Machine Intelligence Research Institute, Jessica Taylor
from
At the EA Global 2016 conference, I gave a talk on “Using Machine Learning to Address AI Risk”:
It is plausible that future artificial general intelligence systems will share many qualities in common with present-day machine learning systems. If so, how could we ensure that these systems robustly act as intended? We discuss the technical agenda for a new project at MIRI focused on this question. [video, 15:04]
Increasingly, preprints are at the center of conversations across the research ecosystem. But disagreements remain about the role they play. Do they “count” for research assessment? Is it ok to post preprints in more than one place? In this paper, we argue that these discussions often conflate two separate issues, the history of the manuscript and the status granted it by different communities. In this paper, we propose a new model that distinguishes the characteristics of the object, its “state”, from the subjective “standing” granted to it by different communities. This provides a way to discuss the difference in practices between communities, which will deliver more productive conversations and facilitate negotiation on how to collectively improve the process of scholarly communications not only for preprints but other forms of scholarly contributions.
TomTom is working with us at Comet Labs with the aim of actively engaging with the Bay Area startup and technologist ecosystem. We are incredibly excited about this partnership and are actively looking to work with startups who will benefit from their mapping data and expertise.
The following is an interview with Dan Weld about the Beneficial AI 2017 conference and The Asilomar Principles that it produced. Weld is Professor of Computer Science & Engineering and Entrepreneurial Faculty Fellow at the University of Washington.
Q: From your perspective what were the highlights of the conference?
“One of the highlights was having the chance to interact with such a diverse group of people, including economists and lawyers as well as the technical folks. I also really enjoyed Yann LeCun’s talk, because I hadn’t previously heard his vision for taking a deep neural-network architecture and extending it to learn full agent capabilities.”
There is one death by suicide in the world every 40 seconds, and suicide is the second leading cause of death for 15-29 year olds. Experts say that one of the best ways to prevent suicide is for those in distress to hear from people who care about them.
Facebook is in a unique position — through friendships on the site — to help connect a person in distress with people who can support them. It’s part of our ongoing effort to help build a safe community on and off Facebook.
Today we’re updating the tools and resources we offer to people who may be thinking of suicide, as well as the support we offer to their concerned friends and family members.
The future of bots is sitting in thousands of documents folders, waiting to be born. At least, that’s the premise of Albert, a bot and bot-creation tool from NoHold, which released a pro version on Monday. The premise behind Albert is straightforward: upload a document, and then ask the Albert-generated bot to answer questions with information based on that document. Albert is a product of the modern era of chatbots, but Albert’s origins are, by tech standards, positively ancient: the key work dates back to a patent filed in 1999.
“Do you know what a bucaneve is?” asks Diego Ventura, the CEO and Founder of NoHold. I admit that I do not. “It’s a flower that magically comes out of the snow. This is what I call our strategy,” Ventura explains. Albert is based on a patent from the last century, and rather than create a tool that makes chatbot then, in the early dotcom boom, Ventura says his company worked instead on a version of the technology used internally by a few big companies, like Dell and Cisco, and that the bot he’s showing me was waiting for the right moment.
The development of robotics and artificial intelligence to improve care of the sick and elderly, and deal with hazardous environments has been boosted with more than £17.3 million ($21.3 million) of investment by the UK Government.
Science funding arm the Engineering and Physical Sciences Research Council (EPSRC) has this week announced two Programme Grants worth a total of £10.8 million for major robotics research projects, involving fiber-optic and photonics-related technologies.
Backed by over £100 million of investment, the RFI will be a national centre of excellence in technology development and innovation. Physical scientists, engineers and life scientists will work together to develop new techniques and instrumentation and apply them to key challenges in the health and life sciences – leading to improved understanding of disease, faster discovery of new treatments for chronic conditions that affect millions of people worldwide, new jobs, and long-term economic growth.
The R Forwards taskforce is also offering diversity scholarships to under-represented individuals who might not otherwise be able to attend. If you qualify and think a scholarship might help you get to useR!2017, the deadline for applications is April 1, at the link below.
Zhou Wang, Alan Conrad Bovik, Hamid Rahim Sheikh and Eero P. Simoncelli, for “Image Quality Assessment: From Error Visibility to Structural Similarity (link is external)”, IEEE Transactions on Image Processing, Volume 13, No. 4, April 2004.
The Minerva Foundation bestowed its Golden Brains award on Eero Simoncelli, a professor of Neural Science, Mathematics, and Psychology at NYU. Dr. Simoncelli was recognized for his “theories of coding efficiency and statistical inference to understand the means by which percepts arise from neural responses.” He was also awarded a Sustained Impact Paper award from IEEE Signal Processing Society. The paper: “Image Quality Assessment: From Error Visibility to Structural Similarity” in the April 2004, IEEE Transactions on Image Processing.
ICON is a comprehensive index of research-quality network data sets from all domains of network science, including social, web, information, biological, ecological, connectome, transportation, and technological networks.
A new collection devoted to neuroscience projects from 2016 Brainhack events has been launched in the open access journal Research Ideas and Outcomes (RIO). At current count, the “Brainhack 2016 Project Reports” collection features eight Project Reports, whose authors are applying open science and collaborative research to advance our understanding of the brain.
For the last two months Code For Science (creators of the Dat Project) have been working with the teams at Data.gov, Data Refuge, the Internet Archive and the California Digital Library to aggregate the government data that has been downloaded so far as part of #datarefuge and create a single metadata dataset. Today we are releasing 38GB of metadata, over 30 million hashes and URLs of research data files.
We are calling this initiative Project Svalbard. The goal is to create a global metadata vault for public research data, especially those at risk of disappearing.
“We are publishing pre-trained word vectors for 90 languages, trained on Wikipedia using fastText. These vectors in dimension 300 were obtained using the skip-gram model described in 1 with default parameters.”