Should We Fear Gen AI? Promises, Pitfalls and the Path Forward | IPWatchdog Unleashed

This week on IPWatchdog Unleashed my conversation is with patent attorney Wen Xie. Many of you probably already know Wen. If you’ve been to IPWatchdog events, you’ve undoubtedly met her already, and you’ve heard her speak on our stage and in our webinars about various topics related to patent practice and procedure. Wen has more than a decade of experience as a U.S. patent attorney specializing in software and artificial intelligence (AI) and she recently founded Lex Lumen Intellectual Property law firm.

We begin our conversation by giving Wen an open mic opportunity to share her thoughts on the state of the industry.

“You asked me just now before we started filming, should we be afraid of AI,” Wen begins. “And my answer is there’s no point in being afraid because it’s coming… AI is coming in every profession, every technology. And we shouldn’t resist it when it comes to patent drafting or patent prosecution.”

Wen also mentioned the need to carefully and thoughtfully prompt AI tools, which we come back to again later in our conversation. But before we get too far away from it, I wanted to dive into the question about whether we should be afraid of AI. Depending on the day, I give a different answer myself. I do think if we are not afraid of AI we should at least be enormously respectful because while governments are starting to regulate use and companies are starting to adopt policies and frameworks, nobody seems to be taking the most basic first step with respect to safety.

Every science fiction aficionado is familiar with Isaac Asimov and his three rules for robots, one of which was simply do no harm to humans. And it just strikes me that AI is potentially a Dr. Evil type technology that could go wrong. So, it seems odd to me that we are not doing more to set up guardrails and categorically require AI to do no harm to people as a foundational building block for everything that will follow.

“A hundred percent,” Wen said in agreement. “And it could harm in, frankly, so many ways. It could be used in so many ways. But from what I can just name off the top of my head, it could be an echo chamber that magnifies wrong beliefs very easily. The bias thing is a huge problem because I almost feel like it could sense what you want to hear. And it reinforces that… And it wants you to use it.”

From here our conversation takes a turn toward innovation and patents, and whether AI can contribute originality at this point in its development that would be properly considered conception within the legal definition of inventing. We also discuss the importance of prompting, and how it can and should be used to streamline patent practice and drafting.

According to Wen, the key is iteratively asking Gen AI for things piece by piece and not just asking it to do something generally, or broadly, and expect useful results. Because the bigger the piece of drafting a Gen AI tool provides, the more irrelevant it seems to be, at least currently. So, coaxing AI step by step is the solution.

“You don’t have AI just start describing stuff,” Wen explained. “It’s piece by piece by piece by piece. It’s harnessing it. And what I’m hoping is that you actually don’t get a big output. I’m hoping for snippets, small outputs consistently… smaller pieces of output that you could piece together.”

We end our conversation with Wen describing what she calls S-C-O-P-E™, which is a five-step, attorney-guided GenAI framework designed specifically to address the challenges of building a structured and layered patent application by blending the speed of generative models with the analytical rigor of regimented attorney oversight.

More IPWatchdog Unleashed

You can listen to the entire podcast episode by downloading it wherever you normally access podcasts or by visiting IPWatchdog Unleashed on Buzzsprout. You can also watch the video on the IPWatchdog YouTube channel. For more IPWatchdog Unleashed, see below for our growing archive of previous episodes.