1. Microsoft announced its AI for Humanitarian Action initiative. The program is a five-year, $40 million to support developers of AI systems for helping humanity. This is the company's third such initiative; in May they launched an AI for accessibility program to help people with disabilities. In 2017, Microsoft kicked off its AI for Earth Innovation Grant program to develop technology to help the planet. The company revealed the AI for Humanitarian Action plan at the  Microsoft Ignite conference. — VENTUREBEAT

2. Google is launching Priority, a new AI feature for Google Drive. Priority was first announced earlier this year at the company's Cloud Next 18 event and is available now in beta. Priority appears in the left-hand column and offers quick access to the files and comments deemed most important by machine learning tools that take into account calendars and activities. — ENGADGET

-- Commentary --

In an interesting post this week, Per Bylund argues that robots can't replace humans because even if they replace us as workers, they can't replace us as consumers.  This is, in my opinion, the biggest conceptual gap in current thinking on AI.  I wrote about this a few weeks ago but I want to revisit it here and clarify the argument a bit.

The most powerful force in the world, in my opinion, is human inertia.  It is the reluctance of humans to change the way we do things.  This is why we are stuck with so many path dependent historical accidents that we now accept as gospel.  Therefore one of the biggest challenges in rolling out AI technology is the adaptability (or lack of) by humans.  As technology innovation continues to increase at an exponential rate, the gating factor to realizing the benefits of new technology will be the speed at which humans can adapt so they can adopt it.

This is important because the coming technology world is going to destroy our faith in many things that we currently believe.  (For a great treatment of this, check out Yuval Harari's "Homo Deus").  Capitalism has worked wonderfully well, but many pundits are speculating that it may not work so well in the future.  It is a system that is based on growth, and that growth, if humans are the primary consumers, may slow considerably.  But transitioning to a brand new system will be difficult and jarring and will probably not happen slowly.  If it happens, it will come fast and be part of some radical event that happens in the world.  I'm betting though, that we try our best to keep capitalism and shape it in an AI world.

The only way that can work is if two things happen.  First, AI's become consumers.  This is what Per Byland isn't thinking about in his article.  I think this is a long way off, > 15 years, but, I believe it will happen.  It will happen because the nature of capitalism is competition.  As your company and my company both design and sell AIs to compete for work, we will eventually stop programming the AIs to do things and instead teach them to mostly be learning machines.  To do such learning, we will have to give them desires and goals.  This will lead to them deciding how to augment themselves.  It's very conceivable that someday trillions of machines are buying things (skills, knowledge, help, work) from each other, and from humans.  This bot economy will eventually surpass the human economy and will continue to drive the growth of capitalism in a world where humans can't.  I know from the past discussion on this that many of you are extremely skeptical, but, no one has so far provided a logical train of thought as to why this doesn't make sense, all the responses were "that seems unlikely because AIs won't have desires."  If you have a stronger reason to believe this can't happen, I'd love to hear it.  But I believe that someday, companies will be talking about AI agents as their target market.

Thanks for reading, and enjoy the rest of your weekend.

@robmay

3. Seven members of the U.S. Congress have sent letters to federal agencies asking about biases in AI used for commerce, surveillance, and employment. Letters were sent to the Federal Trade Commission, FBI, and Equal Employment Opportunity Commission. Signatories include Senators Kamala Harris (D-CA), Cory Booker (D-NJ), Ron Wyden (D-OR), Richard Blumenthal (D-CT), Elizabeth Warren (D-MA), and Patty Murray (D-WA), and Representative Cedric Richmond (D-LA). The letters ask the agencies to respond by the end of this month and to determine whether the technology violates civil rights. — QUARTZ

4. A coalition of humanitarian organizations is partnering with tech companies to launch an AI tool to identify areas most likely to have food shortages. The coalition includes the World Bank, the United Nations and the International Committee of the Red Cross, and the algorithm is called the Famine Action Mechanism (FAM). Famine is the product of many different variables, including war, conflict, food price inflation, political instability, drought, or too much rain; FAM will give the agencies the ability to forecast and respond to potential crises. — WASHINGTON POST

