A key component of fact verification is thevevidence retrieval, often from multiple documents. Recent approaches use dense representations and condition the retrieval of each document on the previously retrieved ones. The latter step is performed over all the documents in the collection, requiring storing their dense representations in an index, thus incurring a high memory footprint. An alternative paradigm is retrieve-and-rerank, where documents are retrieved using methods such as BM25, their sentences are reranked, and further documents are retrieved conditioned on these sentences, reducing the memory requirements. However, such approaches can be brittle as they rely on heuristics and assume hyperlinks between documents. We propose a novel retrieve-and-rerank method for multi-hop retrieval, that consists of a retriever that jointly scores documents in the knowledge source and sentences from previously retrieved documents using an autoregressive formulation and is guided by a proof system based on natural logic that dynamically terminates the retrieval process if the evidence is deemed sufficient. This method is competitive with current state-of-the-art methods on FEVER, HoVer and FEVEROUS-S, while using $5$ to $10$ times less memory than competing systems. Evaluation on an adversarial dataset indicates improved stability of our approach compared to commonly deployed threshold-based methods. Finally, the proof system helps humans predict model decisions correctly more often than using the evidence alone.

The Navy is very much on board for integrating artificial intelligence (AI) and machine learning into its networks, but human decision makers must always be part of the decision process in warfighting, an admiral said. "From a warfighting perspective, artificial intelligence subsets would be enablers or augments to the human in the loop," said Rear Adm. Paul Spedero Jr., director, Fleet Integrated Readiness and Analysis, U.S. Fleet Forces Command, speaking April 8 during a Navy League webinar sponsored by Deloitte. "That has always been our approach. I don't see that changing. There are some things that can't be replaced; the experience of a seasoned warfighter in the field being able to assess things that a machine -- no matter how much we teach it -- may never be able to pick up on. There's always going to be a necessity for [experience-based decision making]. That necessity for war fighting will never go away -- to have a human in the loop. "AI will be our wingmen," he said. "It will not be the lead in a fight." Spedero said in the world of data analysis, his current focus, there "certainly is a place for AI, particularly machine learning, as we try to get to that predictive and prescriptive level of data analytics.

ARLINGTON, Va. — The Navy is very much on board for integrating artificial intelligence (AI) and machine learning into its networks, but human …

L'olivier - assurance auto, an auto insurer and French subsidiary of the English group Admiral, a European leader in automobile insurance, is joining InsurTech Shift Technology to fight against car insurance fraud. According to insurance magazine l'Argus de l'Assurance, in 2014 fraud represented 2.5 billion Euros in damages, only €219M of which have been recovered by insurers. Faced with such high figures, which are also on the increase, the aim of L'olivier is to rely on artificial intelligence and data science to automate and optimize the detection of suspicious claim files. For L'olivier, a direct insurer created in 2011, it was important to reinforce its plan to fight against fraud by establishing a dedicated unit: "Shift will help us in building out our capabilities. Their solution improves our ability to detect fraud by reducing the number of irrelevant cases, at the same time as it enhances our ability to prove fraud by providing administrators with avenues of investigation, enabling us to avoid payment of fraudulent claims," explains Janny Druon, head of the Claims Analytics team at L'olivier.

A Cambridge-based start-up believes machine learning software is the key to autonomous vehicles and Wayve is developing machine learning algorithms for autonomous vehicles. Wayve, which includes the chief scientist at Uber amongst its investors, believes the industry has been doing too much hand-engineering and too little machine learning. The firm is hiring for positions in its Cambridge-based headquarters. "The missing piece of the self-driving puzzle is intelligent algorithms, not more sensors, rules and maps. Humans have a fascinating ability to perform complex tasks in the real world, because our brains allow us to learn quickly and transfer knowledge across our many experiences. We want to give our vehicles better brains, not more hardware."

Facebook has banned an insurance company from using people's profiles to price their insurance. Just hours after Admiral announced that it would launch a new app that scoured Facebook profiles and tried to work out their personalities, the site has said that the plan breaks its terms and so will be banned. Admiral had hoped that by using Facebook information it could build up a picture of people, and hopefully work out whether they were more or less likely to crash. Customers could be given up to a 15 per cent discount if they signed up to the app. Connected company president Shigeki Tomoyama addresses a press briefing as he elaborates on Toyota's "connected strategy" in Tokyo.

Drones won't replace proper sailors anytime soon because, believe it or not, they need more manpower to operate, a Royal Navy admiral has insisted. Naval drones are "not about reducing the requirement for people", Rear Admiral Paul Bennett told a press briefing attended by El Reg on Friday. Instead, they are for putting people into positions where they add "real value". At present, unmanned systems - drones - require on average something like four or five operators each, we understand. Rather than enabling cuts in manpower, if anything they require ever more personnel aboard ships to operate them; not a good situation to be in when the Navy is already critically short of heads.