From patent mining to portfolio pruning, AI has the capability to be a useful assistant when managing a company’s intellectual assets, writes Vaibhav Henry, chief growth officer at Sagacious IP

Artificial Intelligence has been on an evolutionary journey since the 1950s, but it is the recent surge in Generative AI that has truly brought it into the global spotlight.

While the AI/machine learning research and scientific communities had been tracking this journey closely, the public remained largely unaware. This changed, however, with the release of OpenAI’s ChatGPT, which captured the world’s attention.

The release of GPT-3 by OpenAI around 2019-2020 was a defining moment, demonstrating the vast potential of large language models (LLMs). This tool not only showcased the practical applications of LLMs but also made them widely accessible, igniting discussions and interest across the globe.

The past two years have been especially significant, marking a new era in AI that promises to reshape our interaction with technology.

Leveraging AI reasoning in IP workflows

Intellectual property has long been a domain where intricate, knowledge-intensive tasks are handled by seasoned professionals drawing upon years of expertise, and deep subject knowledge.

The powerful reasoning capabilities of LLMs like ChatGPT present opportunities to augment human expertise at various stages of such IP workflows.

A. Patent mining

AI’s impact on patent mining is truly groundbreaking, propelling innovation forward, boosting competitive advantage and reshaping how we safeguard and capitalise on intellectual assets.

Whether it is for innovation benchmarking, competitive analysis, R&D strategic planning or defending a lawsuit or potential acquisition targets, the quality of patent mining depends on the accuracy of patents identified. Given the high stakes involved, patent mining is a sensitive subject where decisions must be astute and well-informed.

Let us dive deep into some specific use cases:

Use Case 1: Benchmarking patents against competitor’s products

The most significant way of strategising your IP is to benchmark your patent portfolio against various competitors’ products and services. This analysis can provide insights into the market value and commercial relevance of your patents.

AI reasoning can streamline this benchmarking process:

Step 1: AI-driven patent analysis

AI can rapidly analyse thousands of patents in your portfolio, performing an initial high-level evaluation to identify potentially relevant ones.

Step 2: Mapping to competitor products

AI can then identify commercial products or services offered by competitors in the market. This analysis can be broken down based on patent technology fields or jurisdictions for added granularity.

Step 3: Identifying high-scoring patents

By mapping your patents to competitor products, the AI can highlight patents that score highly against various competitors’ offerings, indicating potential infringement or licensing opportunities.

Through this process, you can gain valuable insights into the strengths and weaknesses of your patent portfolio relative to those of market competitors. This can inform strategies for licensing programmes or defensive positioning against potential threats.

Use Case 2: Responding to a lawsuit

If a competitor files a lawsuit against your company, AI reasoning can be leveraged to quickly identify patents from your portfolio that can be used for counter-assertion. Suppose your company holds 4,000 patents, and you need to efficiently pinpoint the most relevant ones to assert against the competitor’s products or services. Here is how AI can assist:

Step 1: AI-driven patent analysis

AI systems can rapidly analyse thousands of patents in your portfolio, performing an initial evaluation to identify potentially relevant candidates.

Step 2: Mapping patents to competitor’s products

The AI can, further, align your patents with the competitor’s commercial products or services, providing a scoring system (eg, high/medium/low) to indicate the strength of the mapping for each patent.

Step 3: Prioritising high-scoring patents

With the AI’s analysis, you can sort and prioritise the high-scoring patents marked as ‘high’ representing the most promising candidates for counter-assertion against the competitor.

By leveraging AI reasoning in this manner, you can rapidly identify patents from large portfolio, to assert against a competitor’s litigation, streamlining the process and strengthening your defensive strategy.

Observation: When AI is tasked with providing information on individual products, the depth of the content is substantial. However, if AI is asked to generate in-depth information about several products simultaneously, the resulting content tends to be shallower.

Possible Reason: AI’s response time remains constant for generating any input, whether it is for one product or multiple products. Consequently, as the number of products increases, the AI allocates the same amount of time to each, resulting in less comprehensive information for each product.

B. Patent portfolio management

Effective management of patent portfolios is essential for companies to safeguard their innovations, enhance patent value, and maintain a competitive advantage. In the past, managing patent portfolios was labour-intensive, but the emergence of Generative AI has transformed the process, introducing automation and promising better outcomes. These technological advancements are reshaping how patents are acquired, analysed, and utilised, providing valuable insights. Let us see how:

Use Case: Quick analysis for patent pruning

Maintaining a lean and valuable patent portfolio is essential for companies that file and are granted multiple patents annually. Given that many patents may not reach their full potential over the standard 20-year lifecycle, resulting in eventual abandonment, it is crucial to have a systematic approach for identifying which patents to retain or prune. This not only optimises the IP budget but also aligns the patent portfolio with the company’s evolving business needs.

