Harnessing GenAI in Patent Drafting: Transform Your Firm Now!

Webinar Illustration

Submit your email address to access replay

We Answer Your Questions
Q1: Please describe the starting materials used for your demo. I want a sense of how much time it took to draft quality materials to upload into davinci.

Today davinci requires a set of claims and elements of description of the innovation to generate a 1st draft. If you don’t provide drawings as input, the system won’t save time on the description of the drawings. Like always in software, garbage in garbage out. The higher the quality of the inputs, the higher the quality of the outputs.

Q2: Can you upload a boilerplate application that you wish for davinci to use as a basis for populating the rest of the application (based on your disclosure/claims/drawings)?

You can upload claims, drawings and davinci will use them to populate the rest of the application.

Q3: Is the model further trained on patents and if so from what jurisdictions/authorities?

The model is trained on all patents available on Google Patents, and all legal material available on the internet, IP law, case law, MPEP guide,…

Q4: How do you ensure the correctness of inventorship? If the AI is filling in the specs, and generating embodiments, how does your system make sure those that you put down are what the inventor(s) had in their mind?

1. davinci is an assistant designed to speed up the work of the attorney and help improve the overall quality of the patent. It can’t, in any case, replace the work and the expertise of attorneys.

2. davinci leverages a complex combination of models to generate the output. Some are dedicated to ensuring the consistency of the generated output with the inputs.

Q5: Have you compared the actual results from davinci with those from other AI-based patent drafting tools? How is davinci different?

1. Measuring the performance of generative tools is complex. At davinci we used the time saved for a defined level of quality as a proxy to estimate the performance of davinci.

2. Thanks to 6 years of experience developing AI products for machine learning teams, serving demanding customers such as SAP, Airbus, and the Department of Defense, the davinci teams possess a unique capability to construct scalable, high-performing, and highly secure AI solutions.

Q6: Can we tune the generation level? E.g., force the GenAI to strictly stick to the original language of the claims and not be "creative"?

davinci provides dedicated features to handle a set of constraints within which the AI generates the output. For instance, you can create your generation style that aligns with the vocabulary you typically use or specify the terminology employed during the generation.

Q7: For software inventions involving flow charts, often the great majority of the time goes into preparing the flow chart. Where is the time-saving in this instance?

For the moment, davinci doesn’t generate flowcharts. It's something we have on our roadmap. We prioritize features based on customer needs.

Q8: Does the uploaded invention disclosure need to be in a particular format?

davinci currently manages input in Word (docx), pdf, .txt formats. We regularly add new formats.

Q9: What technologies can be handled?

davinci works best on Software patents, Mechanical patents, and Process & Method patents. Other patents will come soon.

Feel free to share your needs. We regularly review our priorities based on requests.

Q10: What is it about the trained AI that makes it particularly able to generate patent specifications?

davinci’s capabilities are due to the advanced combination of AI technologies to properly understand inputs, orchestrate advanced generation capabilities, and control the quality prior to any generation.

Q11: Does the software have the ability for the user to upload a knowledge base of related applications, both to provide context for the application as well as to maintain consistency of terminology?

davinci lets the user to upload the prior art, which is leveraged to provide the correct context. Terminology is grounded through the invention disclosure, and we also let the user specify terms he does not want to see, with an optional replacement, to ensure that the generation is aligned with the user preferences.

Q12: When revising a draft prepared by davinci, do you simply ask davinci to, e.g., "expand the scope/add support to section X"? I.e., is asking davinci for revisions similar to ChatGPT?

davinci offers features and a user interface to interact with the AI and iterate on improving of the 1st draft. This can be prepackaged prompts such as “expand the scope” or custom prompts for advanced users.

Q13: Can you handle different languages?

davinci handles English and French. It will soon manage all major European languages.

Q14: If one wants to change the drafted paragraph. Can we add a prompt to re-generate a new paragraph?

davinci offers such features that allow it to iterate on the generated text.

Q1: How does the tool ensure data privacy and security?

1. davinci is hosted on Microsoft Azure, the most trusted cloud provider by law firms.

2. davinci matches the highest security standards, SOC2 Type 2, ISO27001, compliant with GDPR. All data is encrypted in transit and at rest. Once the data is processed through the API it is neither retained, nor reused for retraining purposes. Data is strictly segregated.

Q2: When a draft is generated by davinci and then amended by the attorney to conform with the firm drafting format, does davinci learn the firm format?

1. No usage data is retained and reused for training or retraining purposes.

2. davinci can be tailored to adhere to firm-specific formats, wording, or style preferences, thanks to dedicated features.

Q3: Is the LLM you use within the four walls of your system or are you partnering with a third party? Will the data privacy promises extend to them as well?

davinci is a complex AI system leveraging different AI technologies as components of an advanced system. Parts of the davinci AI system are fully proprietary, parts are leveraging third parties, including OpenAI technologies. All data security & privacy policies apply to all providers.

Q4: Can you point to any third-party audits of your security systems?

Our certifications and all the documentation regarding security & compliance are available on our trust center on this link.

Q1: What's the cost?

1.  As long as davinci remains in beta, it is made available to selected firms at no cost in exchange for regular feedback and deep-dive discussions on specific topics on which we want to collect diverse points of view.

2. Kili Technology, the parent company of davinci, is profitable in other product lines, making it sustainable

Q2: What is the reason for the wait list? It looks like the product could be ready for demos already.

1. GenAI is a radically new technology, and it is evolving very fast. Relying on technology to write a patent is a massive change for attorneys. We believe in such a phase, it is important to work very closely with firms that are shaping the future of the industry.

2. Patent writing is a complex job that requires a deep understanding of different domains. We might consider making davinci available when we reach a consistent level of performance across domains.

Q3: How can I get a demo using an invention disclosure from a client, and claims that I've drafted?

You can join the waitlist by clicking this link. We will reach out to you.