Articles tagged with Ai Common Sense Knowledge
Researchers found that realistic AI avatars are rated more positively than cartoon-style ones for perceived competence, integrity, and benevolence. However, individual factors such as prior AI knowledge and trust in science moderate perceptions of trustworthiness.
A recent study from UTSA researchers reveals that large language models (LLMs) can pose a serious threat to programmers who use them to help write code. The study found that up to 97% of software developers incorporate generative AI into their workflow, and 30% of code written today is AI-generated.
A recent study suggests that AI systems replicate and amplify existing social inequalities when left unchecked, particularly in high-stakes areas like hiring and welfare distribution. The researchers argue for the need for greater transparency, inclusivity, and accountability in AI governance to ensure its use serves social justice.
Researchers at the University of Washington developed a new method called variational preference learning, which predicts users' preferences and tailors its outputs accordingly. This approach overcomes the limitations of traditional RLHF, which can lead to averaging user preferences and resulting in incorrect outputs for all users.
Researchers created a measurement scale to assess robot human likeness, revealing four key qualities: appearance, emotional capacity, social intelligence, and self-understanding. To seem lifelike, robots must exhibit these traits, with self-understanding being the most challenging aspect to simulate.
Researchers tested 24 MLLMs on Raven's Progressive Matrices, finding that open-source models struggled significantly. However, closed-source models like GPT-4V performed relatively well, suggesting a need for more advanced resources and training data to improve AI's cognitive abilities.
The study challenges the idea of creating artificial general intelligence (AGI) with human-level cognition, citing limitations in replicating human cognition. Researchers argue that even under ideal circumstances, it is impossible to achieve AGI due to the complexity of cognitive processes.
A recent study published in PNAS Nexus found that widespread adoption of large language models like ChatGPT led to a significant decline in user activity on Stack Overflow. The study highlights the impact of ChatGPT on public knowledge sharing and its implications for AI's future.
A new study published in JAMA Health Forum found that machine learning can be more effective than traditional methods for distributing scarce treatments to patients most vulnerable during a public health crisis. The model reduces expected hospitalizations by about 27 percent compared to actual and observed care.
A recent study published with the Arkansas Agricultural Experiment Station found that ChatGPT can be a useful tool in teaching agriculture students to program microcontrollers. The study, which aimed to determine the confidence of undergraduate agricultural students in using ChatGPT to write Arduino code for moderately difficult progra...
Researchers created a data set of over 10 million documents to test detection ability in current and future detectors. They found that most detectors only work well in specific use cases and can be easily evaded by manipulating the text. The new tool, RAID, aims to provide a standardized benchmark for robust detection.
New research finds that AI explanations can fuel a perception of fairness without being grounded in accuracy or equity. Humans were more likely to override AI recommendations when explanations highlighted gender rather than task-relevance, but this did not improve decision-making accuracy.
Researchers are developing a framework to combine AI and human intelligence in process safety systems, aiming to enhance safety and efficiency. The study identifies challenges and benefits of using Intelligence Augmentation (IA) and proposes strategies for effective implementation to minimize risks.
Scientists at UVA and Toyota Research Institute create language representations of driving behavior to enable robots to associate words with environmental interactions. This allows cars to provide guidance and adjust speed in challenging situations, improving safety and usability.
A Carnegie Mellon University study finds that people with autism are using AI chatbots for workplace advice, but raises questions about the quality of the advice and the need for inclusivity. The research highlights the importance of involving individuals with autism in the development of technology to address their specific needs.
A new study by RMIT University found that only over a third of media organizations have an image-specific AI policy in place, highlighting the need for clearer guidelines. The research highlights the challenges of navigating generative AI technologies in visual journalism, including algorithmic bias and copyright concerns.
Researchers created a system called Holodeck to generate interactive 3D environments, leveraging language models like ChatGPT to control it. The system outperformed earlier tools in evaluating realism and accuracy, with human evaluators preferring its outputs across various indoor environments.
Researchers from the University of Washington created an AI algorithm to analyze infant poses using limited training data. By leveraging generative AI, they were able to produce high-quality results, enabling parents to monitor their babies' daily activities and detect potential health issues early.
