DataDownload: Tim Berners-Lee plans to save the web & NYC is hiring a Chief Algorithm Officer

Some weeks, editing the newsletter can be harder than others. This week was pretty amazing. Ok, it’s an old trope by now — but how can you not be thankful for Tim Berners-Lee. At a time when the World Wide Web needs a compelling conscience — here he is. Bravo. And New York’s Mayor steps up — calling for a Chief Algorithm Officer. Again, a good thing. And the NY Times does a solid job of making the real issues around Deepfakes understandable for non-technical folks. And our friends at TED publish Sara-Jane Dunn’s talk on how stem cells could develop “living software.”

It’s a good week to look to the future, smile, and acknowledge that we all have work to do.

So please read, watch, comment and share. Because without you, we’re just writing words that only the robots on the web are reading.

As always, please ping me if you have thoughts or feedback about the Newsletter. Steve@NYCMediaLab.org.

Steven Rosenbaum
Managing Director
NYC Media Lab

The NYC Media Lab Must-Read

I Invented the World Wide Web. Here’s How We Can Fix It.

Director at W3 and progenitor of the world wide web Tim Berners-Lee had a vision of the internet serving humanity when we invented the web in the 80s. Instead, we got both ends of the spectrum: “projects like Wikipedia, OpenStreetMap and the world of open source software,” and “prejudice, hate and disinformation.” The web, says Berners-Lee, needs radical intervention. So he proposes a way to overcome the stalemate of governments “blaming platforms for inaction” while companies oppose regulation: a Contract for the Web.

The Contract is a global plan of action a year in the making, developed by “activists, academics, companies, governments and citizens,” outlining the steps to prevent the misuse of the web and our data. This includes eliminating incentives that reward clickbait and the spread of disinformation, platform transparency, mindful and inclusive app development, and community building. The Contract site itself is split into nine principles, each explored in depth. The contract has been signed by governments (France, Germany, Ghana), while companies like Microsoft, Reddit, and DuckDuckGo are “committing to action.”

NYC Wants a Chief Algorithm Officer to Counter Bias, Build Transparency

Mayor Bill de Blasio has issued an executive order for a new Algorithms Management and Policy Officer position, which will be responsible for “bringing leadership to algorithmic tool deployment by public agencies, while also developing and supporting policy for best practices by city staff.”

The role will interface with the Office of Information Privacy, the Office of Data Analytics and the Office for Economic Opportunity. Blasio’s stance on algorithmic regulation didn’t end there: this Monday, the mayor told the autonomous FedEx bot prowling NYC streets to get off the sidewalk.

For the Media

Deepfakes — Believe at Your Own Risk

For the latest NY Times “The Weekly” episode, investigative journalist David Barstow dived into the world of deepfakes, taking away three important points:

• The tech is moving surprisingly fast: “it’s astonishing the progress a handful of smart engineers were able to make in a matter of months.”
• Researchers around the world are racing to create techniques for detecting manipulated media, but “some deepfake creators are incorporating the machine-learning algorithms behind those countermeasures to make future deepfakes even harder to detect.”
• Big tech companies are “woefully unprepared” for deepfakes, and are consulting with outside experts to form their own policies.

Alexa’s Voice Can Now Express Disappointment and Excitement

Alexa developers will soon have the ability to add “happy/excited” and “disappointed/empathetic” tones to their skills, according to a recent announcement from Amazon. For example, someone developing a trivia game can add an excited tone when a player answer answers correctly. In January, the Alexa team launched its first intonation in the form of a newscaster voice.

The Case for Sending Robots to Day Care, Like Toddlers

Toddlers learn through play — so why not robots? Roboticist and UC Berkley psychologist Alison Gopnik believes to train smarter robots, engineers may have to instill a “childhood” period in a machine’s development. In more technical terms: create curiosity-driven agents instead of those built to copy movements or learn through random motion — “seemingly illogical fiddling that in the end lands them on an answer.”

Developing laboratory daycare will likely involve quantifying human reasoning at an early age. For Gopnik, this hasn’t been easy: “we ask them what they think about something, and they’ll give you a beautiful monologue about ponies and birthdays.” Gopnik got around this by developing custom toys: “since we’re designing the toy, we know what the problem is that the children are having to solve, and we know what kinds of data they’re getting.”

What We’re Watching

The Next Software Revolution: Programming Biological Cells | Sara-Jane Dunn

Sara-Jane Dunn, a senior scientist at the Biological Computation group at Microsoft Research, took to TED to explain how her team is leveraging embryonic stem cells to develop “living software” that could transform industries.

Events & Announcements

Event: Working with Data with the Housing Data Coalition
Date: December 3, 9AM-11AM
Learn how a group of civic technologists came together to harness open data for housing justice, and how creating a shared infrastructure to work with datasets across NYC agencies has enabled them to meet the needs of New Yorkers. Register Here.

Event: Transit Techies NYC
Date: December 3, 6:30PM-8:30PM
Sunny Ng, creator of goodservice.io, will demo a real-time map of MTA delays and service changes; Ivy Pan from Via will dive into data analysis around ridership demand; and Joshua Gee and Jay Sathe from MTA New York City Transit will talk about the work they did to roll out brand new real-time transit templates throughout the system. Register Here.

Event: Avoiding a Race to the Bottom in the Gig Economy
Date: December 5, 5:30PM-9PM
Behind each app is a global workforce of flexible workers. But the new online jobs markets and commerce platforms, and the startups and tech companies that run them, come with new rules around wages, benefits, and even safety. So, can a gig worker actually make it to the top and earn a higher income with greater stability than before? Register Here. A Deeper Look Powered by AI: Instagram’s Explore Recommender System

Behind Facebook and YouTube, Instagram is the world’s third-largest social media platform, and nobody — besides ByteDance, which has 1.5B MAUs, and Netflix — focuses more on algorithmic curation than Facebook (which owns Instagram) and Google (which owns YouTube). So it’s great to see an intelligible dive into the inner workings of Instagram’s Explore feature, including custom query languages, lightweight modelings techniques, and “high-velocity experimentation” tools. These components make up an AI system that “extracts 65 billion features and makes 90 million model predictions every second.”

Facebook uses a technique similar to word2vec called ig2vec, which infers account-level embeddings using KNN lookup and Facebook’s nearest neighbor retrieval engine, called FAISS. As The Verge summarizes the process: “the Explore system begins by looking at ‘seed accounts,’ which are accounts that users have interacted with in the past…. It identifies accounts similar to these, and from them, it selects 500 pieces of content. These candidates are… ranked based on how likely a user is to interact with each one. Finally, the top 25 posts are sent to the first page of the user’s Explore tab.”

Advanced Machine Learning Helps Play Store Users Discover Personalised Apps

DeepMind has been steadily implementing tech (ex. cooler data centers and better Android battery performance) into Google’s infrastructure. Their latest project dips into Play Store curation. DeepMind collaborated with the app store team to better determine the relevance of apps to a specific user, using a Tansformer model (typically used in NLP) for sequence-to sequence prediction, and a reranker model, which ditches the pointwise method of ranking items one a time for a model that “learns the relative importance of a pair of apps that have been shown to the user at the same time.”

Transactions & Announcements

Rad AI Raises $4M in Seed Funding

AI-powered Regtech Startup Tookitaki Secures US$19.2M in Series A Funding, Pledging to Address Global Money Laundering Issue

Stradigi AI Raises $53M CAD Series A Funding to Fuel North American Expansion

Google-Linked Fund Seeds Berkeley Medical AI Startup