“We decided we wanted to do something that was completely historically unprecedented,” Li said, referring to a small team who would initially work with her. “We’re going to map out the entire world of objects.”
The resulting dataset was called ImageNet. Originally published in 2009 as a research poster stuck in the corner of a Miami Beach conference center, the dataset quickly evolved into an annual competition to see which algorithms could identify objects in the dataset’s images with the lowest error rate. Many see it as the catalyst for the AI boom the world is experiencing today.
LinkedIn is now offering a new free reporting tool to companies. Website demographics let marketers see what types of professionals are clicking through to their websites. It’s “a powerful way to tune your marketing to those visitors, and develop better targeting and content for your campaigns,” according to the LinkedIn announcement.
They’ll be able to see members’ job title, industry, company, location, and more so they can target their efforts accordingly.
Facebook Code, Engineering Blog; Piotr Bojanowski, David Lopez-Paz, Hervé Jegou, Antoine Bordes
from
At Facebook AI Research (FAIR), we wanted to investigate further how artists and professionals typically trained to work with CGI special effects can use these types of tools. Specifically, we set out to answer the following question: Can AI-generated imagery be relevant in the context of directing and producing an immersive film in VR?
To that end, we partnered with the production company OKIO Studio, the Saint George VFX studio, and director Jérôme Blanquet on the virtual reality experience “Alteration.” The goal of our collaboration was to study whether we could generate special effects with AI techniques that were good enough to be part of the final film.
Eugenia Kuyda’s best friend died in 2015. Using a chatbot structure she developed, she entered their messaging history into a Google-built neural network, creating a bot she could interact with. It was the earliest version of Replika, a bot that, as you interact with it, turns into a digital representation of you.
F1000Research, Saulo de Oliveira and Charlotte Deane
from
A large number of structural bioinformatics applications rely on extracting structural features from a protein’s sequence. This is traditionally done by performing multiple sequence alignments (MSAs) of homologues. MSAs have been used as input to predict features such as secondary structure, torsion and bond angles, solvent accessibility, disorder regions, and domain boundaries. The main limitation of most of these descriptors such as predicted secondary structure is that, although often highly accurate, they provide information only about a protein’s local conformation. For instance, they may tell us how a set of residues comprise an alpha-helix, but they do not provide any information as to how different alpha-helices are oriented with respect to one another. Techniques based on co-evolution go a step further by extracting non-local structural information from MSAs. These techniques are based on the notion that two residues which mutate in a correlated fashion, so that a mutation in one is often compensated by a mutation in the other, can be considered to be co-evolving. Co-evolution is interpreted as functional dependence, i.e. if two residues are co-evolving, there is a cost in fitness for mutating only one of these residues. Although these techniques were originally conceived and applied to protein structure prediction, they are now establishe
The City of Seattle has finished searching for its first smart city coordinator, choosing former Kansas City, Missouri, Innovation Policy Advisor Kate Garman.
The city announced the hire on Thursday after a months-long search for a leader who could bring diligence and community-focus to an emerging technology space fraught with both potential and pitfalls. Garman’s duties as the city’s first smart city coordinator will involve leading a smart city program dedicated to establishing “policies, partnerships, systems, platforms, and networks” that serve the needs of Seattle’s residents through the innovative use of technology, according to a press release.
A new camera that builds on technology first described by Stanford researchers more than 20 years ago could generate the kind of information-rich images that robots need to navigate the world. This camera, which generates a four dimensional image, can also capture nearly 140 degrees of information.
“We want to consider what would be the right camera for a robot that drives or delivers packages by air. We’re great at making cameras for humans but do robots need to see the way humans do? Probably not,” said Donald Dansereau, a postdoctoral fellow in electrical engineering.
With robotics in mind, Dansereau and Gordon Wetzstein, assistant professor of electrical engineering, along with colleagues from the University of California, San Diego have created the first-ever single-lens, wide field of view, light field camera, which they are presenting at the computer vision conference CVPR 2017 on July 23.
Translational Data Analytics Institute builds master’s degree program from scratch and seeks input from would-be employers
Extra Extra
Mike Monteiro, an extremely talented designer and communicator, posted a Designer’s Code of Ethics. Go for the advice: “The work you bring into the world is your legacy. It will outlive you. And it will speak for you.” Stay for the unflinching reflexivity: “No one wakes up one day designing to throw their ethics out the window….Have you veered off course? Correct it. Is your workplace an unethical hellmouth? Get another one. Your job is a choice. Please do it right.”