The JGU Center for Lifelong Learning is developing personalized AI-based learning experiences for adult learners, aiming to improve motivation and learning outcomes. The project will also discuss the challenges of AI technologies in teaching modern foreign languages.
The TRAILS AI Institute has awarded eight seed grants totaling $1.5 million to advance AI design, development and governance. The funded projects include developing AI chatbots for smoking cessation and designing animal-like robots for autism support.
A new AI system, Coscientist, has demonstrated its ability to autonomously learn about Nobel Prize-winning chemical reactions and design successful laboratory procedures. The system achieved this in just a few minutes, outperforming human chemists in some cases.
A new study investigates how people perceive and claim authorship of artificially generated texts, revealing that perceived ownership does not always align with declared authorship. Researchers found that participants who wrote the text themselves felt a stronger sense of ownership, while those who relied on AI ghostwriters did not.
The European Union's AI act could enable AI to access our subconscious minds, potentially leading to manipulation. According to Ignasi Beltran de Heredia, only 5% of brain activity is conscious, and the remaining 95% operates subconsciously, making it difficult for us to control or even be aware of.
A paper by Anthony Chemero explains how AI thinking differs from human thinking, highlighting the limitations of large language models trained on biased data. Despite generating impressive text, these models can make up facts and produce biased outputs due to their lack of embodiment and understanding of context.
A Lancaster University academic argues that AI and algorithms contribute to polarization, radicalism, and political violence, posing a threat to national security. The paper examines how AI has been securitized throughout its history, highlighting the need for better understanding and management of its risks.
A study of 33 physicians across 17 specialties found that chatbots provided largely accurate information to diverse medical queries. However, the chatbots had important limitations that require further research and model development.
Biased AI can limit climate predictions and misguide governments due to missing information from under-represented communities. Human-in-the-loop design can fill these 'data holes' by offering a sense check on used data and context.
A recent study found that Google is more current but biased by advertisers, while ChatGPT is more objective but outdated. The researchers suggest combining both strengths to build a better system.
Researchers developed a tool called SKILL that enables AI agents to learn 102 distinct tasks by sharing knowledge in parallel, reducing the time needed to master new skills. The technology has potential applications in medicine, education, and other fields where vast knowledge is required.
Researchers have developed an AI algorithm that accurately estimates coastal fish stocks, providing critical data for sustainable management. The tool has the potential to save millions of dollars in annual research costs and bridge the global data and sustainability divide.
A team of researchers developed an unsupervised entity alignment framework to improve knowledge graph search, avoiding human labor. The framework outperformed most competitors on precision and recall, scoring higher overall across multiple datasets.
A new neural network, CD-GAN, uses common sense knowledge to enhance text descriptions and generate images of birds at three resolution levels. The system achieved competitive scores against other image generation methods, producing vivid and natural-looking images.
A new study by the University of Maryland School of Medicine found that ChatGPT provides correct information on breast cancer screening for most questions. However, inconsistencies and outdated information were detected in some responses. The researchers recommend relying on doctors for advice due to the limitations of ChatGPT.
Researchers seek to develop algorithms providing meaningful explanations for AI decision-making, enabling higher human trust and adoption in fields like science. The project focuses on symbolic reasoning and estimating explanation accuracy, addressing the need for transparent AI systems.
The NERVE Center has developed test methods and metrics for various robots, identifying limitations to improve systems. The center's success grew its research capabilities through partnerships with NIST and the U.S. Army.
Researchers at North Carolina State University developed a blueprint for incorporating ethical guidelines into AI decision-making programs. The new mathematical formula, based on the Agent, Deed, and Consequence (ADC) Model, considers intent, character, and consequences of actions to make more informed decisions.
A study by NTU Singapore found that people have lower trust in health interventions suggested by AI alone, compared to those they perceive as based on human expert opinion. Emphasizing the involvement of a human health expert can improve acceptance and effectiveness of AI-suggested preventive care measures.
Researchers at USC's Information Sciences Institute developed a method to train AI to understand analogies in Aesop's fables, enabling it to make creative connections between familiar and novel situations. The study found that humans approach analogical reasoning subjectively and interpretively, influencing the outcome.