Last Week in AI #175: Cruise Car Outages, AI-generated influencers, AI regulations in Europe, and more!

Top News

Cruise’s Robot Car Outages Are Jamming Up San Francisco

We mentioned this news in last week’s email, but more details emerged this week. On Tuesday night of June 28, nearly 60 Cruise autonomous vehicles stopped moving in the middle of various streets of San Francisco over a 90-minute period after they lost touch. Internal messages seen by WIRED suggest that this was due to the cars losing contact with Cruise servers.

Many cars created a prolonged jam in the city’s downtown, as was reported by the San Francisco Examiner and via photos posted to Reddit. A similar incident happened on the evening of May 18, when Cruise lost touch with its entire fleet and their cars stood still for nearly 20 minutes. Cruise has a system that allows remote operators to drive the vehicles, but was unable to access them in these cases. A letter sent anonymously by a Cruise employee to the California Public Utilities Commission  alleged that the company loses contact with its driverless vehicles “with regularity."

Our take: While autonomous cars may soon drive safely enough to be in wide use on city streets, this incident demonstrates that they may have unexpected bugs that result in strange and potentially especially hazardous behavior. Still, it's surprising how bad this incident is, and that it happened so only two weeks after Cruise was given permission to give paid rides to anyone in SF. Still, no one was injured as a result, unlike a crash that happened to a Cruise vehicle on June 3rd. The National Highway Traffic Safety Agency then said it would open a special investigation into the crash. These incidents unfortunately suggest wide deployment of driverless autonomous vehicles may still not happen for some time.

Fake Friends and the Real Threat of AI-Generated Influencers

Virtual influencers that rely on synthetic images (some generated by AI) instead of real people are gaining popularity, with some having millions of followers. They are typically managed by a marketing company aiming to advertise certain products, like with real influencers. Some hope “that AI influencers could offer a reprieve for real-life influencers who build their lives around the whims of algorithms and advertisers.” Others note that AI influencers are “unpaid, never age the way real-life characters do, and will be free of the kinds of scandals other celebrities can sometimes get mixed up in.” The main worry though is that AI-generated imagery makes it very easy to create realistic pictures that show impossible body standards, which can worsen the mental health problems caused by influencer culture.

Our take: This development is not surprising and will only become more commonplace as generative models, like DALLE-2, become ever more powerful. It’s worth noting that synthetic models have been around for a while, and such virtual idols have been particularly popular in Japan. The difference is that these previous models were obviously animated, but recent advances in AI mean that fans may no longer be able to visually distinguish what’s real from what’s not. So far the AI influencers on Instagram all declare they’re bots in their profile bios. It’s not clear what would happen if there comes a day when such declarations are no longer the norm.

The Fight Over Which Uses of AI Europe Should Outlaw

European countries have already been ahead of the curve in AI regulation—the recently proposed EU AI Act is a notable example of the continent’s efforts. Given its status as a draft, the proposal has evolved over time. Officials from EU nations and members of the European Parliament are considering a ban on AI lie detectors at borders, among other amendments to the Act. These considerations are accompanied by outside pressure: WIRED reports that last month, advocacy groups such as European Digital Rights called for the polygraph ban.

Our take: I’m happy to see continued iteration on and amendments to the AI Act. It’s already been pointed out that the legislation has its flaws, and it can only be made better through revision. Continued input from experts and the public as AI systems are used more in daily life will be vital to curb the worst excesses of these systems. While certain aspects of AI systems are ill-defined and difficult to legislate, even unsuccessful attempts will be important for us to understand how to go about legislation in the future. One notable example from the WIRED article:

An algorithm deployed by the Dutch tax authority to detect potential child benefit fraud between 2013 to 2020 was found to have harmed tens of thousands of people, and led to more than 1,000 children being placed in foster care… The Dutch social benefits scandal might have been prevented or lessened had Dutch authorities produced an impact assessment for the system… the law must have a clear explanation for why a model earns certain labels, for example when rapporteurs moved predictive policing from the high-risk category to a recommended ban.

