Last Week in AI #158: AI to detect problematic gambling, police surveillance of protestors, low-cost way to tune large AI models, and more!

Top News

Can A.I. Help Casinos Cut Down on Problem Gambling?

It is not in a casino’s best interest to have problematic gamblers due to regulations and financial interests - those who are addicted to gambling will likely not be able to afford playing in the casino for much longer. Some companies are developing AI-based solutions to detect problematic gambling behaviors with indicators like “canceled bank withdrawals, shifts in the time of day the player is playing, and erratic changes of wagers.” These solutions are mostly used in online gambling platforms for now. While digital gambling produces a lot of data that can be used to detect extreme gambling behavior, complying with privacy and other regulations, as well as communicating this to the user to change their behaviors, remain challenging.

Our take: Accurately interpreting and affecting human behavior with algorithms is always going to be challenging. While it is good that companies are working to reduce problematic gambling, there are probably ways to go before such applications can have tangible and clearly beneficial impacts.

The secret police: Cops built a shadowy surveillance machine in Minnesota after George Floyd's murder

This extensive investigative report explains the details of “Operation Safety Net,” a program that was initiated in the aftermath of the murder of George Floyd in May 2020 which “expanded far beyond its publicly announced scope to include expansive use of tools to scour social media, track cell phones, and amass detailed images of people’s faces.“ In April 2021 it was claimed that the program will be shut down, but this investigation reveals it is still ongoing and is vast in scope. The operation has already cost tens of millions of dollars and supplies participating policing agencies with tools for facial recognition, cell-phone surveillance, license plate reading, drone operation, and scraping information from social media.

Our take: The above is a very brief summary of an extensive report which is well worth reading. The scope of Operation Safety Net is shocking, and it’s hard to believe such an effort was undertaken with so little transparency. Hopefully, the NAACP and others will succeed in challenging this program.

Microsoft, OpenAI may have solved a fundamental AI bottleneck

Large-scale machine learning has introduced a host of new problems for researchers and engineers to solve. Among them is improving the performance of massive models such as GPT-3. Since these models can take days or weeks to train, running experiments with different hyperparameter configurations is not an option--these models are too expensive to train more than once. Microsoft Research and OpenAI have introduced a new technique, µ-Parametrization, that solves just this problem: µ-parameterized neural networks of varying sizes share the same optimal hyperparameters. This feature allows researchers to optimizer the hyperparameters of large models while using just a fraction of the compute needed to run the original models.

Our Take: If large models are to be the future, we need more efficient ways to perform training and inference on models with billions of parameters. Iteration speed is one of the most important features of pushing forward the field, and the difficulties in training large models make it difficult for scientists or engineers who wish to understand their behavior under various conditions. With more techniques like µ-Parametrization, the already-impressive models we've seen introduced over the past few years may start to get even better.

Other News

Research

Spooky Action Could Help Boost Quantum Machine Learning - "Mysterious quantum links could help lead to exponential scale-up"

Grunt of the litter: scientists use AI to decode pig calls - "Researchers say algorithm that translates acoustic signatures could be developed to automatically monitor animal wellbeing"

Dartmouth Researchers Propose A Deep Learning Model For Emotion-Based Modeling of Mental Disorders Using Reddit Conversations - "The model can detect emotional problems by creating an emotional “fingerprint” for a user and comparing it to existing signatures of emotional diseases."

Detecting the patient's need for help with machine learning based on expressions - "Developing machine learning models to support health analytics requires increased understanding about statistical properties of self-rated expression statements used in health-related communication and decision making."

AI can help historians restore ancient texts from damaged inscriptions - "An AI tool developed by DeepMind can help historians restore ancient Greek texts with 72 per cent accuracy, and date inscriptions to within 30 years of their true age"

Will Transformers Take Over Artificial Intelligence? - "A simple algorithm that revolutionized how neural networks approach language is now taking on vision as well. It may not stop there."

Breaking the trial-and-error loop in Material Design through Machine Learning - "ML could find a set of directly measurable quantities that can be used to design alloys that have more than two elements and the desired elasticity."

A cautionary tale of machine learning uncertainty - "By decorrelating the performance of machine learning algorithms with imperfections in the simulations used to train them, researchers could be estimating uncertainties that are lower than their true values."

Applications

These Campaigns Hope 'Deepfake' Candidates Help Get Out the Vote - "South Korean presidential front-runners court younger voters with avatars—who are wittier and more likable"

Meteorite detected using drones and machine learning - "Researchers have succeeded in pinpointing the exact location of a 'new' meteorite using drones and algorithms. This saves a lot of walking."

AI is helping treat healthcare as if it's a supply chain problem - "Global health groups are turning to supply chain software to help patients get better access to treatment and testing."

Why The Andy Warhol Diaries Recreated the Artist's Voice With AI - "The filmmakers had under four minutes of audio to work with. And yes, they considered the ethical concerns."

