"Expert opinion" on AI gets treated in coverage as a fairly uniform category, but the underlying expertise it draws on varies enormously in ways that matter for how much weight a given opinion should carry on a specific question. A leading AI researcher's opinion on a technical capability question and the same researcher's opinion on enterprise deployment economics are not equally well-grounded, even though both get labeled expert commentary.
Why expertise doesn't transfer cleanly
Deep technical expertise in model architecture or training doesn't automatically confer equally deep insight into how a model performs inside a messy enterprise data environment, what it costs to operate reliably at production scale, or how organizational change management affects whether a deployment succeeds. These are different bodies of knowledge, developed through different kinds of experience, and a researcher without direct deployment exposure is, on deployment-specific questions, in a genuinely different epistemic position than someone who has actually run production AI systems inside a large organization.
What to check before weighting an expert opinion
The most useful diligence question for any quoted expert opinion is whether the specific claim being made falls within the domain the expert has direct, demonstrated exposure to, or adjacent to it. A quote from a well-credentialed AI researcher about the likely trajectory of enterprise AI adoption carries real weight if that researcher also has direct exposure to enterprise deployment, and considerably less if their expertise is purely in model research with no operational exposure.
The incentive to blur this distinction
Coverage benefits from strong, quotable opinions from credentialed sources, which creates some incentive to solicit and publish expert commentary on questions somewhat outside a given expert's core domain, because their credentials still lend the quote authority even where their specific expertise doesn't fully support it. This is worth resisting, even where it means a less impressive-sounding quote from a more directly qualified but less famous source.
What we try to do
Our approach to expert commentary is to match the question to the specific domain of the expert's demonstrated exposure, and to be explicit in framing about the basis of a given opinion, rather than letting general credentials stand in for domain-specific expertise they may not actually cover.