Inside the Man vs. Machine Hackathon

Then there’s Eric Chong, a 37-year-old who has a inheritance successful dentistry and antecedently cofounded a startup that simplifies aesculapian billing for dentists. He was placed connected nan “machine” team.

“I'm gonna beryllium honorable and opportunity I'm highly relieved to beryllium connected nan instrumentality team,” Chong says.

At nan hackathon, Chong was building package that uses sound and look nickname to observe autism. Of course, my first mobility was: Wouldn’t location beryllium a wealth of issues pinch this, for illustration biased information starring to mendacious positives?

“Short answer, yes,” Chong says. “I deliberation that location are immoderate mendacious positives that whitethorn travel out, but I deliberation that pinch sound and pinch facial expression, I deliberation we could really amended nan accuracy of early detection.”

The AGI ‘Tacover’

The coworking space, for illustration galore AI-related things successful San Francisco, has ties to effective altruism.

If you’re not acquainted pinch nan activity done nan bombshell fraud headlines, it seeks to maximize nan bully that tin beryllium done utilizing participants’ time, money, and resources. The time aft this event, nan arena abstraction hosted a chat astir really to leverage YouTube “to pass important ideas for illustration why group should eat little meat.”

On nan 4th level of nan building, flyers covered nan walls—“AI 2027: Will AGI Tacover” shows a bulletin for a taco statement that precocious passed, different titled “Pro-Animal Coworking” provides nary different context.

A half hr earlier nan submission deadline, coders munched vegan meatball subs from Ike’s and rushed to decorativeness up their projects. One level down, nan judges started to arrive: Brian Fioca and Shyamal Hitesh Anadkat from OpenAI’s Applied AI team, Marius Buleandra from Anthropic’s Applied AI team, and Varin Nair, an technologist from nan AI startup Factory (which is besides cohosting nan event).

As nan judging kicked off, a personnel of nan METR team, Nate Rush, showed maine an Excel array that tracked contestant scores, pinch AI-powered groups colored greenish and quality projects colored red. Each group moved up and down nan database arsenic nan judges entered their decisions. “Do you spot it?” he asked me. No, I don’t—the mishmash of colors showed nary clear victor moreover half an hr into nan judging. That was his point. Much to everyone’s surprise, man versus instrumentality was a adjacent race.

Show Time

In nan end, nan finalists were evenly split: 3 from nan “man” broadside and 3 from nan “machine.” After each demo, nan crowd was asked to raise their hands and conjecture whether nan squad had utilized AI.

First up was ViewSense, a instrumentality designed to thief visually impaired group navigate their surroundings by transcribing unrecorded videofeeds into matter for a surface scholar to publication retired loud. Given nan short build time, it was technically impressive, and 60 percent of nan room (by nan emcee’s count) believed it utilized AI. It didn’t.

Next was a squad that built a level for designing websites pinch pen and paper, utilizing a camera to way sketches successful existent time—no AI progressive successful nan coding process. The pianist task precocious to nan finals pinch a strategy that fto users upload soft sessions for AI-generated feedback; it was connected nan instrumentality side. Another squad showcased a instrumentality that generates power maps of codification changes: captious information issues show up successful red, while regular edits look successful green. This 1 did usage AI.