AI Search Concerns

Recently I had the opportunity to test-drive an artificial intelligence (AI) engine used for prior art reference searching — specifically for uncovering relevant patents based on natural language input combined with emphasis on select keywords.

Like the majority of AI search engines presently available, there is one main concern which may be summed up in a single word: Trustworthiness

Though many of the returned references appeared appropriate and relevant, and though there are ratings associated with each returned result indicating relevance, there is a lack of trust that does not exist when undertaking a search oneself.

Perhaps it is the “black box” concept of AI searches, where the human searcher does not have a full understanding of the code and processing behind the search engine.

Perhaps it is the lack of control over traditional boolean searching, control over the exact terms being used for searching and how those terms are being used.

What can be said is that AI search engines have yet to return results which instill confidence in the generated results when compared to results compiled by a knowledgeable human searcher using more traditional tools such as the USPTO’s Patent Public Search.

(Even Google Patent Search tends to return results more relevant and efficiently than AI engines.)

The main concern here is that searchers, legal firms, and inventors are going to rely on these new technologies to return somewhat satisfactory — and possibly erroneous — search results, which will in turn cause extended resource spending in terms of both time and money.

Not that one should disregard AI searching altogether.

A more shrewd approach would be to utilizing artificial intelligent search as an additional tool in the search process, incorporating AI results as one element of a well-rounded whole.