Ear candy for the week, though it is not at all sweet, is this NPR story on hospitals struggling to keep ahead of hackers who could take heart monitors and other medical devices offline.
Michael Harris offers a reasonable overview of where AI and ML meet finance. You know he’s reasonable when he writes, “The majority of funds use fundamental analysis because this is what managers learn in their MBA programs. There are not many hedge funds that rely solely on AI.” Aside from high-frequency trading and a few hedge funds, I’d rank finance well behind precision medicine and AgTech as the industries most likely to benefit from AI in 2017. This is good.
The big tech firms, along with their highly skilled, highly paid workers, have made Pittsburgh younger and more international and helped to transform once-derelict neighborhoods like Lawrenceville and East Liberty.
Indeed, East Liberty has become something of a tech hub, said Luis von Ahn, the co-founder and chief executive of Duolingo, a language-learning platform company with its headquarters in that neighborhood. Google Pittsburgh, with its more than 500 employees, also has part of its offices in East Liberty, as does AlphaLab, a start-up accelerator.
Within easy walking distance from them is the Ace, a branch of the hip hotel chain that opened in 2015 in a former Y.M.C.A. building. The hotel’s in-house Whitfield restaurant and lobby bar have become hangouts for local techies and out-of-towners alike.
On May 16, three months’ worth of data vanished quietly from the website of the U.S. Census Bureau. The figures included age, sex and employment data, crucial for calculating key statistics like the monthly unemployment rate. Their disappearance set off alarm bells among researchers: President Donald Trump had repeatedly cast doubt on the accuracy of the unemployment rate during his campaign. Was some official now actually interfering with the basic measurements of the economy?
That wasn’t the case, it turned out. For all the worry, the data had disappeared because of a technical issue with the online portal, according to a Census spokeswoman. The files were back online before the end of the day.
The scare about the Census data was just one episode in a long and anxious drama that has played out in 2017 about the fate of government data under Trump.
The embedded market for neural network accelerators is heating up, with more systems — ranging from smart speakers and drones to light bulbs — poised to run neural networks locally instead of going back to the cloud for computation.
In an interview with EE Times, Remi El-Ouazzane, vice president and general manager of Movidius, defined the growing trend as “a race for making things smart and autonomous.
Cambridge, MALightbox Gallery asks where artificial intelligence is headed—and how art can help us explore the the world we are making. August 8-13 at Harvard Art Museums (32 Quincy St.) [events are free with museum admission, $$]
Chicago, IL At the University of Chicago, professional science and technology journalists get a respite from deadlines and insight into pressing subjects, such as climate and energy science, the microbiome, advanced batteries, precision medicine and more. Program Dates: Nov. 5-8. Application Deadline: Aug 15.
Imagine if there was a way to identify an unknown substance in real-time, by simply pointing a scanner at the surface in which the substance is located. Chemicals, or areas containing potential chemical residues, could be scanned at standoff range and the substance identified on the spot. Such a development would enable expedited results in time sensitive situations and “one hundred percent” screening for security or process control. The Modeling of Reflectance Given Only Transmission of High-concentration Spectra for Chemical Recognition Over Widely-varying Environments (MORGOTH’S CROWN) Challenge seeks development of algorithms that would help accomplish just that! … Teams will need to register by August 2017, while independent solvers can register through the end of the Challenge.
“We introduce Adversarial Variational Optimization (AVO), a likelihood-free inference algorithm for fitting a non-differentiable generative model incorporating ideas from empirical Bayes and variational inference. We adapt the training procedure of generative adversarial networks by replacing the differentiable generative network with a domain-specific simulator. We solve the resulting non-differentiable minimax problem by minimizing variational upper bounds of the two adversarial objectives. Effectively, the procedure results in learning a proposal distribution over simulator parameters, such that the corresponding marginal distribution of the generated data matches the observations.”
“This blogpost will give an introduction to the architecture and ideas behind TensorForce, a new reinforcement learning API built on top of TensorFlow.”
“This post is about a practical question: How can the applied reinforcement learning community move from collections of scripts and individual examples closer to an API for reinforcement learning (RL) — a ‘tf-learn’ or ‘skikit-learn’ for RL?”