For instance, consider a scenario where you have 500 patents approaching the third annuity fee (11th year) at approximately $7,700 each. The challenge lies in discerning which patents to maintain and which to let go. AI reasoning can streamline this process:

Step 1: Identify potential targets

AI scans the market to detect potential infringing products or services that relate to your patents.

Step 2: Patent scoring

AI assigns scores (eg, high/medium/low) to each patent based on its relevance to the identified products, emphasising commercial importance.

Step 3: Identify low-scoring patents

Focus on patents that receive low scores from the AI analysis, indicating diminished market relevance.

Step 4: Analyse technical problems and alternatives

For patents that return low scores, AI can analyse the technical problems they address and identify alternative solutions that the industry may have adopted, shedding light on why certain patents did not gain traction.

By incorporating regular patent pruning into IP strategy, companies can offload assets that no longer contribute to their business objectives, thereby saving substantial costs. For example, if you decide to prune 200 out of 500 patents deemed unlikely to realise future value, you could save up to $1.5 million in annuity fees.

Thus, incorporating AI into the regular patent pruning strategy enables companies to offload non-essential assets efficiently. This not only saves money but also ensures that the retained patents are those that truly contribute to the company’s innovation and competitiveness. Thus, AI is not just an option; it is a strategic ally in maximising the impact of your IP budget.

C. Patent drafting

AI fanatics may claim that AI tools can draft a complete patent application within a very short timeframe, which is an unrealistic expectation for a task that typically requires significant effort and time. Such claims can be misleading and may undermine the true potential of AI in augmenting intellectual property workflows.

Use Case: Smart use of AI for provisional patent drafting

Though AI cannot be relied upon to generate the final patent application, it can be used to get a head start. AI can assist in:

  • Comparing invention and prior art: on providing the invention and prior art to the AI, it should be able to provide a decent one-on-one comparison of the patent draft and the references. This way, you get to know the exact novel features of the invention, thereby getting a better handle on patent scope.
  • Generating initial claims: AI can generate a decent and small set of initial claims to guide further drafting, so that patent attorneys and drafting professionals may never need to start with a blank slate each time they begin working.

AI is not going to replace patent attorneys, it is merely going to assist them.

Observation: While querying AI to generate a comprehensive claim set (20+) usually results in low-quality AI-generated claims; the same AI generates a small set of claims (six to eight) decently.

Possible reason: As discussed above, the AI’s response time remains constant regardless of the number of claims it needs to generate. Whether it is drafting a single claim or multiple claims, the AI allocates the same amount of time to each claim. Consequently, as the number of claims increases, the AI may not have sufficient time to delve deeply into each one, resulting in less comprehensive and potentially lower-quality claims.

But is it secure?

Security and data privacy are paramount when handling new inventions and proprietary information. Recent incidents of data leakage from public AI tools like ChatGPT have understandably raised alarms. To mitigate these risks, it is not recommended to input sensitive invention disclosures or proprietary data into publicly available AI tools. Instead, consider leveraging the same LLMs like GPT-4, Claude, or PaLM via secure API access.

  • Confidential API access: All major AI providers, such as OpenAI, Anthropic, and Google, offer confidential API access to their LLM models. These confidential APIs ensure that input data remains secure and is not utilised for retraining the models or any other unintended purposes.
  • Open-Source models in private data centres: Despite confidentiality assurances from major AI providers, some patent professionals may still feel cautious about sending invention data to these companies. This concern is valid, given the sensitive nature of intellectual property.

An alternative approach is to use open-source LLM models hosted within a private data centre or a cloud provider like AWS or Google Cloud. In this setup, the LLM models and proprietary data (eg, invention disclosures) never leave the controlled environment, significantly mitigating potential risks.

Patent professionals must carefully evaluate their security needs and comfort levels when integrating AI into their workflows. Implementing robust data protection measures and thoughtfully balancing risks and benefits are crucial for responsibly harnessing the power of AI for patent drafting and IP management.

D. Patent searching

Searching for patents to ensure your idea is original has always been a lot of work and takes up a lot of time. But now, AI is here to help and make this process much easier.

Use Case: AI-assisted patent searching

AI language models like GPT can generate patent numbers in response to queries, but these results may not always be the most accurate or relevant. For comprehensive patent searches, relying solely on an LLM’s reasoning is often insufficient. Instead, proprietary IP data stored in structured databases becomes essential. These databases, combined with AI’s capabilities, offer a more structured and effective approach to patent searching.

The role of AI-based semantic searching tools

AI-based semantic searching tools, designed specifically for patent research, is a robust solution. These tools use AI models to generate concise invention descriptions and iteratively refine search results. This interactive process between the user and the AI model streamlines patent searching, enabling efficient identification of the most applicable prior art.

Proprietary databases maintained by companies such as Amplified, IPRally and Orbit Intelligence combine vast amounts of structured IP data with AI capabilities, delivering precise and relevant search results.