Such proposals are promising for mitigating potential issues and alleviating concerns over the use of AI systems by making them more easily understood and assessed. But areas like interpretability and impact assessment are open research areas and do not admit easy solutions just yet. I suspect we will see a co-evolution of legislation and better ways of understanding some of these more fundamental issues vital to legislation.

Other News

Research

Meta open sources early-stage AI translation tool that works across 200 languages - "Social media conglomerate Meta has created a single AI model capable of translating across 200 different languages, including many not supported by current commercial tools. The company is open-sourcing the project in the hopes that others will build on its work."

DeepMind’s AI develops popular policy for distributing public money - "A “democratic” AI system has learned how to develop the most popular policy for redistributing public money among people playing an online game."

Using machine learning to study parenting styles - "Parental time invested in the early years has an outsized impact on a child’s future. But measuring the effects of different parenting styles remains an elusive task, in part because self-reported survey questions are susceptible to misreporting bias."

Applications

FIFA will track players’ bodies using AI to make offside calls at 2022 World Cup - "FIFA, the international governing body of association football,* has announced it will use AI-powered cameras to help referees make offside calls at the 2022 World Cup."

Words matter: AI can predict salaries based on the text of online job postings - " the words used in a dataset of more than one million online job postings explain 87% of the variation in salaries across a vast proportion of the labor market."

Japan deploys artificial intelligence to detect rip currents as beach season hots up - "AI system identifies currents and bathers, and sends a warning to lifeguards via a smart watch"

New AI-powered app could boost smartphone batteries by 30 per cent - "It is hoped the EOptomizer app will be adapted across the industry and help drive down carbon emissions, by making consumers' goods last longer."

Original essays written in seconds: how ‘transformers’ will change assessment - "Every few years a technology comes along that its proponents claim will transform education. Now there is another one, handily called a “transformer”. Unlike systems to personalise teaching or deliver content, this is a tool for students rather than educators or administrators."

Japan's forestry industry tests robots to address labor shortage - "The yellow, four-legged robot walks up a grass slope, then marches through a forest full of twigs. It even mounts a stump and then climbs down unassisted. It’s part of a trial run by Forestry and Forest Products Research Institute (FFPRI) and SoftBank Corp."

AI can use your brainwaves to see things that you can't - "Artificial intelligence can use your brainwaves to see around corners. The technique, called “ghost imaging”, can reconstruct the basic details of objects hidden from view by analysing how the brain processes barely visible reflections on a wall."

AI can convert almost any 3D scene into the style of a famous artwork - "Artificial intelligence can transfer artistic styles onto a 3D scene, including turning a bulldozer into the style of Van Gogh’s The Starry Night painting Getting immersed in a 3D-version of your favourite painting could be made possible thanks to an artificial intelligence that can transfer artis"

Using GPT-3 to explain how code works - "One of my favourite uses for the GPT-3 AI language model is generating explanations of how code works. It’s shockingly effective at this: its training set clearly include a vast amount of source code. Here are a few recent examples."

87% of Climate and AI Leaders Believe That AI Is Critical in the Fight Against Climate Change - "New Report from AI for the Planet Alliance, BCG, and BCG GAMMA Reveals a Strong Appetite for Using AI to Tackle Climate Change, but Organizations Face Obstacles to Achieving Impact at Scale "

Fair' AI could help redress bias against Black US homebuyers - "An artificial intelligence could guide reparations programmes created to redress decades of US housing discrimination against Black homebuyers."

Business

Why business is booming for military AI startups  - "Exactly two weeks after Russia invaded Ukraine in February, Alexander Karp, the CEO of data analytics company Palantir, made his pitch to European leaders. With war on their doorstep, Europeans ought to modernize their arsenals with Silicon Valley’s help, he argued in an open letter. "

Israeli tech company awarded for 'best speech to text' AI - "Verbit's AI-driven technology can turn any audio information — Be it lectures, podcasts, or television shows — Into a text transcript in seconds.”

