It’s not all slop: how to overcome AI’s problems

🤖 OpenAI released ChatGPT to the general public 595 days ago
Issue #121: the definition of AI slop, armchair activism, and a better tomato sauce
By Zulie @ Medium

Have you seen Shrimp Jesus? If so, you’ve been a victim of AI slop. The term “slop” was popularized on platforms like X (see an example here) and later defined by web developer and programmer Simon Willison as AI-generated content that was “mindlessly generated and thrust upon someone who didn’t ask for it.”

Beyond slop, many of us are familiar with how, for assistive technology, AI is sometimes more of a hindrance than a help. To find better ways to develop AI, researchers can look to the disability community. As Assistant Professor of Computer Science Elaine Short writes, disabled people are very used to dealing with “helpful” experiences that are annoying and useless. She cites the common experience people in wheelchairs have of able-bodied people grabbing wheelchair handles to push them without being asked.

Short explains how the disabled community offers a model of assistance as a collaborative effort. Applying this to AI development can help to ensure that new AI tools support human autonomy rather than taking over. For example, she found that when people use a joystick (limited in motion) to move a robot arm (much less limited), they struggle. But when they collaborate with an AI program that predicts what they want to accomplish based on joystick movements, it works much better.

What about addressing bias in AI, one of the better-known issues plaguing the technology? I used to agree with the standard advice: advocate for better representation in the datasets. But writer Nidhi Sinha calls that attitude out as ineffectual and surface-level (anyone remember Google Gemini’s AI-generated Black founding fathers?), and likely harmful towards vulnerable communities.

For instance, Sinha highlights the Detroit Police Department’s Project Green Light, a mass surveillance project powered by AI facial recognition, as an area where adding more Black and Brown faces to the datasets would not accomplish anything other than wrongful arrests. By contrast, the community-led response, Green Chairs, Not Green Lights, focuses on “what the actual families residing in Detroit need to feel safer in their neighborhood.” The GCGL project runs educational programs about how surveillance doesn’t equal safety and offers green chairs to community members as a way for them to signal that they’ve committed to looking out for their community.

Sinha’s final point brings it home: “The most uncomfortable sentiment in Responsible AI is that sometimes AI should just not be used at all.”

What we’re reading

• I keep coming back to this piece from Joe Guay about wanting to be an activist, but never having attended any protests or marches. Ultimately, he calls himself out for a lack of authenticity and fear of confrontation and scrutiny. His story asks an important and relatable question: What’s enough, in activism?
• Autistic writer Elanor Rice explains how mimicry, a common tactic autistic people use to “mask” (AKA camouflage autistic traits), helped her learn flawless Spanish. “My autistic mimicry allows me to switch accents like I switch phone cases.”
• While dreams can’t predict the future, health writer Kathleen Murphy explains how analyzing dreams can teach you about yourself. One theory I enjoyed: Dreams can be your subconscious engaging in threat simulation, like a practice run of a stressful event.

Your daily dose of practical wisdom: buy whole canned tomatoes

Ever wondered why your homemade spaghetti sauce won’t cook down? Pre-diced canned tomatoes may be the culprit. From cookbook author Dim Nikov in a story with 34 additional cooking tips: “Diced canned tomatoes don’t cook down because they contain calcium chloride, a firming agent that keeps them from turning to mush during canning. Buy your canned tomatoes whole; you can chop them up with a sharp knife or crush them with your hands.”

Did someone forward you this email? Learn something new every day with the Medium Newsletter. Sign up here.

What do you think of this newsletter? Take this super quick survey to let us know.

Edited and produced by Scott Lamb & Carly Rose Gillis

Questions, feedback, or story suggestions? Email us: tips@medium.com