Goto

Collaborating Authors

 Government


Analysis of Disinformation and Fake News Detection Using Fine-Tuned Large Language Model

arXiv.org Artificial Intelligence

The paper considers the possibility of fine-tuning Llama 2 large language model (LLM) for the disinformation analysis and fake news detection. For fine-tuning, the PEFT/LoRA based approach was used. In the study, the model was fine-tuned for the following tasks: analysing a text on revealing disinformation and propaganda narratives, fact checking, fake news detection, manipulation analytics, extracting named entities with their sentiments. The obtained results show that the fine-tuned Llama 2 model can perform a deep analysis of texts and reveal complex styles and narratives. Extracted sentiments for named entities can be considered as predictive features in supervised machine learning models.


Adaptive Selection of the Optimal Strategy to Improve Precision and Power in Randomized Trials

arXiv.org Machine Learning

Benkeser et al. demonstrate how adjustment for baseline covariates in randomized trials can meaningfully improve precision for a variety of outcome types. Their findings build on a long history, starting in 1932 with R.A. Fisher and including more recent endorsements by the U.S. Food and Drug Administration and the European Medicines Agency. Here, we address an important practical consideration: *how* to select the adjustment approach -- which variables and in which form -- to maximize precision, while maintaining Type-I error control. Balzer et al. previously proposed *Adaptive Prespecification* within TMLE to flexibly and automatically select, from a prespecified set, the approach that maximizes empirical efficiency in small trials (N$<$40). To avoid overfitting with few randomized units, selection was previously limited to working generalized linear models, adjusting for a single covariate. Now, we tailor Adaptive Prespecification to trials with many randomized units. Using $V$-fold cross-validation and the estimated influence curve-squared as the loss function, we select from an expanded set of candidates, including modern machine learning methods adjusting for multiple covariates. As assessed in simulations exploring a variety of data generating processes, our approach maintains Type-I error control (under the null) and offers substantial gains in precision -- equivalent to 20-43\% reductions in sample size for the same statistical power. When applied to real data from ACTG Study 175, we also see meaningful efficiency improvements overall and within subgroups.


IRS to use AI to crack down on ultra-rich taxpayers and partnerships

Washington Post - Technology News

According to the announcement, the program will target individuals with reported incomes above $1 million who have more than $250,000 in recognized tax debt. The agency cites 1,600 taxpayers in this category, collectively owing hundreds of millions of dollars in taxes. It also plans to expand its focus on digital assets, noting that 75 percent of taxpayers who use digital assets exchanges do not comply with tax laws.


Axon's Ethics Board Resigned Over Taser-Armed Drones. Then the Company Bought a Military Drone Maker

WIRED

This article was copublished with The Markup, a nonprofit, investigative newsroom that challenges technology to serve the public good. Less than 10 days after the Robb Elementary School shooting in Uvalde, Texas, in May 2022, Axon Enterprises CEO Rick Smith announced the company had formally started developing Taser-equipped drones. The technology, Smith argued, could potentially save lives during mass shootings by incapacitating active shooters within seconds. For Axon, which changed its name from Taser in 2017, the concept seemed a sensible next step for stakeholders who share Axon's public safety mission, Smith said on the company's site. "In brief," he wrote, "non-lethal drones can be installed in schools and other venues and play the same role that sprinklers and other fire suppression tools do for firefighters: Preventing a catastrophic event, or at least mitigating its worst effects."


Ukrainian official claims Elon Musk cost lives by refusing Starlink access during a drone operation

Engadget

Excerpts from Walter Isaacson's Elon Musk biography are coming to light ahead of its release next week, revealing some new details about the billionaire's decision to provide Ukraine with Starlink access amid the country's war with Russia. According to an excerpt CNN reported on, Musk allegedly told SpaceX workers to shut down Starlink access close to the Crimea coast to prevent a Ukrainian drone attack on Russia's naval fleet. Musk, who has reportedly been in contact with Russian officials including President Vladimir Putin, is said to have been worried that the attack would lead to Russia retaliating with nuclear weapons. Ukrainian leaders seemingly begged Musk to reactivate Starlink access but drones that were approaching Russian warships "lost connectivity and washed ashore harmlessly," CNN cites Isaacson as stating. Musk's alleged actions have had significant consequences for Ukraine, according to Mykhailo Podolyak, an advisor to President Volodymyr Zelensky.


