Empire State of Mind

A memory. An open day at the Electrical Engineering department at Imperial College for accepted candidates, the professors wanted to demonstrate their new face recognition software to the group. I demurred (shyness), but they insisted that I be the volunteer.

So I sat and watched for 15 minutes as nothing worked. "Smile", asked the graduate students. "Frown", asked the professors. Same result: no detection. Others were summoned. Invisible man.

It's a minor point, but I can't forget the sheepish grins. As I told the professors and their harried students: "You probably didn't train your algorithms much with many faces like mine."

Cultural sensitivity in technology is one of my perennial themes. And it's hard work even if you acknowledge your blind spots. The anecdotal failures continue to pile up. What was true back in 1991 is much the same in 2021, software and hardware are far more sophisticated and performant but face the same blindspots (provenance of training data, applicability to real world scenarios, ethical framing etc.)

The challenge for software engineering - which is still a craft, is to move beyond curve-fitting phrenology (and Deadwood) into its industrial revolution.

Sidenote: recruiters these days are all "big data, machine learning, cloud yada yada". Buzzword fatigue is an occupational hazard.

I miss the great mass amateurization and view source ethic of early web development and yet the developer tools and frameworks these days almost feel like a golden age is within grasp.

Obligatory citations:

• Planning the Software Industrial Revolution by Brad Cox
• Shop Class as Soulcraft by Matthew B. Crawford
• Commodifying AI, or, Some Half-baked Thinking on Services As Middle-men by Coté

Incidentally, Imperial College was the venue of my worst interviewing experience (and there have been many) - the low point of which was walking past the open door as I left the interview, and walking straight into the wall. This was after having flubbed almost every question I'd been asked.

The laughs and looks exchanged by the professor and the secretary as I turned around, shuffled back muttering an apology (why?) as I rubbed my sore head and headed out the door. Even English reserve and politeness could not deal with my Buster Keaton imitation.

I was admitted to Imperial College a week later.

I occasionally regret not having gone to Imperial or Cambridge for university. Without a doubt, I would be a stronger engineer and yet I suspect the eclectic toli monger you see before you would be repressed.

Imperial Visions, a playlist

A soundtrack for this note (spotify version).

• You Don't See Me by Amel Larrieux
• Invisible by Alison Moyet
• Empire Ants by Gorillaz featuring Little Dragon
• Sensitivity by Ralph Tresvant

I've been neglecting the Toli Technology Series for years now, albeit I occasionally make a few gnomic pronouncements on Twitter, consider this some throat clearing to prompt a reboot.

File under: culture, technology, algorithms, software, race, sensitivity, toli, life, best practices, design, data