ChatGPT is an AI language model developed by OpenAI, based on the GPT (generative pre-trained transformer) architecture. It has been gaining significant attention since its launch due to its ability to generate human-like text and perform a wide range of content-building tasks across various fields, including patent research and analysis.
In the field of patents, ChatGPT is considered to have the potential to revolutionize the conventional patent search and analysis process. Conventional patent searches require significant manual effort and time to thoroughly understand the language, context, and technology behind a patent.
Can ChatGPT's language model enhance accuracy and efficiency in patent operations and assist in various patent processes?
Let’s explore this in detail with an example of the already concluded case i.e., Boeing patent case of a missing comma. The patent number (EP1798872B1) was taken from this case, which was a mere miss of a single comma that invalidated the patent, highlighting the importance of precision in patent writing and the significant impact of even the smallest errors.
Note – The patent was selected only for testing ChatGPT’s patent search algorithm.
LEVEL 1 – Broad approach - We asked ChatGPT to explain a patent “EP1798872B1”
About the ChatGPT test patent
The test patent EP1798872B1 discloses a method for handling aeroplane communications providing numerous broadcast communication networks. The approach includes creating and storing messages with priority, analysing network preferences, and continuously monitoring network availability to guarantee messages are carried over the most appropriate network. Messages are saved if the preferred network is unavailable and the priority falls below a certain level.
What did ChatGPT return when asked – “Can you explain patent EP1798872B1”
When asked to explain this invention, ChatGPT produced an irrelevant response concerning a system for choosing multimedia content components for a media stream, which has nothing to do with the patent’s subject matter.
So why did it give a totally irrelevant response to a simple question?
Since it is an AI language model, ChatGPT does not have access to specific sources or databases for patent information. So we can’t really rely on any information by providing an open ended question. Even free databases like Google Patent, Espacenet, Freepatentsonline, etc. require authorized credentials, security permissions, and a specific understanding of the database’s structure and query language (operators) for returning right set of information.
Even if ChatGPT had access to such databases, it would need to be programmed/ trained with specific instructions on how to fetch the data from different elements of a patent.
LEVEL 2 – Assisted approach
In the second stage, our experts provided the correct title to see how it will change ChatGPT response.
Why did ChatGPT provide a generic response after the title of the patent was provided?
As an AI language model, ChatGPT provides information based on its pre-existing knowledge. It is capable of understanding natural language queries and providing a range of information based on training data which enables ChatGPT to answer a wide range of questions and provide information on a variety of topics.
However, accessing patent databases and retrieving information from them requires specialized knowledge and expertise, such as understanding the database schema, the query language used, and the data retrieving methods.
Although, the next response of ChatGPT brought us back to square one when asked,
LEVEL 3 – Narrow approach
Moving on, when we asked ChatGPT to compare the claim of interest i.e., the 1st claim in its original form (missing comma) and with added comma directly copy-pasting to understand how ChatGPT extract the impact of the missing comma in the claim.
Expert: What’s the difference between 1 and 2
- evaluating a preference to determine a preferred network of the plurality of transmission networks, wherein the preference comprises a preference list identifying a selection of the plurality of broadcast networks in order of preference and identifying the highest in preference of the plurality of broadcast networks in the preference list that is available;
- evaluating a preference to determine a preferred network of the plurality of transmission networks, wherein the preference comprises a preference list identifying a selection of the plurality of broadcast networks in order of preference (,) and identifying the highest in preference of the plurality of broadcast networks in the preference list that is available;
ChatGPT: The difference between claim 1 and claim 2 lies in the wording of the last part of the sentence.
Claim 1 states “identifying the highest in preference of the plurality of broadcast networks in the preference list that is available”. This implies that if the highest preferred network is not available, the system will identify the next highest preferred network that is available and select it for use.
On the other hand, claim 2 states “and identifying the highest in preference of the plurality of broadcast networks in the preference list that is available”. This implies that the system will only identify the highest preferred network that is available and select it for use. It will not check for the availability of other preferred networks if the highest preferred network is not available.
So, the main difference between the two claims is in the handling of the situation where the highest preferred network is not available. Claim 1 allows for the selection of the next highest preferred network that is available, while claim 2 only allows for the selection of the highest preferred network that is available.
Next was to set the valid base for the argument provided during the prosecution of the case.
Expert: What if this different explanation is given in the description and figure of the patent, will it be considered for validation of the patent?
ChatGPT: The explanation given in the patent description and figures can be used to support and clarify the claims made in the patent, but it does not necessarily guarantee the validation of the patent.
The patent claims are the legal definition of the invention and must meet certain criteria, such as novelty, inventiveness, and utility, to be valid. The patent office will examine the claims to determine if they meet these criteria, and may request additional information or modifications to the claims during the examination process.
While the patent description and figures can be helpful in explaining the invention and supporting the claims, they are not a substitute for meeting the legal requirement for patentability. Ultimately, it is up to the patent examiner to determine whether the claims meet the criteria for patentability based on their interpretation of the claims and supporting materials.
What does that mean, can ChatGPT be used for patent search/analysis or not?
Well it depends, ChatGPT is well-suited for semantic recognition and extraction, which means it can extract key information from technical texts and automatically construct a search formula to provide you with the content in response to your question. This could potentially assist the work of patent professional to some extent. Utilising ChatGPT’s capabilities in extracting features from the technical text combined with patent experts we can further explore its potential in patent searches and streamline the process of retrieving relevant information from patent databases.
MCRPL very well understands the importance of human expertise in patent search analysis and related services. Our team of experienced patent professionals not only holds technical expertise but also understands the legal and regulatory aspects of patent filing and prosecution. We provide end-to-end patent services, including patentability searches, prior art searches, patent drafting, patent landscapes, patent portfolio management, and more.
While we acknowledge the potential benefits of AI-based tools such as ChatGPT in streamlining the patent search process, we believe that relying solely on such tools could lead to errors and inaccuracies in the analysis. That is why we use hybrid approach Human plus machine (MCRANK, our proprietary AI based text analytics tool) as an add-on to our human expertise, thereby ensuring the accuracy, precision and reliability of our results. We also continuously invest in training our patent professionals to keep them up-to-date with the latest advancements in AI and other technologies. Our goal is to provide our clients with the best possible service and to help them protect their intellectual property rights.
