Group items tagged

"DeepMind is trying to combine deep learning and algorithms, creating the one algorithm to rule them all: a deep learning model that can learn how to emulate any algorithm, generating an algorithm-equivalent model that can work with real-world data."

""This study shows that if this algorithm were in place earlier, some materials could have conceivably been discovered years in advance," Jain said."

""This is the first time such machine learning tools have been used in this context," says Fluri, "and we found that the deep artificial neural network enables us to extract more information from the data than previous approaches. We believe that this usage of machine learning in cosmology will have many future applications.""

"It is also self-improving. The 10-year-old grading software leverages deep learning algorithms to "compare notes" with human teachers' scores, suggestions, and comments. An engineer involved in the project compared its capabilities to those of AlphaGo, the record-breaking AI Go player developed by Google subsidiary DeepMind."

"Learning from rewards seems like the simplest thing. I make coffee, I sip coffee, I'm happy. My brain registers "brewing coffee" as an action that leads to a reward. That's the guiding insight behind deep reinforcement learning, a family of algorithms that famously smashed most of Atari's gaming catalog and triumphed over humans in strategy games like Go. Here, an AI "agent" explores the game, trying out different actions and registering ones that let it win."

"Train the Deep Learning Ahem Detector with two sets of audio files, "a negative sample with clean voice/sound" (minimum 3 minutes) and "a positive one with 'ahem' sounds concatenated" (minimum 10s) and it will detect "ahems" in any voice sample thereafter."

"What Hamilton and his colleagues found surprising was that this approach could also be applied to helping find problems with existing deep networks, related to the surge of "deepfakes" that have recently cropped up. They applied this data structure to find areas where probabilistic models, such as the generative adversarial networks (GANs) that are often used to create deepfakes, break down. They coined these problematic areas "blind spots," and note that they give us insight into how GANs can be biased. Such blind spots further show that GANs struggle to represent particular areas of a dataset, even if most of their fakes can fool a human. "

"So they trained a deep-learning neural net on tons of examples of deepfaked videos, and produced a model that's better than any previous automated technique at spotting hoaxery. (Their paper documenting the work is here.)"

"Lexplore claims its technology is new-particularly the algorithm that separates typical from atypical readers. But the concepts it's based on aren't. Its tech draws from a deep well of previously-conducted research stretching back decades, which is generally supportive of using a combination of eye tracking and machine learning to screen for dyslexia. "Eye movements is one of the best ways to index reading ability at an incredibly in-depth level," says Julie Kirkby, a psychology professor at Bournemouth University in the United Kingdom, who has studied eye tracking and dyslexia for years."

"One great way to tell the difference is to ask AI recruiting companies what they use artificial intelligence, machine learning and/or deep learning for. Hopefully the hiring firm can what it's using the new technology for and not just that it is. If not it's time to dig a bit deeper."

"Rather than being "taught" medical knowledge, the AI was asked to learn using unsupervised deep neural networks, known as autoencoders, without being given any medical knowledge. The researchers developed a method for translating the features found by the AI-only numbers initially-into high-resolution images that can be understood by humans."

"Hassabis was a child chess prodigy, who learned the game aged four and was able to beat his dad three weeks later - indeed, when he started playing competitively he was so small he had to bring a pillow with him to reach the board - and became a strong player. Yet in AlphaZero's case there was no human input, other than telling it the rules of each game. "In a matter of a few hours it was superhuman," Hassabis says proudly."

"The performance of LipNet compares incredibly favorably to human lipreading experts on GRID corpus, the largest publicly-available sentence-level lipreading dataset. In fact, where human experts got just 52 percent, LipNet scored 93 percent. Its sentence-based approach to lip-reading also smashed the best previous attempt by a machine, which managed 79.6 percent accuracy on the same dataset."