Get 1,200 artificial intelligence (AI) researchers and 2,500 other businesspeople and academics, such as Elon Musk, Stephen Hawking, Ray Kurzweil, and David Chalmers, to all endorse one document about AI ethics. You have the Asilomar AI Principles with serious sound bite power: Experts agree on a humanistic AI ethics program! Do the Principles advance a worthy cause? Reading the text of the Asilomar Principles, however, you get a few vague ethical aspirations offered to guide a top-down control regime. The points do it subtly, so as the holographic Dr. Lanning advised in I, Robot (2004), "you have to ask the right questions."

What is the impact of artificial intelligence (AI) and big data on societies in the Indo-Pacific? How are countries using AI and big data to enhance their national security and advance their national interests? And what are the major regulatory issues? For a perspective on these and other matters, Jongsoo Lee interviewed Simon Chesterman, dean and provost's chair professor of the National University of Singapore Faculty of Law and senior director of AI Governance at AI Singapore. What are nations in the Indo-Pacific doing to develop their artificial intelligence (AI) and big data capabilities?

Goal-directed evaluation of Answer Set Programs is gaining traction thanks to its amenability to create AI systems that can, due to the evaluation mechanism used, generate explanations and justifications. s(CASP) is one of these systems and has been already used to write reasoning systems in several fields. It provides enhanced expressiveness w.r.t. other ASP systems due to its ability to use constraints, data structures, and unbound variables natively. However, the performance of existing s(CASP) implementations is not on par with other ASP systems: model consistency is checked once models have been generated, in keeping with the generate-and-test paradigm. In this work, we present a variation of the top-down evaluation strategy, termed Dynamic Consistency Checking, which interleaves model generation and consistency checking. This makes it possible to determine when a literal is not compatible with the denials associated to the global constraints in the program, prune the current execution branch, and choose a different alternative. This strategy is specially (but not exclusively) relevant in problems with a high combinatorial component. We have experimentally observed speedups of up to 90x w.r.t. the standard versions of s(CASP).

In the past ten years artificial intelligence has encountered such dramatic progress that it is seen now as a tool of choice to solve environmental issues and in the first place greenhouse gas emissions (GHG). At the same time the deep learning community began to realize that training models with more and more parameters required a lot of energy and as a consequence GHG emissions. To our knowledge, questioning the complete environmental impacts of AI methods for environment ("AI for green"), and not only GHG, has never been addressed directly. In this article we propose to study the possible negative impact of "AI for green" 1) by reviewing first the different types of AI impacts 2) by presenting the different methodologies used to assess those impacts, in particular life cycle assessment and 3) by discussing how to assess the environmental usefulness of a general AI service.

Machine unlearning, i.e. having a model forget about some of its training data, has become increasingly more important as privacy legislation promotes variants of the right-to-be-forgotten. In the context of deep learning, approaches for machine unlearning are broadly categorized into two classes: exact unlearning methods, where an entity has formally removed the data point's impact on the model by retraining the model from scratch, and approximate unlearning, where an entity approximates the model parameters one would obtain by exact unlearning to save on compute costs. In this paper we first show that the definition that underlies approximate unlearning, which seeks to prove the approximately unlearned model is close to an exactly retrained model, is incorrect because one can obtain the same model using different datasets. Thus one could unlearn without modifying the model at all. We then turn to exact unlearning approaches and ask how to verify their claims of unlearning. Our results show that even for a given training trajectory one cannot formally prove the absence of certain data points used during training. We thus conclude that unlearning is only well-defined at the algorithmic level, where an entity's only possible auditable claim to unlearning is that they used a particular algorithm designed to allow for external scrutiny during an audit.

