Mini Briefs

China makes it a criminal offense to publish deepfakes or fake news without disclosure

China has released a new government policy designed to prevent the spread of fake news and misleading videos created using artificial intelligence, otherwise known as deepfakes. Failure to disclose that deepfakes or false information posted online were created by AI is now considered a criminal offense. The rules, to be enforced by the Cyberspace Administration of China, will go into effect on January 1, 2020.

This rule follows in the wake of California’s recent action last month, in which it became the first US state to criminalize the use of deepfakes in political campaign promotion and advertising. Given that Congress and US platforms are in the process of analyzing the potential harm of deepfakes and creating tools to detect them, this rule by China represents a broader move forward in the fight against technology-aided misinformation.

Why the fight against disinformation, sham accounts and trolls won’t be any easier in 2020

The big tech companies have announced aggressive steps to keep trolls, bots and online fakery from marring another presidential election - from Facebook’s removal of billions of fake accounts to Twitter’s spurning of all political ads. Unfortunately, as we approach the 2020 election, disinformation techniques are only growing more subversive and sophisticated, leaving tech companies with difficult and subjective choices on how to combat them.

Politico considers a number of the evolving challenges Silicon Valley faces as it tries to counter such misinformation heading into the election cycle. These include the danger of American trolls, the trickiness of policing domestic content, the fact that bad actors are learning and improving, and the difficulty of labeling information as misleading with certainty.

Advances & Business

• AI-enabled assistant robot returning to the Space Station with improved emotional intelligence - The Crew Interactive Mobile Companion (or CIMON for short) recorded a number of firsts on its initial mission to the International Space Station, which took place last November, including becoming the first ever autonomous free-floating robot to operate aboard the station, and the first ever smart astronaut assistant.

• 2020 AI Predictions - PwC’s third annual AI Predictions report shows that only 4% of executives surveyed plan to deploy AI enterprise-wide in 2020. A year ago, nearly 20% hoped to do the same.

• AWS announces DeepComposer, a machine-learning keyboard for developers - On the first day of AWS re:Invent, Amazon announced DeepComposer, a machine-learning driven keyboard aimed at developers. The keyboard is supposed to help developers learn about machine learning in a fun way, and maybe create some music along the way.

• Helping machines perceive some laws of physics - MIT researchers have designed a model that demonstrates an understanding of some basic “intuitive physics” about how objects should behave. The model could be used to help build smarter artificial intelligence and, in turn, provide information to help scientists understand infant cognition.

• Demis Hassabis on AI’s potential - Demis Hassabis argues that artificial intelligence (AI) could usher in a new renaissance of discovery, acting as a multiplier for human ingenuity, opening up entirely new areas of inquiry and spurring humanity to realise its full potential.

• Japanese Researchers Teaching Robots to Repair Themselves - At Humanoids 2019 in Toronto, researchers from the University of Tokyo showed how they taught a PR2 to perform simple repairs on itself by tightening its own screws.

• Neurosymbolic AI Advances State of the Art on Math Word Problems - Researchers at Microsoft have demonstrated a new technique called Neurosymbolic AI which has shown promising results when applied to difficult scenarios such as algebra problems stated in words. The PureAI editors were given a sneak peek at the draft of a research paper that describes the work.

• Deep-Learning the Hardest Go Problem in the World - Earlier this year, David Wu, researcher at Jane Street, posted about our project KataGo and research to improve self-play learning in Go, with an initial one-week run showing highly promising results. Several months later in June, KataGo performed a second, longer 19-day run with some major bugfixes and minor optimizations.

• A radical new neural network design could overcome big challenges in AI - Researchers borrowed equations from calculus to redesign the core machinery of deep learning so it can model continuous processes like changes in health.

• A silicon chip that mimics the brain’s neurons could help fight paralysis - Artificial neurons that mimic the way our body’s nerve cells transfer electrical signals could one day help patients with nerve damage.

• Using Artificial Intelligence To Analyze Markets: An Interview With Ainstein CEO Suzanne Cook - To learn more about the use of artificial intelligence at it may be applied to analyzing stocks and markets, Forbes asked the CEO and originator of Ainstein AI about her work in this area.

