While graph neural networks (GNNs) are widely used for node and graph representation learning tasks, the reliability of GNN uncertainty estimates under distribution shifts remains relatively under-explored. Indeed, while post-hoc calibration strategies can be used to improve in-distribution calibration, they need not also improve calibration under distribution shift. However, techniques which produce GNNs with better intrinsic uncertainty estimates are particularly valuable, as they can always be combined with post-hoc strategies later. Therefore, in this work, we propose G-$\Delta$UQ, a novel training framework designed to improve intrinsic GNN uncertainty estimates. Our framework adapts the principle of stochastic data centering to graph data through novel graph anchoring strategies, and is able to support partially stochastic GNNs. While, the prevalent wisdom is that fully stochastic networks are necessary to obtain reliable estimates, we find that the functional diversity induced by our anchoring strategies when sampling hypotheses renders this unnecessary and allows us to support G-$\Delta$UQ on pretrained models. Indeed, through extensive evaluation under covariate, concept and graph size shifts, we show that G-$\Delta$UQ leads to better calibrated GNNs for node and graph classification. Further, it also improves performance on the uncertainty-based tasks of out-of-distribution detection and generalization gap estimation. Overall, our work provides insights into uncertainty estimation for GNNs, and demonstrates the utility of G-$\Delta$UQ in obtaining reliable estimates.

We introduce the package ddml for Double/Debiased Machine Learning (DDML) in Stata. Estimators of causal parameters for five different econometric models are supported, allowing for flexible estimation of causal effects of endogenous variables in settings with unknown functional forms and/or many exogenous variables. ddml is compatible with many existing supervised machine learning programs in Stata. We recommend using DDML in combination with stacking estimation which combines multiple machine learners into a final predictor. We provide Monte Carlo evidence to support our recommendation.

Americans in Silicon Valley fear advanced artificial intelligence in campaign ads will influence and manipulate voters' decisions in the 2024 election. Americans in Silicon Valley are predicting advanced artificial intelligence could significantly influence and manipulate voters in the 2024 elections, with a potential for "disturbingly false" political advertising to push agendas. "I've seen some hilarious videos and some concerning ones where it's getting too realistic," Travis, of San Jose, said. As advanced artificial intelligence applications proliferate across industries, the rapidly evolving technology has raised concerns about its ability to manipulate elections, with some 2024 presidential campaigns already utilizing the tool. Former President Trump's presidential campaign, for example, triggered an uproar on X after using artificial intelligence to recreate Florida Gov. Ron DeSantis' 2024 presidential announcement with fictional guests, including billionaire Democratic donor George Soros, World Economic Forum Chair Klaus Schwab, former Vice President Dick Cheney, Adolf Hitler, the devil and the FBI.

Senior foreign affairs correspondent Greg Palkot provides details on the major strike on an Iraqi militia leader and the U.S.'s response to Houthi attacks in the Red Sea Iraq has condemned the United States after U.S. forces carried out a drone strike in central Baghdad on Thursday that killed a high-ranking militia commander. Iraqi Prime Minister Mohammed Shia al-Sudani said Friday that the U.S. targeting and killing Mushtaq Taleb al-Saidi -- or "Abu Taqwa," the leader of the Harakat Hezbollah al-Nujaba, an Iraqi Shi'ite militant group -- was a violation of Iraqi sovereignty. A U.S. defense official confirmed that Thursday's drone strike on a vehicle containing the militia leader and three other militia members was authorized as he was responsible for recent attacks on U.S. personnel. Sudani said Friday that the U.S. had bypassed the Iraqi government, which is "the body authorized to impose the law." In his statement, he also reiterated calls for U.S. troops to withdraw from the country.

Tesla is recalling more than 1.6m Model S, X, 3 and Y electric vehicles exported to China for problems with their automatic assisted steering and door latch controls. The recall, Tesla's largest ever in China, affects the majority of the cars it has sold in the country, according to Bloomberg and the Wall Street Journal. China's state administration for market regulation announced the recall on Friday. The agency said Tesla in Beijing and Shanghai would use remote upgrades to fix the problems, so in most cases car owners would not need to visit Tesla service centers. The recall follows another in the US last month of more than 2m Tesla EVs to improve its system for monitoring drivers. The Chinese recall, due to problems with the automatic steering assist function, applies to 1.6m imported Tesla Model S, Model X, Model 3 and Model Ys.

