We develop an algorithm to semantically parse linear ordering problems, which require a model to arrange entities using deductive reasoning. Our method takes as input a number of premises and candidate statements, parsing them to a first-order logic of an ordering domain, and then utilizes constraint logic programming to infer the truth of proposed statements about the ordering. Our semantic parser transforms Heim and Kratzer's syntax-based compositional formal semantic rules to a computational algorithm. This transformation involves introducing abstract types and templates based on their rules, and introduces a dynamic component to interpret entities within a contextual framework. Our symbolic system, the Formal Semantic Logic Inferer (FSLI), is applied to answer multiple choice questions in BIG-bench's logical_deduction multiple choice problems, achieving perfect accuracy, compared to 67.06% for the best-performing LLM (GPT-4) and 87.63% for the hybrid system Logic-LM. These promising results demonstrate the benefit of developing a semantic parsing algorithm driven by first-order logic constructs.

This story originally appeared on WIRED Japan and has been translated from Japanese. A project to develop autonomous vehicles for self-driving taxis has begun in earnest in Japan. The plan put forward by Tier IV, a startup specializing in autonomous-driving technology, has been selected for a demonstration project by the Japanese Ministry of Economy, Trade, and Industry. Now, a prototype development project has officially begun. Tier IV became known for developing open-source self-driving software and conducting demonstrations of self-driving taxis in May and June in Odaiba, an entertainment district of Tokyo.

While the recent Chain-of-Thought (CoT) technique enhances the reasoning ability of large language models (LLMs) with the theory of mind, it might still struggle in handling logical reasoning that relies much on symbolic expressions and rigid deducing rules. To strengthen the logical reasoning capability of LLMs, we propose a novel Symbolic Chain-of-Thought, namely SymbCoT, a fully LLM-based framework that integrates symbolic expressions and logic rules with CoT prompting. Technically, building upon an LLM, SymbCoT 1) first translates the natural language context into the symbolic format, and then 2) derives a step-by-step plan to solve the problem with symbolic logical rules, 3) followed by a verifier to check the translation and reasoning chain. Via thorough evaluations on 5 standard datasets with both First-Order Logic and Constraint Optimization symbolic expressions, SymbCoT shows striking improvements over the CoT method consistently, meanwhile refreshing the current state-of-the-art performances. We further demonstrate that our system advances in more faithful, flexible, and explainable logical reasoning. To our knowledge, this is the first to combine symbolic expressions and rules into CoT for logical reasoning with LLMs. Code is open at https://github.com/Aiden0526/SymbCoT.

Despite excellent average-case performance of many image classifiers, their performance can substantially deteriorate on semantically coherent subgroups of the data that were under-represented in the training data. These systematic errors can impact both fairness for demographic minority groups as well as robustness and safety under domain shift. A major challenge is to identify such subgroups with subpar performance when the subgroups are not annotated and their occurrence is very rare. We leverage recent advances in text-to-image models and search in the space of textual descriptions of subgroups ("prompts") for subgroups where the target model has low performance on the prompt-conditioned synthesized data. To tackle the exponentially growing number of subgroups, we employ combinatorial testing. We denote this procedure as PromptAttack as it can be interpreted as an adversarial attack in a prompt space. We study subgroup coverage and identifiability with PromptAttack in a controlled setting and find that it identifies systematic errors with high accuracy. Thereupon, we apply PromptAttack to ImageNet classifiers and identify novel systematic errors on rare subgroups.

Standing around an empty parking lot in Arizona on a sunny, 100 F summer day was not a pleasant way to spend an afternoon. But excitement and anticipation overpowered our fears of heat exhaustion as we waited for Waymo's self-driving taxi. I was there with Kelly Funkhouser, CR's manager of vehicle technology. We looked back and forth wondering which direction the minivan would arrive from. "What if it gets stuck on the way here and never shows up?" Kelly was a little more optimistic.

Led by Sumit Jha, professor in the Department of Computer Science at UTSA, the team has changed the conventional approach employed in explaining machine learning decisions that relies on a single injection of noise into the input layer of a neural network. The team shows that adding noise -- also known as pixilation -- along multiple layers of a network provides a more robust representation of an image that's recognized by the AI and creates more robust explanations for AI decisions. This work aids in the development of what's been called "explainable AI" which seeks to enable high-assurance applications of AI such as medical imaging and autonomous driving. "It's about injecting noise into every layer," Jha said. "The network is now forced to learn a more robust representation of the input in all of its internal layers. If every layer experiences more perturbations in every training, then the image representation will be more robust and you won't see the AI fail just because you change a few pixels of the input image."

Burn-In: A Novel of the Real Robotic Revolution is a technothriller that follows the hunt for a terrorist through the streets of a future Washington, DC. More than 300 factual explanations and predictions (with endnotes) are baked into the story, and the research for it ranged from assembling the latest job automation reports to interviews with AI scientists and water-system cybersecurity experts. This is the first chapter, where we meet the main character, FBI special agent Lara Keegan, who is responding to an emergency alert at Washington's Union Station. Soon Keegan will be assigned to test out a robotic policing tool and launched into a conspiracy whose mastermind is using cutting-edge tech to tear the nation apart. The man's greasy red beard and braided Viking-style Mohawk had likely not been washed in a couple weeks, but the way that he cradled his AR-15 assault rifle made it clear he took care of what most mattered to him.

Waymo, whose driverless minivans are already shuttling a limited number of passengers in suburban Phoenix, Arizona will soon begin delivering packages for UPS as part of a new strategic partnership announced on Wednesday. Why it matters: Widespread use of robo-taxis is still years away, but automated trucks are quickly gaining momentum toward deployment. Waymo's ambition is to use the same self-driving technology in its minivans -- what it calls the Waymo driver -- to automate big rigs and delivery trucks like the ones UPS uses every day.

I'm in the middle seat of a Chrysler Pacifica minivan, heading north on Dobson Road in Chandler, Arizona, when I notice we may have taken a wrong turn. Under normal circumstances, I would just lean forward and ask the driver for an explanation. There is, after all, no driver to ask. Last October, Alphabet's self-driving subsidiary Waymo emailed its customers in the suburbs of Phoenix to let them know that "completely driverless Waymo cars are on the way." For several years, Waymo has offered its autonomous taxi service to a small group of people, but the rides typically included a safety driver behind the steering wheel. Now, Waymo is saying more of those rides will take place sans safety driver, a sign that the company is growing confident in the accuracy of its technology.

Sometimes when you are on the brink of a rebellion, it's hard to see what's happening around you. Chandler, Arizona, has become a hot bed of attacks on autonomous vehicles (AVs). Over the past three years, people have assaulted self-driving cars in the city nearly two dozen times, pelting them with rocks, trying to run them off the road, challenging them to games of chicken, and slashing their tires. One man even threatened an AV with a .22-caliber But police chief Sean Duggan says Chandler is "absolutely not" at the forefront of a rebellion between humans and machines.