Copy
Machine Learnings

Awesome, not awesome.

#Awesome
"The Library Innovation Lab at the Harvard Law School Library has completed its Caselaw Access Project, an endeavour to digitize every reported state and federal US legal case from the 1600s to last summer. The process involved scanning more than 40 million pages...One of the biggest hurdles to developing artificial intelligence for legal applications is the lack of access to data. To train their software, legal AI companies have often had to build their own databases by scraping whatever websites have made information public and making deals with companies for access to their private legal files...Now that millions of cases are online for free, a good training source will be easily available.” - Erin Winick, Editor Learn More from MIT Technology Review >

#Not Awesome
"I don't think necessarily that there are people at Amazon saying, 'let's not deliver to black people in Roxbury," Gilliard said. "What typically happens is there's an algorithm that determines that for some reason not delivering there made the most sense algorithmically, to maximize profit or time… And there are often very few people at companies that have the ability or the willingness or the knowledge to look at these things and say, 'hey, wait a minute. While these decisions are often made by AI algorithms, that doesn't mean humans aren't responsible for the results. Gilliard said that when he sees the sort of AI algorithms Amazon and others use, "...my antenna sort of go up, because much of that is based on training data that… probably [reflects] the biases that are already built into society." - Learn More from CBC Radio-Canada >

What we're reading.

1/ Researchers use AI models to run simulations of real-life social problems to better understand how religious violence can break out. Learn More from Motherboard >

2/ If we don't have a global conversation about how to build algorithms that make "split-second decisions that will result in life or death," we will introduce software that changes the physical world in ways that violate the value systems of different regions. Learn More from Quartz >

3/ When companies optimize algorithms to extract as money from consumers who are most willing to pay, anyone can becomes a victim - from elderly people with dementia to the "rich and busy" who don't monitor their receipts. Learn More from Tim Harford >

4/ The days of self-driving cares are almost upon us - when they finally arrive we'll have tens of thousands of algorithm trainers in Kenya to thank. Learn More from BBC News >

5/ Deep learning algorithms make it possible for researchers to pinpoint signals of natural selection within specific regions of people's genomes in ways that it's never been studied before. Learn More from Nature >

6/ One of the biggest potential risks of unchecked AI algorithms is the stripping of people's political agency. Learn More from openDemocracy >

7/ For autonomous vehicles to make it onto the roads, they'll need to prove to federal regulators that they're much better than humans at driving in every possible situation. Learn More from WIRED >

What we're building.

Meet Journal
We started Journal excited to answer a single question.

What would be possible if our information — about people, projects, and ideas — was connected and easily accessible?

In a year after closing our Seed funding round, we’ve started moving towards an answer. We are building a new kind of Journal. You write notes in it, save interesting links, and drop in important documents and messages for later. When you need something, ask Journal, and it will actually help you find it.

Journal integrates with the services where your information lives (like Slack, Gmail, Evernote, Pocket, and Dropbox) — so that you have one connected home for all your stuff. You can use the best services for messaging, documents, and more — and Journal will tie them all together so you stay in control.

For the past few years we’ve seen ourselves, friends, and colleagues stretched thin trying to manage many different silos of information. Journal is the new way forward. It’s a connected home to enhance the way we gather and share knowledge.

So far, beta community members have used Journal to launch new products, prepare for a new baby, remodel a home, write a book, teach a graduate course, plan a vacation, prepare for meetings, and much more.


On the surface, Journal looks and feels simple.

Beneath the surface, the Journal platform has two unique attributes that will enable more flexible and personalized product experiences than we’ve seen from existing knowledge management services:

  1. A state-of-the-art machine learning and natural language processing model that conceptually understands people’s digital information across formats
     
  2. An architecture and UX that understands and shows information as distinct types (e.g. files, tasks, articles, messages, products, and more)

Today, these attributes mean less friction for people interacting with their information in Journal. You organize links, emails, and other types of information without losing visual context, find anything easily across the apps you integrate, and see relevant items automatically grouped alongside your contacts and events. They are also the building blocks to make Journal the connected home for people to gather their thoughts, get organized, discover and share knowledge. We will use them to empower people to harness their information rather than to be burdened by it.

In the future, Journal will change depending on where you are and what you’re doing. When you check your phone from bed, it may look like a blank pad to dump thoughts. When you’re assigned a task at work it could become a list of suggested resources for you and your team. On the weekend, it may just show a map of nearby bakeries and long-form articles recommended by friends — it will help you do more with your information no matter the shape it takes.

We are so grateful to the nearly 2,000 beta users from our Machine Learnings and Noteworthy communities whose feedback has helped shape our product. In recent weeks we’ve observed an inflection point in deepening engagement, retention, and enthusiastic user feedback — So, starting tomorrow, we’re going to share Journal for Mac, our Web app, and our Chrome extension beyond our own communities. Our iOS app will follow in the coming weeks.

It’s been a hell of a challenge building Journal so far, and we have a lot of work ahead of us. We’ll only be able to make our vision for Journal a reality with the contributions of passionate, smart teammates and a supportive community. If you’re a distributed systems or machine learning engineer interested in building the future of knowledge management, please say hello. If you’d like early access to Journal, add your name to our waitlist >.

We would love for you to come a long for the ride.

Samiur Rahman
CEO, Co-founder
Journal

Links from the community.

"If I Can You Can (and you should!)" submitted by James Dellinger (@jamrdell). Learn More from Noteworthy >

"An AI physicist can derive the natural laws of imagined universes" submitted by Samiur Rahman (@samiur1204). Learn More from MIT Technology Review >

"Scaling Machine Learning at Uber with Michelangelo" submitted by Avi Eisenberger (@aeisenberger). Learn More from Uber >

First time? Subscribe to Machine Learnings 🤖
04/11/18 View this email in your browser
Copyright © 2018 Machine Learnings, All rights reserved.


Done learning about AI?
Opt out of emails