿงฎ ํŠนํ—ˆ ์ฒญ๊ตฌํ•ญ ๊ตฌ์„ฑ ์‹ค๋ฌด ๊ฐ€์ด๋“œ (์ƒ์„ธํŽธ)

๊ทธ๋ ‡๋‹ค๋ฉด ‘์ธ๊ฐ„์˜ ๊ฐœ์ž…’์ด๋ผ๋Š” ๊ณต๊ฒฉ์„ ํ”ผํ•˜๊ณ  ‘๊ธฐ์ˆ ์  ๋ฐœ๋ช…’์ž„์„ ๋ช…ํ™•ํžˆ ํ•˜๋ ค๋ฉด ํŠนํ—ˆ ์ฒญ๊ตฌํ•ญ์„ ์–ด๋–ป๊ฒŒ ์ž‘์„ฑํ•ด์•ผ ํ• ๊นŒ์š”? ํ•ต์‹ฌ ์ „๋žต 4๊ฐ€์ง€๋ฅผ ๋” ๊ตฌ์ฒด์ ์œผ๋กœ ์‚ดํŽด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.

1. ์ฃผ์–ด๋ฅผ ‘์‚ฌ๋žŒ’์ด ์•„๋‹Œ ‘์ปดํ“จํ„ฐ(ํ”„๋กœ์„ธ์„œ)’๋กœ ์„ค์ •ํ•˜์„ธ์š”.

์ด๋Š” ๋ฐœ๋ช…์˜ ์ค‘์‹ฌ์„ ‘์‚ฌ์šฉ์ž์˜ ์ •์‹  ํ™œ๋™’์—์„œ ‘๊ธฐ๊ณ„์˜ ๊ธฐ์ˆ ์  ๋™์ž‘’์œผ๋กœ ์˜ฎ๊ธฐ๋Š” ๊ฐ€์žฅ ์ค‘์š”ํ•œ ๋‹จ๊ณ„์ž…๋‹ˆ๋‹ค. ์ฒญ๊ตฌํ•ญ์˜ ๋ชจ๋“  ๋‹จ๊ณ„๊ฐ€ ์ปดํ“จํ„ฐ ํ•˜๋“œ์›จ์–ด(ํ”„๋กœ์„ธ์„œ, ๋ฉ”๋ชจ๋ฆฌ ๋“ฑ)์— ์˜ํ•ด ์ˆ˜ํ–‰๋จ์„ ๋ช…์‹œํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.

  • Bad ๐Ÿ‘Ž: ์‚ฌ์šฉ์ž๊ฐ€ LLM์— ํŽ˜๋ฅด์†Œ๋‚˜๋ฅผ ์ง€์ •ํ•˜๊ณ  ์ฝ”๋“œ๋ฅผ ์ƒ์„ฑํ•˜๋Š” ๋ฐฉ๋ฒ•.
  • Good ๐Ÿ‘: ํ”„๋กœ์„ธ์„œ๊ฐ€(A processor), ์‚ฌ์šฉ์ž์˜ ์ž…๋ ฅ์„ ์ˆ˜์‹ ํ•˜์—ฌ ํŠน์ • ํ”„๋กœ๊ทธ๋ž˜๋ฐ ์–ธ์–ด์˜ ์ „๋ฌธ๊ฐ€ ํŽ˜๋ฅด์†Œ๋‚˜๋ฅผ LLM์— ๋ถ€์—ฌํ•˜๋Š” ๋‹จ๊ณ„.

2. ํ”„๋กฌํ”„ํŠธ๋ฅผ ‘๊ตฌ์กฐํ™”๋œ ๋ฐ์ดํ„ฐ’๋กœ ๊ตฌ์ฒดํ™”ํ•˜์„ธ์š”.

‘์ž์—ฐ์–ด ํ”„๋กฌํ”„ํŠธ’์™€ ๊ฐ™์€ ์ถ”์ƒ์ ์ธ ํ‘œํ˜„ ๋Œ€์‹ , ์ปดํ“จํ„ฐ๊ฐ€ ์ฒ˜๋ฆฌํ•˜๋Š” ๊ตฌ์ฒด์ ์ธ ๋ฐ์ดํ„ฐ ๊ตฌ์กฐ์ž„์„ ๋ช…ํ™•ํžˆ ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ์ด๋Š” ๋ฐœ๋ช…์ด ๋‹จ์ˆœํ•œ ์•„์ด๋””์–ด๊ฐ€ ์•„๋‹˜์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.