The NFL and lawyers for thousands of retired NFL players have reached an agreement to end race-based adjustments in dementia testing in the $1 billion settlement of concussion claims, according to a proposed deal filed Wednesday in federal court. The revised testing plan follows public outrage over the use of "race-norming," a practice that came to light only after two former NFL players filed a civil rights lawsuit over it last year. The adjustments, critics say, may have prevented hundreds of Black players suffering from dementia to win awards that average $500,000 or more. The Black retirees will now have the chance to have their tests rescored or, in some cases, seek a new round of cognitive testing, according to the settlement, details of which were first reported in The New York Times on Wednesday. "We look forward to the court's prompt approval of the agreement, which provides for a race-neutral evaluation process that will ensure diagnostic accuracy and fairness in the concussion settlement," NFL lawyer Brad Karp said in a statement. The proposal, which must still be approved by a judge, follows months of closed-door negotiations between the NFL, class counsel for the retired players, and lawyers for the Black players who filed suit, Najeh Davenport and Kevin Henry.

Artificial Intelligence Technology Solutions, Inc., ( OTCPK:AITX), a global leader in AI-driven security and productivity solutions for enterprise clients today increased its revenue guidance for its fiscal Q3 ending November 30, 2021. This press release features multimedia. AITX CEO Steve Reinharz introduces RAD's yet unnamed robotic dog, part of the RAD 3.0 product offering, at AITX Investors Open House held October 13, 2021, at the company's REX (RAD Excellence Center) in Detroit, Michigan (Photo: Business Wire) As previously disclosed, AITX through its wholly owned subsidiary Robotic Assistance Devices Inc. (RAD), is expected to close October 2021 with Recurring Monthly Revenue (RMR) of over $80,000. "Sales activity has now accelerated to the pace where we can anticipate topping the $100,000 RMR milestone by the end of November 2021," commented Mark Folmer, RAD COO and President. "This is such exciting progress for AITX. We expect significant revenue events before Q3 closes," said Steve Reinharz, CEO and President of AITX.

She has been described as "a vision of the future" who is every bit as good as other abstract artists today, but Ai-Da – the world's first ultra-realistic robot artist – hit a temporary snag before her latest exhibition when Egyptian security forces detained her at customs. Ai-Da is due to open and present her work at the Great Pyramid of Giza on Thursday, the first time contemporary art has been allowed next to the pyramid in thousands of years. But because of "security issues" that may include concerns that she is part of a wider espionage plot, both Ai-Da and her sculpture were held in Egyptian customs for 10 days before being released on Wednesday, sparking a diplomatic fracas. "The British ambassador has been working through the night to get Ai-Da released, but we're right up to the wire now," said Aidan Meller, the human force behind Ai-Da, shortly before her release. According to Meller, border guards detained Ai-Da at first because she had a modem, and then because she had cameras in her eyes (which she uses to draw and paint).

Existing intellectual property laws don't allow AI systems to be recognized as inventors, which threatens the integrity of the patent system and the potential to develop life-changing innovations. Current legislation only allows humans to be recognized as inventors, which could make AI-generated innovations unpatentable. This would deprive the owners of the AI of the legal protections they need for the inventions that their systems create. The Artificial Inventor Project team has been testing the limitations of these rules by filing patent applications that designate a machine as the inventor-- the first time that an AI's role as an inventor had ever been disclosed in a patent application. They made the applications on behalf of Dr Stephen Thaler, the creator of a system called DABUS, which was listed as the inventor of a food container that robots can easily grasp, and a flashing warning light designed to attract attention during emergencies.

Undesired bias afflicts both human and algorithmic decision making, and may be especially prevalent when information processing trade-offs incentivize the use of heuristics. One primary example is \textit{statistical discrimination} -- selecting social partners based not on their underlying attributes, but on readily perceptible characteristics that covary with their suitability for the task at hand. We present a theoretical model to examine how information processing influences statistical discrimination and test its predictions using multi-agent reinforcement learning with various agent architectures in a partner choice-based social dilemma. As predicted, statistical discrimination emerges in agent policies as a function of both the bias in the training population and of agent architecture. All agents showed substantial statistical discrimination, defaulting to using the readily available correlates instead of the outcome relevant features. We show that less discrimination emerges with agents that use recurrent neural networks, and when their training environment has less bias. However, all agent algorithms we tried still exhibited substantial bias after learning in biased training populations.