Founders are raising too much money đź’¸, the auctioned AI art controversy, and another look at the trolley problem
November 4, 2018 Issue
The AI and Robotics Newsletter from Comet Labs. Subscribe here. Past issues here.
đź‘‹ Hey there,
As you’re reading this, a few of us at Comet are 30,000 feet in the air heading to Shanghai for the first stop on the Assessing China — Chipsets program. Check out the six semiconductor startups that will be joining us:
Decawave — Deep Vision* — Esperanto* — Eta Compute — GreenWaves — Oculii*
* denotes a Comet portfolio company
We’ll let you know about the ideas and challenges we explore with the startups and our partners in China in the next newsletter. We’re excited to share what we learn with you!
2019 is right around the corner, and we’re already planning 4 more industry-focused programs in China for the year. If you want to learn more and be involved in the next one, head over to the website here and drop us your email address.
Have a great Sunday, and enjoy these reads from the team!
Follow Comet Labs for more stories and thoughts about the impact of AI and robotics.
đź’¬ THIS WEEK AT OUR LUNCH TABLE:
In yet another AI-related controversy, a piece of artwork was recently sold on an auction that was produced using a neural network based on open-source code. It’s a fascinating story, and brings up a very important question:
If open-source code is used to make a profit, does the original author need to give permission?
Yes. The author should be asked first, and deserves a cut of the profit. |
No. This is exactly how open-source code is meant to be used. |
It depends on what the code is being used for. |
This isn’t the right question. (reply and let us know!)Pepper the robot’s testimony was the subject of last week’s poll, and you’ve spoken: it was all just a publicity stunt. We agree and hope that the robot doesn’t come back to haunt us later.
đź“š FEATURED ARTICLE:
How Three French Students Used Borrowed Code To Put The First AI Portrait In Christie’s — The Verge
At the intersection of technology, art, law, and ethics, you’ll find this wild story. The Verge breaks down the situation well, though reading the tweets from the original code author and replies from the artists add even more interesting context. Check them out here, and let us know what you think in the poll above.
đź“ą WATCH THIS:
đź“– WORTH READING:
Why Founders Should Be Cautious About Over-Raising — Atrium Blog
Ethan Kurzweil, Partner at Bessemer Ventures, discusses the dangers to founders of over-fundraising, despite how tempting additional cash might seem. Typically, it leads to wasteful spending and an inflated sense of product-market fit.
No, A.I. Won’t Solve the Fake News Problem — NY Times
Deciding whether or not a story is likely true is a cognitively intense task involving inference, correlation and causation analysis, and context clues, all of which are currently very difficult to model with AI. Drs. Gary Marcus and Ernest Davis explain more here.
The future of photography is code — TechCrunch
While we may be hitting a relative plateau in the innovation of camera hardware, computational software on the inside might still lead to photography breakthroughs with algorithms for object detection, tracking, and other optimizations. Learn more about the specifics of modern camera techniques here.
Look to Africa to advance artificial intelligence — Nature
The entrepreneurial ecosystem in Africa has been expanding rapidly, and now Google is opening an AI lab in Ghana to help give founders and engineers even more resources they need to succeed in creating AI applications. Moustapha Cisse, the Ghana lab’s director, highlights the remaining challenges that still need to be addressed.
Haste Makes Waste — Collaborative Fund
With the sharp increase in VC check sizes in the past few years (thanks in part to SoftBank) startups are more than ever being encouraged to grow as rapidly as possible. But with rapid growth comes a huge risk of losing the product.
Driver monitoring systems are here — and so are privacy concerns — Axios
Given so many mishandlings of data in recent past, the concern for privacy must be top of mind for any developer of a computer vision application, particularly when it involves a camera pointed directly at someone’s face and recording.
AI Researchers Fight Over Four Letters: NIPS — WIRED
The #ProtestNIPS hashtag has gained a lot of traction in the past week with some of the top AI researchers jumping in, including Google Brain’s Jeff Dean. Even so, the board of the most popular AI conference is remaining hesitant to change its name.
Should a self-driving car kill the baby or the grandma? Depends on where you’re from. — MIT Technology Review
A new study compares answers to the famed trolley problem with regard to autonomous vehicles between populations from different countries. Whether or not you agree that the question is relevant to AVs, the answers are a fascinating glimpse into the nuances of cultural influence.
Waymo, take the wheel: Self-driving cars go fully driverless on California roads — TechCrunch
Though you won’t find them wandering the streets of San Francisco yet, this is still big news for Waymo and AVs. The company is permitted to test the cars in essentially any condition at speeds of up to 65mph.
đź‘Ź PORTFOLIO NEWS:
Grocers Enlist Robots to Chase E-Commerce — Simbe
In a story from the WSJ that explores the ways grocery retailers are integrating robots into their operations, check out the profile of Simbe’s partnership with Schnuck’s Market.
Doc.ai acquires Crestle.ai to help physicians join the AI revolution — Doc.ai
AI-driven medical data management company Doc.AI just acquired Crestle.AI, a platform for quickly deploying deep learning packages and used by over 11,000 data scientists. The goal of the acquisition is to enable more medical practitioners to learn and create AI models.
No Headset Required: Lightform Is AR In The Real World — Lightform
Check out this profile of Lightform from Forbes and learn from the company’s founders about what inspired the idea for an AR projector and how they envision it will enhance displays everywhere.
đź“ť READ THIS PAPER:
Intrinsic Social Motivation via Causal Influence in Multi-Agent RL — Jaques, et. al
This fascinating study looks at how multiple agents influence each other’s actions in challenging “social” situations without knowing each others’ reward functions. The team suggests that this approach more closely mimics real-world scenarios in which individuals try to cooperate to achieve a common goal. They identify social interaction as a key under-explored area of intrinsic motivation (or an agent’s general reward function) and the results have been very promising. In their constructed environment, agents are rewarded for influencing another’s actions with the hypothesis that it would increase information sharing and cooperation. As a result, multiple agents all showed better performance toward a common goal when they were rewarded for influence, especially when able to use explicit communication to convey messages to one another.
Bonus research: Reinforcement Learning with Prediction-Based Rewards — OpenAI
Interesting things happen when agents rewarded for curiosity (and without knowing an objective function) and told to explore video game scenarios. Check out the videos!
đź› STUFF YOU CAN USE:
Horizon: The first open source reinforcement learning platform for large-scale products and services — Facebook AI Research
In this post, Facebook also mentions that they have already deployed Horizon to try and improve how users receive notifications, suggestions from M (the company’s AI messenger bot), and other applications.
100 Brilliant Women in AI Ethics to Follow in 2019 and beyond — Medium
This list is an amazing resource where you can find women who are researchers, engineers, and thought leaders working to examine the ethical implications of AI technology. Follow them!
đź“Ł HIRING:
Mobile Engineer — doc.ai
The team at doc.ai is on a mission to revolutionize the use of healthcare data with blockchain and AI, and they’re looking for a mobile engineer to join the team.
Several Positions Available — Creator
From engineering to design, on both hardware and software teams, Creator is looking for talented folks for many different roles.
Several Available Positions — Grabango
Grabango, our portfolio company providing technology for checkout-free shopping, is looking for new members to join its stellar team.
And that’s it for this week! Find us on Twitter to stay updated with the latest news in AI and robotics.
Subscribe here to get this newsletter in your inbox and participate in the poll!