  • Bad ๐Ÿ‘Ž: LLM์— ์ž์—ฐ์–ด ํ”„๋กฌํ”„ํŠธ๋ฅผ ์ œ๊ณตํ•˜๋Š” ๋‹จ๊ณ„.
  • Good ๐Ÿ‘: ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ ๋ช…์นญ ๋ฐ ๋ฒ„์ „ ์ œ์•ฝ ์กฐ๊ฑด์„ ํฌํ•จํ•˜๋Š” ๊ธฐ๊ณ„ ํŒ๋… ๊ฐ€๋Šฅํ•œ ์ปจํ…์ŠคํŠธ ์Šคํ‚ค๋งˆ(a machine-readable context schema)๋ฅผ ์ƒ์„ฑํ•˜์—ฌ LLM์— ์ œ๊ณตํ•˜๋Š” ๋‹จ๊ณ„.

3. ๊ฒฐ๊ณผ๋ฌผ์ด ์•„๋‹Œ ‘์‹œ์Šคํ…œ ์„ฑ๋Šฅ ๊ฐœ์„ ’์„ ์ฒญ๊ตฌํ•˜์„ธ์š”.

‘์ข‹์€ ์ฝ”๋“œ’์™€ ๊ฐ™์€ ์ฃผ๊ด€์ ์ธ ๊ฒฐ๊ณผ๋ฌผ์ด ์•„๋‹ˆ๋ผ, ์ปดํ“จํ„ฐ์˜ ๊ธฐ๋Šฅ์„ ์‹ค์งˆ์ ์œผ๋กœ ํ–ฅ์ƒ์‹œํ‚ค๋Š” ๊ฐ๊ด€์ ์ด๊ณ  ์ธก์ • ๊ฐ€๋Šฅํ•œ ํšจ๊ณผ๋ฅผ ๋ช…์‹œํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ์ด๊ฒƒ์ด ๋ฐ”๋กœ ‘๊ธฐ์ˆ ์  ํšจ๊ณผ’์˜ ํ•ต์‹ฌ์ž…๋‹ˆ๋‹ค.

  • Bad ๐Ÿ‘Ž: ์ตœ์ ํ™”๋œ ์ฝ”๋“œ๋ฅผ ์ƒ์„ฑํ•˜๋Š” ๋‹จ๊ณ„.
  • Good ๐Ÿ‘: ์ƒ๊ธฐ ์Šคํ‚ค๋งˆ๋ฅผ ํ†ตํ•ด LLM์˜ ํ† ํฐ ์ƒ์„ฑ ํ™•๋ฅ ์„ ์ œ์–ดํ•˜์—ฌ, ์ฝ”๋“œ ํ˜ธํ™˜์„ฑ ์˜ค๋ฅ˜๋ฅผ ๊ฐ์†Œ์‹œํ‚ค๊ณ  GPU ๋ฉ”๋ชจ๋ฆฌ ์‚ฌ์šฉ๋Ÿ‰์„ ์ ˆ๊ฐํ•˜๋Š” ์ตœ์ ํ™”๋œ ์ฝ”๋“œ๋ฅผ ์ƒ์„ฑํ•˜๋Š” ๋‹จ๊ณ„.

4. ‘์ž๋™ํ™”’ ๊ณผ์ •์„ ๋ช…ํ™•ํžˆ ํ•˜์„ธ์š”.

์ตœ์ดˆ์˜ ์ž…๋ ฅ์„ ์ œ์™ธํ•œ ๋ชจ๋“  ๊ณผ์ •(๋ฐ์ดํ„ฐ ๊ตฌ์กฐํ™”, LLM ์ œ์–ด, ๊ฒฐ๊ณผ ์ƒ์„ฑ ๋“ฑ)์€ ์ธ๊ฐ„์˜ ์ถ”๊ฐ€์ ์ธ ํŒ๋‹จ ์—†์ด ์‹œ์Šคํ…œ์— ์˜ํ•ด ์ž๋™์œผ๋กœ(Automatically) ์ด๋ฃจ์–ด์ง„๋‹ค๋Š” ์ ์„ ๋ช…์‹œํ•˜์—ฌ, ์žฌํ˜„ ๊ฐ€๋Šฅํ•œ ๊ธฐ์ˆ  ํ”„๋กœ์„ธ์Šค์ž„์„ ๋ณด์—ฌ์ค˜์•ผ ํ•ฉ๋‹ˆ๋‹ค.