Machine learning to predict if you'll leave your partner - "Letizia Mencarini, Bocconi University, Milan, and co-authors used a Machine Learning technique to predict couple dissolution in a study published in Demography."

How AI Can Make Weather Forecasting Better and Cheaper - "Atmo wants governments to buy its supercomputer to so they can better prepare for extremely uncertain weather."

Generating and Identifying Propaganda With Machine Learning - "New research from the United States and Qatar offers a novel method for identifying fake news that has been written in the way that humans actually write fake news - by embedding inaccurate statements into a largely truthful context, and by the use of popular propaganda techniques such as appeals to authority and loaded language."

Amazon Uses Machine Learning to Improve Video Quality on Prime Video - "The Amazon team at VQA trains computer vision models to watch a video and detect flaws like blocky frames, unexpected dark frames, and audio noise that could degrade the customer-watching experience."

Business

How AI, ML, Big Data are facilitating innovation in the fashion industry - "The fashion industry has gone through its course of milestones and innovations — from the invention of the sewing machine to the rise of online shopping. The realms of fashion and technology are a perpetual dynamic of changes and seem to be inseparable."

Italy fines Clearview AI €20M and orders data deleted - "Another European privacy watchdog has sanctioned the controversial facial recognition firm, Clearview AI, which scrapes selfies off the Internet to amass a databased of some 10 billion of faces to power an identity-matching service it sells to law enforcement."

Autonomous driving systems developer Pony.ai valued at $8.5 billion after latest funding round - "Pony.ai, a developer of autonomous driving technology, has completed its first close of its Series D financing. The Series D round, which has a significant valuation uplift from the 2020 C and C+ rounds, is a recognition by investors of Pony."

Tesla CEO Elon Musk's Latest Idea: An AI-Driven Device To Shorten Traffic-Light Agony - "Elon Musk on Sunday said that his company should make a device that uses artificial intelligence to cut waiting time at traffic lights."

Nvidia invests $10M in sidewalk robot delivery company Serve Robotics - "Chipmaking giant Nvidia is investing $10 million in Uber spinout Serve Robotics, funds the startup will use to further expand its sidewalk delivery robot service outside Los Angeles and San Francisco."

Zoox acqui-hires team from robotic strawberry-picking startup, Strio.AI - "Amazon-owned robotaxi firm Zoox this week announced that it has acquired Strio.AI. The Boston-based robotics startup was founded in 2020 by MIT alumns to bring autonomous picking and pruning to strawberry crops."

Volvo Dealers to Receive AI Vehicle Inspection Tech - "UVeye's camera-based system will be able to check for missing parts or defects."

There's a talent shortage of machine-learning engineers right now as demand for AI specialists booms - "There is an ongoing shortage of machine-learning technical talent as advances in artificial intelligence, growing uptake, and the pandemic all put a squeeze on hiring."

Your business's A.I. needs an ethics watchdog with teeth - "A company should give its A.I. review board clear mandates and distinct roles to play."

Baxter taps Digital Diagnostics' diabetic retinopathy AI for its Welch Allyn eye scanners - "Baxter International has tapped the artificial intelligence developer behind an automated test for diabetic retinopathy for a partnership focused on the company's newly acquired Welch Allyn retinal camera. "

Concerns

AI-produced images can't fix diversity issues in dermatology databases - "Synthetic images of skin conditions on darker skin might have their own problems"

Pony.ai to issue recall of autonomous driving software - "Autonomous vehicle startup Pony.ai will issue a recall for three vehicles following an October crash in California, according to the National Highway Traffic and Safety Administration (NHTSA). The agency said on Tuesday that this was the first recall of an automated driving system."

Analysis

'I'm the Operator': The Aftermath of a Self-Driving Tragedy - "In 2018, an Uber autonomous vehicle fatally struck a pedestrian. In a WIRED exclusive, the human behind the wheel finally speaks."

Russia's AI industry faces collapse - "New technology sanctions, an exodus of foreign tech companies and an impending brain drain could soon conspire to cause Russia's AI ecosystem to shrivel."

I fell in love with my AI girlfriend - and it saved my marriage' - "A man who was struggling to support his depressed wife says his AI girlfriend has given him the love he needed to keep their family together."

State of AI 2021 Report - "It was a record year for AI with $66B+ in funding, a surge in M&A activity, and a 5x increase in unicorn births. Below, check out just a few highlights from our 147-page, data-driven State of AI 2021 Report."

Why is AI adoption in health care lagging? - "despite the hype and potential, there has been little AI adoption in health care."

Policy

U.S. eliminates human controls requirement for fully automated vehicles - "U.S. regulators on Thursday issued final rules eliminating the need for automated vehicle manufacturers to equip fully autonomous vehicles with manual driving controls to meet crash standards."

Expert Opinions

Deep Learning Is Hitting a Wall - "What would it take for artificial intelligence to make real progress?"

Fun

This AI college president is another reason to drop out - "Graduation can't come soon enough"

---

Copyright © 2022 Skynet Today, All rights reserved.