DataDownload: R2D2 as a model for AI collaboration

7 min readDec 5, 2020

DataDownload: R2D2 as a model for AI collaboration A weekly summary of all things Media, Data, Emerging Tech View this email in your browser

This week, we’re looking back — and looking forward.

If you had to pick a model, what would be the best character for human-machine collaboration? Alexis Lloyd posts a fun and provocative piece that comes to an unexpected conclusion.

The author of Ready Player One, Ernie Cline — is on with Trevor Noah to consider the dangers and benefits of technology in his just released “Ready Player Two”. As if “One” wasn’t bleak enough.

And, information is coming from some strange places — including Amazon’s smell-o-vision COVID locator (of sorts). NPR’s Code Switch gets an earned appreciation this year when we’re discussing race in a new way, hopefully with meaningful outcomes.

The NYC Media Lab team is planning a calendar of events, programs, and new initiatives in 2021. So keep an eye out.

Ideas, feedback, thoughts? We always love hearing from you at Steve@NYCMediaLab.org.

Best-
Steve

Steven Rosenbaum
Executive Director
The NYC Media Lab
Steve@NYCMediaLab.org Must-Read R2D2 as a Model for AI Collaboration

“Welcome to 2020, where humans pretend to be bots who are in turn pretending to be human.”

Why is that C3PO is mildly detestable, despite being programmed to be an agreeable, intelligent humanoid? Why is Hanson Robotics’ Sophia (or its Silicon Valley counterpart, Fiona) so unnerving? VP of Product Design at Medium Alexis Lloyd has an idea. Actually, a framework for thinking about the different ways we can interact and collaborate with machines, segmented into three archetypes: C3PO, Iron Man/Jarvis, and R2D2.

C3PO: As early as the 1920s, we’ve been sidestepping conversations with awkward humanoids.

A century later and we’re still perturbed by the likes of Sophia. According to Lloyd, the issue is, in part, due to our own difficulties with interacting with each other. “It’s hard to program a computer to [interact like a human] in any satisfying way because in many ways it’s hard for humans to do successfully either…. We don’t all talk to each other the same way. We don’t all have the same set of cultural backgrounds or conversational expectations.”

Iron Man/Jarvis: Stark’s archetype is a human augmented by machine — initially to survive and later to gain superhuman abilities. We see medical devices like cochlear implants or prosthetic limbs or BCIs being reimagined as augmentations: “someone with a prosthetic hand who has a USB drive in his index finger, or runners who have prosthetics designed to run more efficiently than a biological foot.” But augmentation isn’t necessarily collaboration — for that, we look to the R2D2 archetype.

R2D2: The little robot’s beeps and boops comfortably fit into what we imagine as “robot language,” despite being functionally useless (as an aside — check out Timo Arnall’s 2012 clips of how robots equipped with computer vision “see”). The droid is useful precisely because it can do the things we cannot, instead of trying to mimic human qualities: machine-to-machine communication, diagnostics, repairs, and generally acting as a user-friendly interface between complex technology.

10 min read

Polis director Charlie Beckett sat down with R&D Chief and Product Lead at WSJ Alyssa Zeisler for his interview series with women working at the intersection of AI and journalism. Zeisler’s dual role is a reflection of WSJ’s own integration of its R&D and newsroom tools departments: “Bringing these teams together was an opportunity to embed the R&D approach into our legacy systems and operations, and at the same time bring in a little bit more product thinking and tooling to R&D.”

The excellent Talk2020 tool we previously featured is one example of a newsroom- and user-facing product to come from the hybrid department. A few others: “a tool that merges different levels of geometry was used in this recent piece to connect previously non-comparable regional data and in this piece to analyse the impact of Black Banks in the US. A new approach to information and data gathering was used in this piece showing that Google is giving preference to YouTube over other video sources.”

7 min read

Read More Tech+Media Amazon Reviews Complaining About a ‘Lack of Smell’ From Scented Candles Appeared to Increase as Coronavirus Cases Spiked Correlation doesn’t equal causation, but you have to wonder… researcher Kate Petrova found that Yankee Candle reviews on Amazon that mentioned a “lack of smell” grew from 2% in January to 6% in November. The tweet blew up, but Petrova herself warned that “we need to be careful in interpreting these numbers.”

2 min read Read More Mapping Black Media

As part of the Center for Community Media’s Black Media Initiative, Newmark J-School released a map of nearly 300 media outlets that serve primarily Black communities. “Black media has been historically classified as newspapers — often weekly — owned by an African-American publisher that print in cities with significant Black populations. We’ve included those… alongside digital startups that cover niche topics, publications that serve West Indian and African audiences, radio stations, nonprofit newsrooms and more.” Find the Airtable database here.

1 min read Read More Measuring Risk-Taking by Watching People Move Computer Mice

In a recent experiment, researchers found that the way people’s mouse drifted towards a safer gambling option on-screen — “even when they ended up taking the risky bet” — hinted that they were “more risk-averse than their choice would indicate.” In other words, your mouse movements are a more nuanced metric for risk-taking, since the final choice is only binary — you either make it or you don’t.

In one analysis, even a set of people who made the exact choices had different inclinations towards risk based on their mouse movements. It’s not difficult to imagine AI-backed hiring platforms like HireVue incorporating this data into their automated candidate assessments.

4 min read

Trevor Noah features Ready Player One author Ernest Cline, who discusses the themes in his sequel, Ready Player Two.

8 min watch

Watch Now What We’re Listening To Podcast: Code Switch

If you haven’t listened to NPR’s Code Switch yet, it’s “the fearless conversations about race that you’ve been waiting for! Hosted by journalists of color, our podcast tackles the subject of race head-on. We explore how it impacts every part of society — from politics and pop culture to history, sports and everything in between.”

Listen Now Virtual Events Free Event: Ask Me Anything Live — Steve Case, Co-Founder of AOL
Date: December 8, 1PM-2PM EST
Join virtually for an afternoon with Steve Case, Co-Founder of AOL and Chairman & CEO of Revolucion LLC, as part of Betaworks Studios series, “Ask Me Anything Live.” Register Here.

Free Event: How I Made My First Data-Driven Decision
Date: December 8, 5PM-5:30PM EST
Product Manager at Google Sasha Latsenia will discuss balancing knowledge and personal experience with what’s best for a wider user group. Register Here. A Deeper Look ‘It Will Change Everything’: DeepMind’s AI Makes Gigantic Leap in Solving Protein Structures

At this year’s Critical Assessment of protein Structure Prediction (CASP) competition, DeepMind’s protein-folding-prediction model, AlphaFold, has made a leap equivalent to the 2012 ImageNet moment. The last time AlphaFold participated in the biennial protein folding competition was in 2018, where it surprised researchers, but used an approach that was similar to many other teams. This year, AlphaFold made a “gargantuan leap” in performance.

John Moult, who co-founded CASP in 1994, said the results are “a big deal. In some sense the problem is solved.” The problem, in this case, is predicting how proteins fold: “Proteins are the building blocks of life, responsible for most of what happens inside cells. How a protein works and what it does is determined by its 3D shape — ‘structure is function’ is an axiom of molecular biology…. Accurately predict protein structures would vastly accelerate efforts to understand the building blocks of cells and enable quicker and more advanced drug discovery.”

Andrei Lupas, an evolutionary biologist, says AlphaFold’s latest results are a game-changer: “This will change medicine. It will change research. It will change bioengineering. It will change everything.” Structural biologist Janet Thornton hopes “the approach could help to illuminate the function of the thousands of unsolved proteins in the human genome, and make sense of disease-causing gene variations that differ between people.”