 

๐Ÿ“œ ๊ฐ•ํ™”๋œ ์ฒญ๊ตฌํ•ญ ์˜ˆ์‹œ

์•ž์„œ ์„ค๋ช…ํ•œ ์ „๋žต๋“ค์„ ๋ชจ๋‘ ํ†ตํ•ฉํ•˜๋ฉด ๋‹ค์Œ๊ณผ ๊ฐ™์ด ๊ฐ•ํ™”๋œ ํŠนํ—ˆ ์ฒญ๊ตฌํ•ญ์„ ๊ตฌ์„ฑํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

[์ฒญ๊ตฌํ•ญ] ์ตœ์ ํ™”๋œ ์ฝ”๋“œ๋ฅผ ์ƒ์„ฑํ•˜๋Š” ์ปดํ“จํ„ฐ ๊ตฌํ˜„ ๋ฐฉ๋ฒ•์œผ๋กœ์„œ,

  1. (a) ํ”„๋กœ์„ธ์„œ๊ฐ€ ์‚ฌ์šฉ์ž์˜ ์ž์—ฐ์–ด ์ž…๋ ฅ์„ ํŒŒ์‹ฑํ•˜์—ฌ, ํŠน์ • ํ”„๋กœ๊ทธ๋ž˜๋ฐ ์–ธ์–ด์˜ ์ „๋ฌธ๊ฐ€ ์—ญํ• ์„ ์ •์˜ํ•˜๋Š” ํŽ˜๋ฅด์†Œ๋‚˜ ์‹๋ณ„์ž๋ฅผ ์ƒ์„ฑํ•˜๋Š” ๋‹จ๊ณ„;
  2. (b) ํ”„๋กœ์„ธ์„œ๊ฐ€ ์ƒ๊ธฐ ์ž…๋ ฅ ๋ฐ ์™ธ๋ถ€ ์ฝ”๋“œ ์ €์žฅ์†Œ๋ฅผ ์ฐธ์กฐํ•˜์—ฌ, ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ ๋ช…์นญ๊ณผ ๋ฒ„์ „ ์ œ์•ฝ์กฐ๊ฑด, ํ•˜๋“œ์›จ์–ด ๋ฉ”๋ชจ๋ฆฌ ์‚ฌ์šฉ ํ•œ๊ณ„์น˜๋ฅผ ํฌํ•จํ•˜๋Š” ๊ตฌ์กฐํ™”๋œ ์ปจํ…์ŠคํŠธ ๋ฐ์ดํ„ฐ(structured context data)๋ฅผ ์ƒ์„ฑํ•˜๋Š” ๋‹จ๊ณ„;
  3. (c) ํ”„๋กœ์„ธ์„œ๊ฐ€ ์ƒ๊ธฐ ํŽ˜๋ฅด์†Œ๋‚˜ ์‹๋ณ„์ž ๋ฐ ๊ตฌ์กฐํ™”๋œ ์ปจํ…์ŠคํŠธ ๋ฐ์ดํ„ฐ๋ฅผ ํฌํ•จํ•˜๋Š” ์ œ์–ด ํ”„๋กฌํ”„ํŠธ๋ฅผ ์ƒ์„ฑํ•˜์—ฌ LLM์— ์ „์†กํ•จ์œผ๋กœ์จ, LLM์˜ ๋‚ด๋ถ€ ํ† ํฐ ์ƒ์„ฑ ๊ณผ์ •์„ ์ž๋™์œผ๋กœ ์ œ์–ดํ•˜๋Š” ๋‹จ๊ณ„;
  4. (d) ์ƒ๊ธฐ ์ œ์–ด๋œ LLM์œผ๋กœ๋ถ€ํ„ฐ, ์ƒ๊ธฐ ์ œ์•ฝ์กฐ๊ฑด์„ ๋งŒ์กฑํ•˜๊ณ  ์ปดํŒŒ์ผ ์˜ค๋ฅ˜์œจ์ด ์‚ฌ์ „ ์ •์˜๋œ ์ž„๊ณ„๊ฐ’ ๋ฏธ๋งŒ์ด๋ฉฐ GPU ๋ฉ”๋ชจ๋ฆฌ ์‚ฌ์šฉ๋Ÿ‰์ด ๊ฐ์†Œ๋œ ์ตœ์ ํ™”๋œ ์ฝ”๋“œ๋ฅผ ์ˆ˜์‹ ํ•˜๋Š” ๋‹จ๊ณ„๋ฅผ ํฌํ•จํ•˜๋Š” ๋ฐฉ๋ฒ•.

