Follow

much like AI & HCI are “two fields divided by a common focus” (aaai.org/ojs/index.php/aimagaz), I’m gradually realizing that all the AI problems I find interesting are also fundamentally PL problems, because programming languages are how we communicate nuanced human intent to computers

hypothesis: the AI/HCI relationship is cyclic because AI booms focus on building software that’s smart enough to guess what the user wants without explicitly asking; HCI then comes in to clean up the mess, bring the human back into the loop & restore some control to users

we're in a machine learning boom right now, & ML is *all about* guessing over asking. letting the user explicitly express their intent to the computer has fallen out of style, but I expect it to come back with a vengeance before long

in the absence of a “sufficiently smart machine”, it becomes obvious that a one-size-fits-all approach – “just make the computer Do The Right Thing for everyone” – won’t work. different users & different communities have different needs, & need to be able to express them somehow

the languages we use to communicate these individualized needs to the computer won’t necessarily look like conventional programming languages on the surface, but they will be *languages* – less expressive interfaces are insufficient to capture the full range of desired behaviors

@maxkreminski A contrarian hypothesis: HCI excites the public and investors with futuristic new interfaces, which is enough to get them interested but don't *really* work (many are wizard-of-ozzed or very brittle), which then kicks off a round of AI people getting funding to build the missing backend tech.

(Works better in some cases than others. I think works especially well for the AI boom kicked off post-Sketchpad, which more or less turned into the AI EDAM community.)

@mjn honestly I think these hypotheses are fully compatible with one another! I'm giving AI a bit of a hard time here since I'm generally closer to it & it fits the narrative I'm trying to put together

@maxkreminski I agree, don't think they're really incompatible. I think there's also more complexity than 'AI' and 'HCI' here, esp. the role of computational design as a quasi-discipline. For example, imo it's a huge stretch to call Sutherland's thesis 'HCI'. It influenced later HCI, but its own methodology was really pretty un-HCI-ish (much more a mix of design studies, graphics, and GOFAI).

Sign in to participate in the conversation
Mastodon

Server run by the main developers of the project 🐘 It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!