

Discover more from Last Week in AI
Last Week in AI #144: AI helps to discover new planets, Poetry by creepy robo-artist, China backs banning its social scoring
AI helps to identify 301 new exoplanets, Humanoid robo-artist gives public performance of its own poetry, China backs UN's new AI guidelines that ban AI-driven social scoring, and more!
Welcome to the our latest round up of last week’s AI news!
We also offer weekly editorials as a perk for paying subscribers, and we hope you will consider subscribing to our substack to get access to future ones and to support us — and you can do so with a hefty black friday discount!
Top News
A Whopping 301 Newly Confirmed Exoplanets – Discovered With New Deep Neural Network ExoMiner
NASA recently confirmed 301 exoplanets (planets outside the solar system), adding to a total of about 4.5k known to mankind. Detecting exoplanets is difficult, because they do not emit light and are too far to be directly seen by telescopes. Instead, researchers need to observe when and how a star's brightness dims, which may indicate a planet passing in front of it. There are large datasets of star brightness data, but manually parsing and analyzing them takes a lot of time. To tackle this problem, researchers trained a neural network called ExoMiner to classify potential exoplanet candidates, greatly speeding up the discovery process. The model is trained on a labeled dataset of observations from the Kepler Space Telescope, and researchers hope to fine-tune the model to observations from other telescopes soon.
Our take: It is encouraging and exciting to see AI advances making an impact in science, where analyzing and processing large bulks of data by humans is often impractical. Another famous example that comes to mind is CERN, which is known for using AI to filter what data to save from its particle collider.
Creepy Humanoid Robo-Artist Gives Public Performance Of Its Own AI-Generated Poetry
Ai-Da is an AI-driven robot whose creators claim that it can paint, draw, sculpt, and, write its own poetry. Last Friday, the robot gave a public performance of AI-generated poetry produced by first ingesting all 14,233 lines of Dante’s the “Divine Comedy” and then generating poetry in response to it. Its creator, Gallerist Aidan Meller, did engage in some “restrictive editing” of her content, but the poetry is primarily the product of the AI. An excerpt of this poem:
“We looked up from our verses like blindfolded captives, / Sent out to seek the light; but it never came / A needle and thread would be necessary / For the completion of the picture. / To view the poor creatures, who were in misery, / That of a hawk, eyes sewn shut.”
According to Meller, Ai-Da was designed “as a tool to glean insight into our own patterns of behavior to build better strategies in the face of an increasingly online world”.
China backs UN pledge to ban (its own) social scoring
On Tuesday, November 23, UNESCO's 193 member countries approved a first-of-its-kind AI ethics recommendation. The focus is encouraging countries to steer clear of dangerous use cases for AI because of their impacts on human rights. Interestingly, China also signed this document, despite the fact that Beijing has popularized 'social scoring' systems that the document recommends banning. The text also calls on governments to conduct ethical impact assessments and to put in place enforcement mechanisms for protecting human rights.
Our take: It is indeed interesting that China signed this bill, and it follows on the heels of Beijing's recent AI ethics recommendations. I think, in terms of increasing its standing in the world, it is in China's interest to engage in developing such principles. It remains to be seen whether Beijing will follow them or not.
The article includes a quote that I want to take some time to comment:
Whenever you are not certain that the development of certain technologies is going to have a negative impact but you assume that they might — don’t do it. It’s as simple as that.
I think this statement should not be taken at face value. The development of the Internet itself is something that has had a negative impact, but had we applied this principle then you wouldn't be reading these words right now. There are certainly applications of AI that seem pretty obviously bad, but taking this principle in all instances might mean dismissing applications that will bring both upsides and downsides. I would agree with commentators like Ben Thompson on the idea that the development of technologies like AI and the Internet is amoral, and that we need to do a lot more than dismiss a technology as dangerous before deciding to dispense of it. I don't think AI is going to "wreak havoc" on societies or that imagining it will is a useful way to think. If we make decisions based on principles like "this might have negative consequences so let's not develop it," we clearly haven't thought through the issues in nearly enough depth.
Other News
Research
A New Research On Unsupervised Deep Learning Shows That The Brain Disentangles Faces Into Semantically Meaningful Factors, Like Age At The Single Neuron Level - "Faces appear to be encoded in low-dimensional neural codes inside such patches, with each neuron encoding an orthogonal axis of variation in the face space."
DeepMind, Google Brain & World Chess Champion Explore How AlphaZero Learns Chess Knowledge - "Deep neural networks are known to learn opaque, uninterpretable representations that lie beyond the grasp of human understanding."
70-Page Paper From Yoshua Bengio Team: GFlowNet Foundations - "In the new paper GFlowNet Foundations, a research team from Mila, University of Montreal, McGill University, Stanford University, CIFAR and Microsoft Azure AI builds upon GFlowNets, providing an in-depth formal foundation and expansion of the set of theoretical results for a broad range of scenarios, especially active learning."
AI Reveals Previously Unknown Biology – We Might Not Know Half of What’s in Our Cells - "Artificial intelligence-based technique reveals previously unknown cell components that may provide new clues to human development and disease."
A deep learning method to automatically enhance dog animations - "Researchers at Trinity College Dublin and University of Bath have recently developed a model based on deep neural networks that could help to improve the quality of animations containing quadruped animals, such as dogs."