Musk 'committed evil' with Starlink order, says Ukrainian official

The Guardian

A senior Ukrainian official has accused Elon Musk of "committing evil" after a new biography revealed details about how the business magnate ordered his Starlink satellite communications network to be turned off near the Crimean coast last year to hobble a Ukrainian drone attack on Russian warships. In a statement on X, the social media platform formerly known as Twitter, which Musk owns, the Ukrainian presidential adviser Mykhailo Podolyak wrote that Musk's interference led to the deaths of civilians, calling them "the price of a cocktail of ignorance and big ego". "By not allowing Ukrainian drones to destroy part of the Russian fleet via Starlink interference, @elonmusk allowed this fleet to fire Kalibr missiles at Ukrainian cities. As a result, civilians, and children are being killed," Podolyak wrote. "Why do some people so desperately want to defend war criminals and their desire to commit murder? And do they now realise that they are committing evil and encouraging evil?" Sometimes a mistake is much more than just a mistake.


Russia detains man accused of plotting rail bombing in Crimea

Al Jazeera

Russia's Federal Security Service (FSB) has detained a man for plotting a rail bombing in Crimea as a drone was downed over the Moscow-occupied peninsula. Crimea, which Russia seized from Ukraine in 2014, has been targeted by Ukrainian drone raids and sabotage attacks as Kyiv tries to retake the peninsula. The FSB said on Friday the suspect โ€“ a Russian citizen in his mid-40s โ€“ had been "collecting information on the deployment of Russian defence ministry facilities and units" and was preparing a railway bombing. "In a hiding place he had organised [we] found and seized an improvised explosive device made using foreign-made plastic explosives," it said. It said the man had been acting on the "instructions of Ukrainian military intelligence" and had been remanded in custody. Russia's TASS news agency said the man was a resident of the Crimean city of Sevastopol.


Zelenskyy adviser claims Elon Musk allowed Russians to hit Ukrainian cities

Al Jazeera

Ukrainian presidential adviser Mykhailo Podolyak has slammed Elon Musk for indirectly allowing Russian forces to attack Ukrainian cities after it was revealed his Starlink satellite communications interfered with a drone operation. Details of the incident are laid out in a biography of Musk by Walter Isaacson, due out on Tuesday. The book describes how the network turned off communications near the coast of the Russian-occupied Crimean peninsula as Ukrainian drones were approaching Russian warships, resulting in "lost connectivity". Musk allegedly ordered Starlink engineers to turn off the communications as he feared Russian President Vladimir Putin would respond with nuclear weapons to a Ukrainian attack on Crimea, according to Isaacson's book. "I think if the Ukrainian attacks had succeeded in sinking the Russian fleet, it would have been like a mini Pearl Harbor and led to a major escalation," Musk is quoted as saying.


The Coming Wave by Mustafa Suleyman review โ€“ a tech tsunami

The Guardian

On 22 February1946, George Kennan, an American diplomat stationed in Moscow, dictated a 5,000-word cable to Washington. In this famous telegram, Kennan warned that the Soviet Union's commitment to communism meant that it was inherently expansionist, and urged the US government to resist any attempts by the Soviets to increase their influence. This strategy quickly became known as "containment" โ€“ and defined American foreign policy for the next 40 years. The Coming Wave is Suleyman's book-length warning about technological expansionism: in close to 300 pages, he sets out to persuade readers that artificial intelligence (AI) and synthetic biology (SB) threaten our very existence and we only have a narrow window within which to contain them before it's too late. Unlike communism during the cold war, however, AI and SB are not being forced on us.


Three top takeaways from the Senate Energy committee hearing on DOE and AI

FOX News

Fox News correspondent Gillian Turner has the latest on the president's focus amid calls for an impeachment inquiry on'Special Report.' Lawmakers on the Senate Energy Committee were warned on Thursday about both the threats and opportunities that come with artificial intelligence being integrated into the U.S. energy sector and everyday life as a whole. The committee held a hearing on the rapidly advancing technology, and experts present spent a significant amount of time not only discussing AI but the ever-looming threat of China and its efforts to steal and recreate emerging U.S. capabilities. "China released their new generation of AI Development Plan, which includes [research and development] and infrastructure targets. The U.S. currently does not have a strategic AI plan like this," Committee Chair Joe Manchin, D-W.Va., said at the hearing's outset.