Google, Mayo Clinic build new type of AI algorithm to map interactions between areas of the brain - "Mayo Clinic and Google Research's Brain Team are developing a new type of artificial intelligence algorithm to chart out the neural connections spanning each region of the brain."

Switzerland Moves Ahead With Underground Autonomous Cargo Delivery - "After half a decade of study, Cargo Sous Terrain is ready to start on its first tunnel"

Olive rakes in $400M to turbocharge growth of 'humanized' AI for healthcare - "Healthcare automation startup Olive raked in $400 million in fresh capital to build out its enterprise AI for hospitals. The round, led by Vista Equity Partners, boosts the company's valuation to $4 billion, it claims. Base10 Partners Advancement Initiative also participated in the round."

Ford, VW-backed Argo AI lays off 150 workers, slows hiring - "Argo AI, the autonomous vehicle technology startup backed by Ford and Volkswagen, has laid off about 150 people and slowed the pace of hiring, making it the latest tech company to reduce its workforce as recession fears grow."

Concerns

Today’s Robotic Surgery Turns Surgical Trainees Into Spectators - "Medical training in the robotics age leaves tomorrow's surgeons short on skills"

People who regularly talk to AI chatbots often start to believe they're sentient, says CEO - "In brief Numerous people start to believe they're interacting with something sentient when they talk to AI chatbots, according to the CEO of Replika, an app that allows users to design their own virtual companions."

China Says It's Developing an AI to Detect Party Loyalty - "The latest dystopian nightmare AI just dropped — and it's raising a lot of questions that are absolutely NOT answered by its social media announcement."

How IBM Watson Overpromised and Underdelivered on AI Health Care - "Within the comfortable confines of the dome, Watson never failed to impress: Its memory banks held knowledge of every rare disease, and its processors weren't susceptible to the kind of cognitive bias that can throw off doctors. It could crack a tough case in mere seconds."

US safety regulators open special investigation into Cruise AV crash - "The National Highway Traffic Safety Administration has opened a special investigation into a crash in San Francisco involving a Cruise autonomous vehicle that resulted in minor injuries."

More and more CS students are interested in AI – and there aren't enough lecturers - "Computer-science departments across US universities do not have enough lecturers to teach increasing numbers of students interested in AI, a report from the Center for Security and Emerging Technology (CSET) this month suggested."

Analysis

Neuromorphic Chips Are Destined for Deep Learning—or Obscurity - "People in the tech world talk of a technology “crossing the chasm" by making the leap from early adopters to the mass market. A case study in chasm crossing is now unfolding in neuromorphic computing."

Community colleges can become America’s AI incubators - "Community colleges could create pathways to good-paying jobs across the United States and become tools for training a new generation of AI-literate workers.  "

AI inventors may find it difficult to patent their tech under today's laws - "Comment Future AI could be a challenge for US Patent and Trademark Office (USPTO) officials, who need to wrap their heads around complex technology that's perhaps not quite compatible with today's laws."

These simple changes can make AI research much more energy efficient - "Deep learning is behind machine learning’s most high-profile successes, such as advanced image recognition, the board game champion AlphaGo, and language models like GPT-3. But this incredible performance comes at a cost: training deep-learning models requires huge amounts of energy."

All the talk of computers having ‘brains’ is all wrong. AI advancements don’t work that way - "It is a truth, universally acknowledged, that the machines are taking over. What is less clear is whether the machines know that. Recent claims by a Google engineer that the LaMBDA AI Chatbot might be conscious made international headlines and sent philosophers into a tizz."

Sure, GitHub's AI-assisted Copilot writes code for you, but is it legal or ethical? - "GitHub Copilot, Microsoft's AI pair-programming service, has been out for less than a month now, but it's already wildly popular. In projects where it's enabled, GitHub states nearly 40% of code is now being written by Copilot. That's over a million users and millions of lines of code."

Fun

---

Copyright © 2022 Skynet Today, All rights reserved.