Several AI-powered patent-searching tools are discussed below:

Challenges in AI-powered patent searching

While AI-powered patent databases offer significant benefits, users often encounter several challenges that can hinder their search experience:

  • Semantic search limitations: AI struggles with understanding the context of system claims, often matching components individually instead of recognising the central innovation. If you provide a system claim, for instance, the AI may attempt to match various components individually rather than identifying the core novelty, which is the typical goal in patentability, invalidity, and other patent-related searches.
  • Query structuring: Users must learn to effectively structure search queries, which involves more than just entering the invention details into a text box.
  • Understanding AI logic: There is a need for users to comprehend the AI’s search logic to refine their searches for better outcomes.
  • Data overload: AI can sometimes provide an overwhelming amount of data, making it difficult to pinpoint relevant patents.
  • Language and terminology: AI may not fully grasp the nuances of patent language, leading to missed or irrelevant results.
  • Constant updates: AI databases require regular updates to include the latest patents, which can be a logistical challenge.
     

Overcoming challenges for effective patent searching:

Achieving optimal results from AI-powered patent databases requires adhering to certain best practices, such as:

  • Write the semantic query (invention disclosure or claim reading) in your own language, focusing on the inventive aspects and novelty that need to be searched within available patent documents.
  • Practice and refine your semantic searching skills to yield better results.

These tips help bridge the gap between the user’s intent and the AI’s processing capabilities, ensuring more accurate and relevant search outcomes.

It is important to note that while the above discussed use cases are selected for their relevance, they may not encompass the full range of possibilities. There could be other use cases that contribute to the comprehensiveness of the analysis, depending on specific needs and contexts.

Table 1: A comparison of the qualities of proprietary databases

Feature/advantage Amplified IPRally Orbit Intelligence
AI-powered Search Advanced semantic search algorithms Knowledge graph-based method Semantic and Boolean search options
Contextual understanding Understands context and nuances Grasp technical context and relationships Comprehensive search capabilities
Technical precision General patent relevance High accuracy for complex technical searches Detailed insights into patent landscapes
Ideal use case Quick, relevant patent searches Complex technical patent searches Versatile, detailed patent searches with analytical insights

Best practices for integrating AI in IP

Streamlining IP work further, AI can be used for multiple roles – an intelligent intern automating mundane tasks and analyses, an expert bridging gaps in human expertise by providing guidance on complex concepts, or a peer/copilot collaborating synergistically alongside professionals.

1. AI model security

  • Cloud-based API models: Preferred for companies that trust reputable providers with their data, offering robust security measures and compliance with data protection regulations.
  • Open-source models: Chosen for highly sensitive IP data where companies need full control without relying on external servers, providing transparency and community-driven security enhancements.
     

2. Prompt chaining

  • Depending on the depth of information required, consider the below approaches:
  • Broad-level data: If you require a general overview or a wide range of information, inputting multiple data points at once can be effective. This approach provides a broader context and allows for a more comprehensive response.
  • Depth and precision: However, when depth and precision are crucial, consider using prompt chaining or cascaded prompts. By breaking down the inquiry into smaller, focused questions for individual data points, you can explore specific details thoroughly and generate higher-quality responses.
     

3. Task decomposition

  • Given the breadth (number) and depth (scope) of IP services, a one-size-fits-all approach is ineffective. Instead, a meticulous deconstruction of each service into its constituent tasks is necessary to determine where AI can be most impactful.
     

4. User involvement

  • As AI becomes more ubiquitous, the uniqueness of its outputs may diminish. It is the human touch that can elevate these outputs, infusing them with creativity and insight. Balanced engagement from users is crucial to refine and personalise AI-generated results, ensuring they meet the nuanced needs of IP services.
     

5. Feedback mechanisms for evolving AI

  • Continuous learning: Implementing feedback mechanisms is vital for enhancing AI’s accuracy and adaptability. AI models should actively seek user guidance on ambiguous tasks, improving decision-making over time.
  • Periodic retraining: Regular updates and retraining with fresh data ensure AI remains effective and relevant in evolving IP landscapes.

The path forward

From patent mining and portfolio management to drafting and searching, AI reasoning can augment human expertise at various stages. However, it’s crucial to balance the potential benefits with data security concerns, implementing robust protocols like secure API access or private cloud deployments.

By adhering to best practices like prompt chaining, task decomposition, and continuous learning through feedback mechanisms, companies can responsibly integrate AI into their IP processes, unlocking new levels of efficiency and competitive advantage. Furthermore, as AI systems learn and evolve through ongoing interactions, they become more sophisticated, providing increasingly nuanced support for IP tasks.

Looking forward, the integration of AI into IP management goes beyond simple automation; it entails creating a dynamic partnership between technological capabilities and human understanding. This collaboration has the potential to drive innovation in IP strategy and administration by combining AI’s computational power with human intuition.



Source link

Leave a Comment

Your email address will not be published. Required fields are marked *