Last week, I made a brief comment on LinkedIn about the legally compliant use of AI in adherence to the professional duties of patent attorneys, sparking a lively discussion.
What started as a casual observation about the benefits of local AI models quickly turned into a full-blown debate among patent professionals (the LinkedIn discussion thread is in German). The question at the heart of the conversation:
Are local AI models only for computer nerds, or can they be a viable solution for everyone in the patent industry?
My point was simple: For a confidentiality-sensitive task like patent drafting, it makes sense to use a local AI solution as long as it produces suitable results. Only when local AI reaches its limits should we consider cloud-based solutions, while also factoring in their confidentiality implications.
The debate quickly brought together a variety of opposing perspectives from patent professionals and AI experts. Here’s a summary of the key arguments made during the discussion.
- Bastian Best: I advocated for local AI models, especially for patent drafting, arguing that they’re cost-effective, easy to install and use, and offer higher confidentiality.
- Dr. Christian A. Mohr: He preferred cloud-based AI solutions for their scalability, continuous updates, and flexibility.
- Andreas Graf: Andreas chimed in, emphasizing the high hardware requirements for fine-tuning AI models, which are often only feasible in the cloud.
- Dr. Jan-Christian Schütte: He echoed Dr. Mohr’s point, stressing that cloud solutions allow patent attorneys to focus on their core tasks rather than spending time on prompt engineering.
- Anna von Berg: Anna supported cloud-based solutions, highlighting the superior hardware infrastructure, the availability of larger models, and the reduced maintenance burden.
This gets right to the core of the great AI debate about local vs. cloud-based models. Let’s dive deeper into the core arguments of the debate and weigh the pros and cons of each side.
The Case for Cloud-Based AI Models
- Scalability and Continuous Updates:
It was argued that cloud-based AI systems are scalable and benefit from ongoing updates. As Dr. Christian A. Mohr pointed out, these updates ensure that you’re always working with the latest models and data. This level of scalability is difficult to achieve with local setups, according to some debaters. - Flexibility and Accessibility:
Dr. Mohr also highlighted the flexibility and accessibility of cloud solutions. In a highly connected world, being able to access powerful AI from anywhere, without needing specialized hardware, is seen as a big advantage for him. - Up-to-Date Models:
Cloud-based solutions give users access to the most current language models and training data, keeping them aligned with the latest technological advancements. Both Dr. Mohr and Dr. Schütte emphasized that this was a major advantage in fast-moving industries like AI. - Utilizing Trusted Cloud Infrastructure:
Many companies and clients already trust cloud platforms like Microsoft Azure, which handle massive amounts of sensitive data every day. Dr. Schütte pointed out that this can ease concerns over privacy and security, as these established platforms often come with built-in safeguards.
The Case for Local AI Models
- Confidentiality:
When it comes to patent drafting, confidentiality is paramount, especially when handling sensitive information about an invention before a patent application has been filed. As I argued, local AI models provide superior control over sensitive data because everything stays on your own system—no external cloud provider has access to your information. I find this to be a key advantage, especially for patent professionals handling confidential invention details. - Cost-Effectiveness and Simplicity:
Local AI models can be surprisingly affordable, and many are even free. They’re also simpler to install and operate than you might think. As I mentioned, I believe they’re not just for “computer nerds”—any patent attorney who can install software on his or her computer can get them up and running with minimal hassle. This free video course teaches you how to set up your local AI toolchain. - Stability and Reproducibility:
A significant benefit of local models is the ability to tailor them to your specific needs. You can train and fine-tune a local model to produce consistent results in a specific style, which is crucial for patent drafting. Unlike cloud models that update frequently, local models provide the stability needed for long-term projects. I, personally, prioritize a stable AI model that consistently writes in the style I need.
Additional Considerations from the Debate
- Prompt Engineering Matters:
Regardless of whether you’re using local or cloud-based products, AI proficiency is key, i.e., the skill to effectively guide and interact with the AI to get the results you need. I stressed that the ability to guide AI models through well-crafted prompts is critical to getting quality outputs. - Specialized Models vs. Generalization:
There’s an ongoing debate about whether specialized small language models (“SLMs”) can outperform general-pupose large language models for specific tasks, like patent drafting. Anna von Berg and I both acknowledged the benefits of more targeted AI tools. - Style vs. Cutting-Edge Technology:
There were differing opinions on whether the focus should be on achieving a particular style or leveraging the latest AI models. Dr. Mohr advocated for staying current with the latest cloud-based models, while I emphasized the importance of having a model that consistently delivers the style of writing I need.
Conclusion: The Right AI Solution Depends on Your Needs
There’s no one-size-fits-all answer when it comes to AI in patent drafting. The choice between local and cloud-based AI depends on your individual needs, priorities, and resources.
If confidentiality and control are your top concerns, local AI models could be the way to go. But if you need scalability, flexibility, and access to the latest advancements, a cloud-based solution might be a better fit.
As this debate continues to evolve, one thing remains clear: AI will continue to play a pivotal role in the future of patent drafting, and AI proficiency is a future skill every patent attorney needs to master. Regarding the specific tooling, it’s up to each of us to find the right balance that works for our specific workflows.
Stay curious, and never stop learning!
Bastian
1 thought on “Cloud vs. Local AI: What’s Best for Patent Attorneys?”
Comments are closed.