5. Some researchers believe that the first human being to live for 1,000 years may have already been born. Co-founder of the SENS Research Foundation Aubrey de Grey says scientists are working to solve the seven types of aging damage that prevent people from living much longer. Founder and CEO of SingularityNET Ben Goertzel says that AI will play a key role in simulating, predicting, and understanding the effects of therapies to stop aging damage. — INVERSE

6. Researchers from the AI robotics company Kindred published a paper about establishing benchmarks for reinforcement-learning algorithms for robots. The paper's authors, A. Rupam Mahmood, Dmytro Korenkevych, Gautham Vasan, William Ma, and James Bergstra, used three commercial robots to test the performance of four algorithms; most studies use simulated robots in a software program. According to their research, the reinforcement-learning solutions were not as effective as the traditional script training methods. The team presented their paper last week at the Conference on Robot Learning in Zürich, Switzerland. — ZDNET

7. University of Connecticut cognitive scientist and philosopher Susan Schneider writes about the intersection of SETI (Search for Extraterrestrial Intelligence) and AI. Schneider makes the argument that if aliens do exist, it is possible that they are "postbiological," superintelligent machines. She says two theories make this plausible: the short window observation, which says that once a society has technology to reach space it is only a few hundred years away from a paradigm change from biology to AI; and the belief by many SETI researchers that any alien civilization is much older than ours and would be vastly more intelligent. — DAILY GALAXY

8. BMW engineers have developed an autonomous motorcycle. The company does not plan to produce and sell the self-driving motorcycles commercially. — THE STREET

9. The Royal Institution of Australia, publisher of Cosmos, is hosting a roundtable discussion about AI and machine learning in Adelaide, Australia on October 5. Guests include Alan Finkel, Australia’s Chief Scientist since 2016,  Anton van den Hengel, a computer science professor and director of the Australian Institute for Machine Learning, and Ellen Broad, an author and data policy expert. — COSMOS

10. Facebook AI researchers trained a system on 303,358 pairs of stories and writing prompts and had it generate its own prompts and stories. Human judges preferred the stories produced by the hierarchical structure they created over those produced by a conventional language models. Their research is published in the arXiv online repository. — @AEONOFDISCORD

1. Researchers from Chinese tech company Tencent developed two AI agents that beat the StarCraft II video game's AI cheater mode. The most difficult levels of play in StarCraft II pits players against the game's AI, which has unfair advantages over the players. Tencent's TSTARBOT1 neural network is a macro-level controller agent that monitors algorithms that handle lower level functions; the TSTARBOT2 neural network is a macro-micro controller with modules that work independently. The research is published in the arXiv online repository. — THE NEXT WEB

2. More than 2,000 AI experts in Europe are joining forces to collaborate on research and draw funding for projects. The alliance is called the Confederation of Laboratories for Artificial Intelligence Research in Europe, or CLAIRE. Scientists from 29 different countries have joined the group in an effort to keep the EU competitive with AI efforts in the U.S. and China. North America spent $23 billion on AI research in 2016, and China spent $7 billion, while the EU only spent $4 billion. — SOUTH CHINA MORNING POST

4. A team of AI scientists from Yale University, led by Rebecca Kramer-Bottiglio, developed an AI "skin" that can be wrapped around an object to turn it into a functioning robot. NASA had put out a call for soft robotics technology that an astronaut can wear and relay information about breathing, posture, and other data. The skin can also be used to animate objects and will be able to learn from the information it collects and adapt to perform multiple tasks. — DAILY STAR

5. Researchers at Deezer have developed a system that can associate certain music tracks with moods. The research has been published in the arXiv online repository. They designed a neural network and trained it using the Million Song Dataset (MSD) and 14,000 English words tagged by valence and intensity or "arousal," ranging from calm to energetic. The system was tested on audio signals and word2vec embeddings from lyrics, and it performed as well or better than existing classification methods. — VENTUREBEAT

6. Chinese e-commerce company Alibaba is launching AI robots for service industries, starting with the hospitality sectors. The robots are called Space Eggs; they are less than a meter tall and cased in aluminum with sensors. The robots have a semantic map, autonomous navigation, facial and voice recognition, and AliGenie voice assistant for responding to commands. They are being tested in hotels to perform tasks including delivering food and laundry to guests.  — ZDNET