When you arrive at the Game Developers Conference in San Francisco this March you’re going to be confronted with a smorgasbord of fantastic talks, so today organizers want to quickly highlight one you won’t want to miss!
It’s a remarkable entry in the Programming track of talks at GDC: a presentation from Valve’s John McDonald about how the company has been using deep learning methods to try and combat cheating in its competitive team-based shooter Counter-Strike: Global Offensive.
Because the aim of any machine learning algorithm is to identify patterns or regularities in data, it is only natural that these algorithms will identify bias. Removing the algorithmic black-box will not eliminate the bias. You may be able to get a better idea of what the algorithm is doing but it will still enforce the biased patterns it `sees’ in the data. Because, what we really don’t want is the machine to act on bias, i.e. to be prejudiced, not that the machine does not use heuristics.
Transparency is then better served by proper treatment of the learning process than by removing the black box. Trust in the system will improve if we can ensure openness of affairs in all that is related to the system. The following principles to the design of AI systems should be required from all models that use human data or affect human beings or can have other morally significant impact
Klaus Schwab, Founder and Executive Chairman of the World Economic Forum, talks with Google Chief Executive Sundar Pichai on the age of artificial intelligence, training the new workforce,. tax regulations and more.
There’s an intriguing new player on the cybersecurity block, and it’s called Chronicle. Notable because it’s part of Google’s parent company, it emerged out of Alphabet’s “moonshot” incubator, known as X. Announced last week in two different blog posts on Medium, Chronicle will focus on helping companies comprehend their own security data and, according to the company’s CEO, “stop cyber attacks before they cause harm.”
In an era of global computer infections like WannaCry, or vulnerabilities in computer processors like Meltdown and Spectre, a Google-like company turning its focus and resources to cybersecurity is a good thing. But there’s limited information available about how it might function.
In a blog post, the company’s cofounder and CEO, Stephen Gillett, wrote that one prong of the firm will be “a new cybersecurity intelligence and analytics platform that we hope can help enterprises better manage and understand their own security-related data.”
Howard University today announced that Howard West, the university’s academic partnership with Google will expand to cover the full academic year beginning fall 2018. The announcement comes after a successful three-month pilot program during the Summer of 2017 at Google’s Silicon Valley headquarters.
“Howard West is an extension of our commitment to produce industry-ready Black computer science graduates who will enter the workforce with the added invaluable knowledge gained by working alongside the leading experts at Google,” says Howard University President Dr. Wayne A. I. Frederick. “The program has greatly enhanced the learning process for our students and faculty while producing technology professionals equipped with first-hand knowledge and practical applications from the industry.”
Computers can beat chess champions, simulate star explosions, and forecast global climate. We are even teaching them to be infallible problem-solvers and fast learners.
And now, physicists at the Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) and their collaborators have demonstrated that computers are ready to tackle the universe’s greatest mysteries. The team fed thousands of images from simulated high-energy particle collisions to train computer networks to identify important features.
The researchers programmed powerful arrays known as neural networks to serve as a sort of hivelike digital brain in analyzing and interpreting the images of the simulated particle debris left over from the collisions. During this test run the researchers found that the neural networks had up to a 95 percent success rate in recognizing important features in a sampling of about 18,000 images.
Hoping to accelerate developments in artificial intelligence by attracting top research talent, DeepMind has announced funding for an endowed chair in the University of Alberta’s Department of Computing Science.
National Bureau of Economic Research, James Bessen
from
Artificial intelligence (AI) technologies will automate many jobs, but the effect on employment is not obvious. In manufacturing, technology has sharply reduced jobs in recent decades. But before that, for over a century, employment grew, even in industries experiencing rapid technological change. What changed? Demand was highly elastic at first and then became inelastic. The effect of artificial intelligence on jobs will similarly depend critically on the nature of demand. This paper presents a simple model of demand that accurately predicts the rise and fall of employment in the textile, steel and automotive industries. This model provides a useful framework for exploring how AI is likely to affect jobs over the next 10 or 20 years.
Imagine that spring has finally arrived and you’re planning your weekend. The weather forecast looks great. You could go to the beach – but what if it’s closed because of an algal bloom? Maybe you could go for a hike – will the leaves be out yet? What might be in flower? Will the migratory birds be back? Oh, and you heard last year was bad for ticks – will this spring be better or worse?
We all take weather forecasts for granted, so why isn’t there a ‘nature forecast’ to answer these questions? Enter the new scientific field of ecological forecasting. Ecologists have long sought to understand the natural world, but only recently have they begun to think systematically about forecasting.
Particle physics and machine learning have long been intertwined.
One of the earliest examples of this relationship dates back to the 1960s, when physicists were using bubble chambers to search for particles invisible to the naked eye. These vessels were filled with a clear liquid that was heated to just below its boiling point so that even the slightest boost in energy—for example, from a charged particle crashing into it—would cause it to bubble, an event that would trigger a camera to take a photograph.
Our ability to predict species responses to environmental changes relies on accurate records of animal movement patterns. Continental-scale acoustic telemetry networks are increasingly being established worldwide, producing large volumes of information-rich geospatial data. During the last decade, the Integrated Marine Observing System’s Animal Tracking Facility (IMOS ATF) established a permanent array of acoustic receivers around Australia. Simultaneously, IMOS developed a centralised national database to foster collaborative research across the user community and quantify individual behaviour across a broad range of taxa. Here we present the database and quality control procedures developed to collate 49.6 million valid detections from 1891 receiving stations. This dataset consists of detections for 3,777 tags deployed on 117 marine species, with distances travelled ranging from a few to thousands of kilometres. Connectivity between regions was only made possible by the joint contribution of IMOS infrastructure and researcher-funded receivers. This dataset constitutes a valuable resource facilitating meta-analysis of animal movement, distributions, and habitat use, and is important for relating species distribution shifts with environmental covariates.
arXiv, Computer Science > Learning; Dmitriy Serdyuk, Nan Rosemary Ke, Alessandro Sordoni, Adam Trischler, Chris Pal, Yoshua Bengio
from
We propose a simple technique for encouraging generative RNNs to plan ahead. We train a “backward” recurrent network to generate a given sequence in reverse order, and we encourage states of the forward model to predict cotemporal states of the backward model. The backward network is used only during training, and plays no role during sampling or inference. We hypothesize that our approach eases modeling of long-term dependencies by implicitly forcing the forward states to hold information about the longer-term future (as contained in the backward states). We show empirically that our approach achieves 9% relative improvement for a speech recognition task, and achieves significant improvement on a COCO caption generation task.
“We are super pumped to be launching a weekly data science podcast called DataFramed, in which Hugo Bowne-Anderson (me), a data scientist and educator at DataCamp, speaks with industry experts about what data science is, what it’s capable of, what it looks like in practice and the direction it is heading over the next decade and into the future.”