Don't think of a goalpost
Welcome to Reailty Check. Thank you for reaidng!
False dichotomies and loaded questions can throttle a conversation in more than one way. Once you ask, “Will AI become Skynet?”, you’ve tarred AI with the brush of apocalypse even if the answer is No for the foreseeable future. That’s George Lakoff’s famous “elephant” problem with framing. And there’s also a narrowing/limiting problem: What if “Apocalypse or Utopia?” isn’t even a good axis for the discussion?
In a 2021 paper about AI benchmarks (e.g., standard performance tests to determine the quality of an algorithm), authors Raji, Denton, Bender et al warn about the easy blurring between “the software passed the test” and “we’re one step closer to sentient robots.” It’s a good reminder how dangerous the word “toward” is. (“Toward generative AI…”, “Toward the AI apocalypse…”, etc.)
AI is hardly the only topic in technology, or society, where our current habits of conversation don't support complexity. Some of this is that technology and politics are complicated and as audiences we crave simplicity. But some of it is a race to the bottom by people who should know better. Editors. Senators. And—as the paper and this article about it note—researchers themselves, who have understandable incentives to describe progress in simplified terms that tend to blur good results and epic futures.
“If the public at large has an inaccurate sense of what ‘AI’ can do (due to hyped results from purportedly general benchmarks), the public at large is ill-positioned to reject the application of these systems. The result: we will be stuck with systems presented as ‘objective’ (because they’re computers) which are in fact perpetuating racism, sexism, ableism, transphobia, ageism, etc. ...”
Professor Emily Bender also talks about these framing challenges in a close reading of a 2022 Steven Johnson NYT Magazine article.
What does a healthy narrative approach look like? How do we write and strategize about new tools with “narrative guardrails” that discourage extreme conclusions and make complexity more palatable?
Thank you for being a subscriber and participant! What is Reailty Check?