The arrival center stage of Chat GPT and generative AI has generated a great deal of excitement in many areas of economic activity. I previously reported my observations concerning the attitude to these technologies amongst IP law practitioners based on exchanges at the American Intellectual Property Law Association (AIPLA) annual conference.
While the AIPLA annual conference is heavily attended by practitioners in private practice and industry, I am now following up with regard to Tech Transfer professionals working for US Universities and research institutions.
Interestingly, my impression is that while practitioners in this area are less gung ho concerning the adoption of generative AI in core activities such as drafting and negotiating contracts or drafting patent applications, they may already be using the technology more effectively in other areas of their daily activities.
Tech Transfer has a promotional aspect that is outside the experience of many IP professionals, and it is in this area where Tech Transfer professionals seem already to be employing generative AI to good effect. Generative AI provides an efficient mechanism for producing general articles on uncontroversial subjects, or a rough first draft of a document, that can then be adjusted by a qualified human. Tech Transfer professionals report developing sophisticated prompt engineering expertise for these applications, and recommend in particular defining carefully the form of the required output in prompts, including the level of experience or technical expertise of the intended recipient, the length of sentences and incorporating one, or preferably multiple examples of texts in the required style. Similarly, it is suggested to include restrictions as the eligible sources of material to be incorporated in the response as appropriate. The availability of plugins for the generation of Powerpoint Presentations in a defined house style, or graphical tools such as Adobe Firefly are mentioned as particularly valuable in this context.
It has been found that different Generative AI tools often produce results of different quality for a given query, even if they work on the same underlying technology (e.g. Chat GPT and Bing), and it is recommended to try the same prompt with different tools, and to choose the best response. Chat GPT in particular is noted to tend towards a promotional style, singing the praises of whatever the subject of a text might be, and it is often necessary to tone down this orientation.
On the other hand, Tech Transfer Professionals seem to take a conservative stance on many points. Although as reported in my previous article many view restrictions on data retention by Generative AI tools as the silver bullet in addressing confidentially concerns, in the Tech Transfer field, the general attitude still seems to be that any exposure of confidential information to an online Generative AI platform is out of the question. Insofar as Generative AI is used to generate a published text, it is considered generally appropriate to indicate this as an attribution in the text. In my previous article, the availability of Local LLMs is mentioned as a solution to many of the IP problems associated with Generative AI, in particular since they can use an entirely open source model, and operate in isolation. Tech Transfer professionals report real experience with these technologies, made possible through the technical resources of the Universities or Research bodies they work for, or through hosting platforms such as HuggingFace, yet remain circumspect with regard to the use of such tools in their daily work, in particular due to the unknown provenance of the training dataset upon which the model is based, and the copyright risks that is perceived to raise.
In conclusion, it seems that while perhaps approaching legal risks from a different point of view, Tech Transfer professionals are quietly gathering actionable expertise that will no doubt be of value to the whole IP world.