→ ์ด ์˜ˆ์‹œ๋Š” ๋‹จ์ˆœํ•œ ๊ฒฐ๊ณผ๋ฌผ์ด ์•„๋‹ˆ๋ผ, ‘์ปดํŒŒ์ผ ์˜ค๋ฅ˜์œจ ๊ฐ์†Œ’, ‘GPU ๋ฉ”๋ชจ๋ฆฌ ์‚ฌ์šฉ๋Ÿ‰ ๊ฐ์†Œ’์™€ ๊ฐ™์€ ์‹œ์Šคํ…œ ์ˆ˜์ค€์˜ ์ธก์ • ๊ฐ€๋Šฅํ•œ ๊ธฐ์ˆ ์  ํšจ๊ณผ๋ฅผ ๋ช…ํ™•ํžˆ ํ•จ์œผ๋กœ์จ ํŠนํ—ˆ ๋“ฑ๋ก ๊ฐ€๋Šฅ์„ฑ์„ ํฌ๊ฒŒ ๋†’์ž…๋‹ˆ๋‹ค.

 

์ž์ฃผ ๋ฌป๋Š” ์งˆ๋ฌธ ❓

Q: “๊ณ ์–‘์ด ์‹œ๋ฅผ ์จ์ค˜” ๊ฐ™์€ ๊ฐ„๋‹จํ•œ ํ”„๋กฌํ”„ํŠธ๋„ ํŠนํ—ˆ๋ฅผ ๋ฐ›์„ ์ˆ˜ ์žˆ๋‚˜์š”?
A: ์•„๋‹ˆ์š”, ๊ทธ ์ž์ฒด๋Š” ์•„์ด๋””์–ด์— ๋ถˆ๊ณผํ•˜์—ฌ ํŠนํ—ˆ๋ฅผ ๋ฐ›๊ธฐ ์–ด๋ ต์Šต๋‹ˆ๋‹ค. ํŠนํ—ˆ์˜ ๋Œ€์ƒ์€ ์‹œ๋ฅผ ์ƒ์„ฑํ•˜๊ธฐ ์œ„ํ•ด ํŠน์ • ๋ฐ์ดํ„ฐ ๊ตฌ์กฐ(์˜ˆ: ์‹œ์  ์žฅ์น˜, ์šด์œจ ๊ตฌ์กฐ๋ฅผ ์ •์˜ํ•œ ์Šคํ‚ค๋งˆ)๋ฅผ ๊ฐ€์ง„ ํ”„๋กฌํ”„ํŠธ๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ LLM์„ ์ œ์–ดํ•˜๊ณ , ๊ทธ ๊ฒฐ๊ณผ ์ปดํ“จํŒ… ์ž์›์„ ๋œ ์‚ฌ์šฉํ•˜๊ฑฐ๋‚˜ ํŠน์ • ์Šคํƒ€์ผ์˜ ์‹œ๋ฅผ ๋” ์ •ํ™•ํ•˜๊ฒŒ ์ƒ์„ฑํ•˜๋Š” ๋“ฑ์˜ ๊ธฐ์ˆ ์  ๋ฐฉ๋ฒ•์ด๋‚˜ ์‹œ์Šคํ…œ์ž…๋‹ˆ๋‹ค.
Q: ํ”„๋กฌํ”„ํŠธ ๊ธฐ์ˆ ์˜ ‘๊ธฐ์ˆ ์  ํšจ๊ณผ’๋Š” ๊ตฌ์ฒด์ ์œผ๋กœ ์–ด๋–ค ๊ฒƒ๋“ค์ด ์žˆ๋‚˜์š”?
A: ๋Œ€ํ‘œ์ ์œผ๋กœ ์ฝ”๋“œ ์ƒ์„ฑ ์‹œ ์ปดํŒŒ์ผ ์˜ค๋ฅ˜์œจ ๊ฐ์†Œ, GPU๋‚˜ ๋ฉ”๋ชจ๋ฆฌ ๊ฐ™์€ ์ปดํ“จํŒ… ๋ฆฌ์†Œ์Šค ์‚ฌ์šฉ๋Ÿ‰ ์ ˆ๊ฐ, ์‘๋‹ต ์ƒ์„ฑ ์‹œ๊ฐ„ ๋‹จ์ถ•, ํŠน์ • ๋ฐ์ดํ„ฐ ํ˜•์‹(JSON, XML ๋“ฑ)์˜ ์ถœ๋ ฅ ์ •ํ™•๋„ ํ–ฅ์ƒ ๋“ฑ์ด ์žˆ์Šต๋‹ˆ๋‹ค. ์ค‘์š”ํ•œ ๊ฒƒ์€ ์ด ํšจ๊ณผ๊ฐ€ ์ธก์ • ๊ฐ€๋Šฅํ•˜๊ณ  ์žฌํ˜„ ๊ฐ€๋Šฅํ•ด์•ผ ํ•œ๋‹ค๋Š” ์ ์ž…๋‹ˆ๋‹ค.
Q: ๊ตญ์ œ ์ถœ์› ์‹œ ๊ตญ๊ฐ€๋งˆ๋‹ค ์ฒญ๊ตฌํ•ญ์„ ๋‹ค๋ฅด๊ฒŒ ์ž‘์„ฑํ•ด์•ผ ํ•˜๋‚˜์š”?
A: ๋„ค, ํ•ต์‹ฌ ์ „๋žต์€ ๊ฐ™์ง€๋งŒ ๊ฐ ํŠนํ—ˆ์ฒญ์ด ์ค‘์‹œํ•˜๋Š” ํฌ์ธํŠธ์— ๋งž์ถฐ ๊ฐ•์กฐ์ ์„ ๋‹ฌ๋ฆฌํ•˜๋Š” ๊ฒƒ์ด ์œ ๋ฆฌํ•ฉ๋‹ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด, ๋ฏธ๊ตญ(USPTO) ๋ช…์„ธ์„œ์—๋Š” ‘์ปดํ“จํ„ฐ ๊ธฐ๋Šฅ์˜ ๊ตฌ์ฒด์  ๊ฐœ์„ ’์„, ์œ ๋Ÿฝ(EPO)์—๋Š” ‘๊ธฐ์ˆ ์  ๋ฌธ์ œ ํ•ด๊ฒฐ์„ ํ†ตํ•œ ๊ธฐ์ˆ ์  ํšจ๊ณผ’๋ฅผ, ํ•œ๊ตญ(KIPO)์—๋Š” ‘์‹œ์Šคํ…œ ๊ตฌ์„ฑ๊ณผ ์ฒ˜๋ฆฌ ํ๋ฆ„์˜ ๊ตฌ์ฒด์„ฑ’์„ ๋” ๋ถ€๊ฐ์‹œํ‚ค๋Š” ๋ฐฉ์‹์ž…๋‹ˆ๋‹ค.