• This AI writes a text adventure while you play it - It’s easy to imagine advances in AI will have an impact on strategy games and digital versions of board games like Chess and Go. This article considers an implementation of AI technology in a text adventure.

• This Year’s AI (Artificial Intelligence) Breakthroughs - When it comes to AI (Artificial Intelligence), VCs (venture capitalists) continue to be aggressive with their fundings. During the third quarter, 965 AI-related companies in the US raised a total of $13.5 billion. In this article, Forbes details this year’s breakthroughs in AI.

• Global AI Survey: AI proves its worth, but few scale impact - Most companies report measurable benefits from AI where it has been deployed; however, much work remains to scale impact, manage risks, and retrain the workforce. A group of high performers shows the way.

Concerns & Hype

• What is the law when AI makes the “decisions”? - It should be no surprise that the Federal Court last week ruled that the Australian government’s scandal-plagued automated welfare-debt collection system was illegal. The real surprise is that the system, colloquially dubbed “Robodebt”, was ever allowed to happen in the form it did.

• What should newsrooms do about deepfakes? These three things, for starters - Three researchers argue the dangers of deepfakes are overblown, but they will still require journalists to give thought to how they handle unconfirmed information.

• Artificial intelligence: How to measure the “I” in AI - While AI has had significant success in areas like game playing, contemporary AI also fails miserably at basic tasks that almost every human can perform. In this article, TechTalks summarizes key recommendations on measuring AI systems from AI researcher Francois Chollet.

• Dismantling AI Myths and Hype - This article considers the various predictions and overhyping surrounding AI.

• What Microsoft and Google Are Not Telling You About Their A.I. - While lauded A.I. technologies like Google’s search algorithm and AlphaGo are not fake, they may be less “artificial” than Silicon Valley’s PR machine wants us to believe.

• Biased Algorithms Are Easier to Fix Than Biased People - Racial discrimination by algorithms or by people is harmful - but that’s where the similarities end.

• Facebook’s Head of AI Says the Field Will Soon “Hit the Wall” - In this interview, Wired talks with Facebook’s Head of AI.

• Robots in Finance Could Wipe Out Some of Its Highest-Paying Jobs - Robots have replaced thousands of routine jobs on Wall Street. Now, they’re coming for higher-ups.

• Magnolia residents’ AI-powered surveillance camera tracks people, cars at entrance to neighborhood, experts caution bias - An AI security camera in Land’s End is part of a larger trend of increased surveillance in a time when technology has become cheaper and more effective.

Analysis & Policy

• AI takes on popular Minecraft game in machine-learning contest - In this contest, competitors may take up to four days and use no more than eight million steps to train their AIs to find a diamond.

• Exclusive: government blocks full publication of AI review - The government is blocking the full publication of an official report assessing the ways artificial intelligence could be deployed across Whitehall and the wider public sector, NS Tech can reveal.

• Yes, China is probably outspending the US in AI - but not on defense - New estimates show how much China’s investments in AI have been overblown.

Expert Opinions & Discussion within the field

• The Doer and The Clarion Caller - There is an unnecessary drama unfolding on Twitter on “the war” between connectionists and symbolists. This drama is absurd and perpetuated by people whose relevance exists only in the discussion of the “difference” and the “war.”

Explainers

• Increase model performance by… removing data? - In any given dataset, not every sample will contribute equally to training a machine learning model. This is obvious when you say it out loud. Some data will teach a model a ton. But some will be irrelevant, or redundant, and won’t really move the needle at all.

• This is how Facebook’s AI looks for bad stuff - The context: The vast majority of Facebook’s moderation is now done automatically by the company’s machine-learning systems, reducing the amount of harrowing content its moderators have to review.

• arXiv Machine Learning Classification Guide - We are excited to see the adoption of arXiv in the rapidly growing field of machine learning. Given the interdisciplinary nature of machine learning, it is becoming a challenge for our volunteer moderators to keep up with verifying the appropriate categories for machine learning applications.

That’s all for this week! If you are not subscribed and liked this, feel free to subscribe below!