Abandoning fossil fuels and adopting lower-carbon technologies are our best options for warding off the accelerating threat of climate change. And access to rare earth elements, key ingredients in many of these technologies, will partly determine which countries will meet their goals for lowering emissions. Some nations, including the US, are increasingly worried about whether the supply of those elements will remain stable. As a result, scientists and companies alike are intent on increasing access and improving sustainability by exploring secondary or unconventional sources. This story is from the next magazine edition of MIT Technology Review, set to go live on January 8--and it's all about innovation.

If 2023 was the year lawmakers agreed on a vision, 2024 will be the year policies start to morph into concrete action. AI really entered the political conversation in the US in 2023. There was also action, culminating in President Biden's executive order on AI at the end of October--a sprawling directive calling for more transparency and new standards. Through this activity, a US flavor of AI policy began to emerge: one that's friendly to the AI industry, with an emphasis on best practices, a reliance on different agencies to craft their own rules, and a nuanced approach of regulating each sector of the economy differently. Next year will build on the momentum of 2023, and many items detailed in Biden's executive order will be enacted.

A U.S. Special Operations drone strike in Baghdad on Thursday killed a senior figure in an Iran-linked militant group that is part of Iraq's security apparatus, drawing sharp criticism from the Iraqi government, as well as allied groups. The Pentagon acknowledged responsibility for the strike, saying in a statement that U.S. forces had taken "necessary and proportionate action," adding that the attack "was taken in self-defense" and that no civilians had been harmed. A missile fired by the drone struck a vehicle carrying three men near the logistics headquarters for the 12th brigade of the group, Harakat al-Nujaba, killing a brigade commander known as Abu Taqwa and two others, according to Iraqi security officials. The group, closely linked to Iran, was designated as a global terrorist organization by the State Department in 2019. Nujaba, however, has remained part of Iraq's Popular Mobilization Forces, an umbrella organization that is in turn part of the government's broader security forces.

Artificial intelligence based on neural networks has made significant progress. However, there are concerns about the reliability and security of this approach due to its lack of transparency. This is the black box problem of AI. Here we show how this problem can be solved using symbolic AI, which has a transparent white box nature. The widespread use of symbolic AI is hindered by the opacity of mathematical models and natural language terms, the lack of a unified ontology, and the combinatorial explosion of search options. To solve the AI black box problem and to implement general-purpose symbolic AI, we propose to use deterministic logic cellular automata with rules based on first principles of the general theory of the relevant domain. In this case, the general theory of the relevant domain plays the role of a knowledge base for the cellular automaton inference. A cellular automaton implements automatic parallel logical inference at three levels of organization of a complex system. Our verification of several ecological hypotheses provides a successful precedent for the implementation of white-box AI. Finally, we discuss a program for creating a general-purpose symbolic AI capable of processing knowledge and ensuring the reliability and safety of automated decisions.

In the largest survey of its kind, 2,778 researchers who had published in top-tier artificial intelligence (AI) venues gave predictions on the pace of AI progress and the nature and impacts of advanced AI systems The aggregate forecasts give at least a 50% chance of AI systems achieving several milestones by 2028, including autonomously constructing a payment processing site from scratch, creating a song indistinguishable from a new song by a popular musician, and autonomously downloading and fine-tuning a large language model. If science continues undisrupted, the chance of unaided machines outperforming humans in every possible task was estimated at 10% by 2027, and 50% by 2047. The latter estimate is 13 years earlier than that reached in a similar survey we conducted only one year earlier [Grace et al., 2022]. However, the chance of all human occupations becoming fully automatable was forecast to reach 10% by 2037, and 50% as late as 2116 (compared to 2164 in the 2022 survey). Most respondents expressed substantial uncertainty about the long-term value of AI progress: While 68.3% thought good outcomes from superhuman AI are more likely than bad, of these net optimists 48% gave at least a 5% chance of extremely bad outcomes such as human extinction, and 59% of net pessimists gave 5% or more to extremely good outcomes. Between 38% and 51% of respondents gave at least a 10% chance to advanced AI leading to outcomes as bad as human extinction. More than half suggested that "substantial" or "extreme" concern is warranted about six different AI-related scenarios, including misinformation, authoritarian control, and inequality. There was disagreement about whether faster or slower AI progress would be better for the future of humanity. However, there was broad agreement that research aimed at minimizing potential risks from AI systems ought to be prioritized more.