๊ฒฐ๋ก ์ ์œผ๋กœ, AI ํ”„๋กฌํ”„ํŠธ๋ฅผ ํŠนํ—ˆ๋กœ ๋ณดํ˜ธ๋ฐ›๋Š” ๊ธธ์€ ๋ถ„๋ช…ํžˆ ์กด์žฌํ•ฉ๋‹ˆ๋‹ค. ๋‹ค๋งŒ, ‘๋ฌด์—‡์„ ์š”์ฒญํ•˜๋Š”๊ฐ€’๋ผ๋Š” ์•„์ด๋””์–ด์˜ ์ฐจ์›์„ ๋„˜์–ด, ‘์–ด๋–ป๊ฒŒ ์ปดํ“จํ„ฐ ์‹œ์Šคํ…œ์„ ๊ธฐ์ˆ ์ ์œผ๋กœ ์ œ์–ดํ•˜๊ณ  ๊ฐœ์„ ํ•˜๋Š”๊ฐ€’๋ฅผ ๋ช…ํ™•ํžˆ ๋ณด์—ฌ์ฃผ๋Š” ์ „๋žต์  ์ ‘๊ทผ์ด ํ•„์š”ํ•ฉ๋‹ˆ๋‹ค. ์ด ๊ธ€์ด ์—ฌ๋Ÿฌ๋ถ„์˜ ํ˜์‹ ์ ์ธ ์•„์ด๋””์–ด๋ฅผ ๊ฐ•๋ ฅํ•œ ์ง€์‹์žฌ์‚ฐ์œผ๋กœ ๋งŒ๋“œ๋Š” ๋ฐ ์ž‘์€ ์‹ค๋งˆ๋ฆฌ๊ฐ€ ๋˜์—ˆ์œผ๋ฉด ํ•ฉ๋‹ˆ๋‹ค. ๋” ๊ถ๊ธˆํ•œ ์ ์ด ์žˆ๋‹ค๋ฉด ๋Œ“๊ธ€๋กœ ๋ฌผ์–ด๋ด ์ฃผ์„ธ์š”~ ๐Ÿ˜Š

Patentability of LLM Prompts: Overcoming Abstract Idea Rejections

  Can a Simple Command to an AI Be Patented? This article provides an in-depth analysis of how LLM prompt techniques can transcend mere...