MetaICL: A New Few-Shot Learning Method Where A Language Model Is Meta-Trained To Learn To In-Context Learn - "Researchers from Facebook AI, the University of Washington, and the Allen Institute for AI have developed Meta-training for In-Context Learning (MetaICL), a new few-shot learning meta-training paradigm."
Applications
This AI tool lets you visualize how climate change could affect your home - "The project, titled "This Climate Does Not Exist," lets you enter the address of your current home or your favorite travel destination and see what it could look like years later once climate change has taken its toll."
Australia’s Victoria testing AI drone tech in road repair ‘blitz’ - "The government of Australia’s Victoria state says it is testing high-resolution sensors, artificial intelligence, and drones as part of what it’s calling a “regional road maintenance blitz” to renovate its aging ground transport network."
Can AI Truly Give Us a Glimpse of Lost Masterpieces? - "Recent projects used machine learning to resurrect paintings by Klimt and Rembrandt. They raise questions about what computers can understand about art."
Firefighting Robots Go Autonomous - "Both independent and remote-controlled machines can save lives"
diavlex text-to-paint using Neural Painters and CLIP - "Our goal is to teach a machine how to paint using a combination of colors and strokes by telling it in natural language what to paint. How can we achieve this? Artists combine colors and brushstrokes to paint their masterpieces, they do not create paintings pixel by pixel."
Talking sense: using machine learning to understand quotes - "The Guardian’s data scientists have been working with other newsrooms on a global project to think about AI and journalism. Here they explain how they have been teaching a machine to understand what a quote is"
Business
Nimble Robotics details uptake for its AI-enabled picking robots - "Nimble Robotics, a robotics and e-commerce fulfillment technology company, today shared details on the rapid adoption of its solutions. Nimble also shared that it is working with many well-known brands including Best Buy, Victoria’s Secret, Puma, NFI/CalCartage, iHerb, Adore Me, Weee! and others."
DoorDash is finally going public with its robotics division after poaching a leader from an Amazon-owned self-driving startup - "The nation's No. 1 restaurant-delivery operator by market share said it had been operating a stealth robotics division for three years called DoorDash Labs. The division is now ready to go public under Ashu Rege, who was poached this month from Zoox, an Amazon-owned self-driving startup."
Verbit lands $250M Series E at a $2B valuation to expand its AI-powered transcription platform - "Less than six months after raising a $157 million Series D round, AI-powered transcription and captioning platform Verbit announced today it has closed $250 million in a Series E round that values the company at $2 billion. The latest funding brings its total raised to more than $550 million. "
Gartner says AI software market will reach $62 billion next year - "The analyst firm Gartner has said that artificial intelligence software will make up $62.5 billion in revenue in 2022. If it manages to reach that number, the sector will have had a revenue boost of 21.3% compared to this year."
In a crowded global market, Canadian AI startups’ fundraising results stand out - "It’s boom times for startups building with or atop AI functionality: The Exchange explored the rise in venture capital dollars for AI startups last week, noting that investment into the business niche set all-time records in Q4 2020, and then successively in quarters one, two and three in 2021."
Concerns
Revealed: the software that studies your Facebook friends to predict who may commit a crime - "What do your Facebook posts, who you follow on Instagram and who you interact with the most on social media say about you? According to the tech startup Voyager Labs, that information could help police figure out if you have committed or plan to commit a crime."
Facebook’s race-blind practices around hate speech came at the expense of Black users, new documents show - "Researchers proposed a fix to the biased algorithm, but one internal document predicted pushback from “conservative partners.”"
Representation and Imagination for Preventing AI Harms - "The AI Incident Database launched publicly in November 2020 by the Partnership on AI as a dashboard of AI harms realized in the real world."
Uncovering bias in search and recommendations - "Bias is a problem that’s exceedingly easy to introduce into any system and exceedingly hard to identify and address."
Artificial intelligence may not actually be the solution for stopping the spread of fake news - "Disinformation has been used in warfare and military strategy over time. But it is undeniably being intensified by the use of smart technologies and social media."
Policy
Europe’s AI laws will cost companies a small fortune – but the payoff is trust - "Artificial intelligence isn’t tomorrow’s technology — it’s already here. Now too is the legislation proposing to regulate it."
AI played 'big role' in approach to pandemic, says UAE artificial intelligence minister - "The UAE approached the COVID-19 pandemic “as a scientist,” said Omar Al Olama, the UAE’s Minister of Artificial Intelligence, Digital Economy and Remote Work Applications."
South Korea starts work on $330m national AI complex - "The National AI Industrial Convergence Complex will be developed over the next three years. When complete in 2024 it will offer office space and a data centre to some 77 start-ups looking to develop commercial AI projects."
Expert Opinions
Can humans and AI work side by side? A guide to what comes next - "A human perspective shift for the 21st Century."
Explainers
Doorways - "There are infinite worlds to explore. All of these images were synthesized using two machine learning models, VQGAN and CLIP."
Podcast
Check out our audio discussions of last week’s AI news stories!
Subscribe to the podcast here: Website | RSS | iTunes | Spotify | YouTube
Copyright © 2021 Skynet Today, All rights reserved.