A recent study supports the notion that using AI for drafting–something lawyers are eager to do–can effectively make you stupid over time. “Over four months, LLM [large language model] users consistently underperformed at neural, linguistic, and behavioral levels,” including having difficulty recalling their own work, compared to “brain-only” users and those using search engines. These users were all drafting written products.
There’s been some pushback, including the charge that the study “looks only at the downside of large language models (LLM) and rules potential benefits out of consideration.” One benefit is that reducing one’s “cognitive load” frees time to do other more important or more enjoyable things, which some contend is the real measure of the usefulness of LMM. Although even the doubters question the use of fully LLM answerable questions being posed in educational settings. Perhaps that only encourages hoovering up data and not learning how to think critically.
Back to lawyers using LLM to draft. Given the high rate of mistaken information, including nonexistent cases, LLM should probably be used with caution–perhaps providing an initial draft but one that is then thoroughly reviewed and ingested as to make it your own.