Using Social Media and Machine Learning to Spot Child Sexual Exploitation

Category Machine Learning

tldr

Social media researchers are creating a machine learning program that can detect unwanted sexual advances on Instagram. The program was trained on data from over 5 million direct messages, annotated by a group of young users who experienced conversations that felt unsafe or uncomfortable. The data is helping to increase protection for young users by identifying and cutting off risky interactions sooner.

hashtags

Revealing what lies beneath: New AI technique uncovering materials' secrets from the surface

Category Engineering

tldr

MIT researchers have developed a new AI technique to figure out what's inside a material simply by observing its surface. The technique can analyze surfaces of uniform, composite and even complex materials, and apply for biological tissues as well. It can identify any damages, cracks, stresses and details of internal microstructures, in a completely noninvasive way.

hashtags

Machine Learning for Detecting Risky Conversations on Instagram

Category Computer Science

tldr

A team of four research universities have proposed a way to use machine learning technology to flag risky conversations on Instagram without having to eavesdrop on them. This new system is 1,000 times more effective at detecting conversations that contained cyberbullying and other risk factors than existing software, as it takes context and intent into consideration.

hashtags

The Toxicity of ChatGPT: Unmasking the Dark Side of AI

Category Science

tldr

ChatGPT is a popular AI tool which recently has been receiving attention concerning the generation of racist and other harmful dialogue. Research from Allen Institute for AI, Princeton University and Georgia Tech has demonstrated its capacity to produce biased and hurtful commentary, leading to a warning to its users. OpenAI is currently working to rectify problems, however responses to the latest report is unheard of.

hashtags

The Discovery of the Aperiodic Monotile: How David Smith Changed Mathematics

Category Science

tldr

David Smith, a retired print technician from Yorkshire, England, and a team of mathematicians, programmers and tiling enthusiasts, have solved a nearly 60-year-old mathematical problem by discovering a single shape that tiles the infinite plane without repeating and overlapping. This discovery, known as the aperiodic monotile, has delighted mathematicians since its discovery by Robert Berger in the mid-1960s.

hashtags

AI-Powered Model EDGE Choreographs Human Dance Animation to Match Any Piece of Music

Category Computer Science

tldr

Stanford University researchers have developed an AI-enabled model called Editable Dance GEneration (EDGE) that can choreograph human dance animation to match any piece of music and be edited by the user. It will be formally introduced at the Computer Vision and Pattern Recognition conference in June, where anyone can pick the tune and watch as EDGE creates a new dance sequence on the 'EDGE Playground' website.

hashtags

Introducing LMQL: A New Programming Language that Interacts with Large Language Models

Category Computer Science

tldr

Researchers from ETH Zurich have developed a new programming language called LMQL that increases user control and interaction with large language models like ChatGPT. This programming language combines natural and programming language to guide the language models in a more precise way. LMQL enables a novel way of programming and provides safety constraints to prevent unexpected or controversial outputs.

hashtags

VCU's Security and Forensics Engineering Lab: Combating Digital Crime

Category Computer Science

tldr

VCU's Security and Forensics Engineering Lab, under the direction of Irfan Ahmed, focuses on protecting and investigating industrial control systems from malicious cyberattacks. Ahmed's team creates tools and simulations to investigate, detect, and secure an ever-evolving type of cybercrime. Their newly invented solutions are applied in the industrial sector and to assist law enforcement in preventing and responding to both digital and non-digital crimes.

hashtags

New Feature Selection Method Improves the Discrimination of Terms for Text Categorization

Category Computer Science

tldr

A research team from East China Normal University, led by Li Zhang, has proposed a new method, Max-Difference Maximization Criterion (MDMC), for estimating the importance of terms in text categorization. The method introduces a new weight based off of class information occupancy, which is then combined with Accuracy2 to yield better discrimination of terms. MDMC has been proven to outperform other methods such as ISCA, PCA, and NMF.

hashtags

How Feces Can Improve Our Understanding of Nutrition and Health

Category Technology

tldr

Analyzing feces has become a useful tool towards understanding nutrition and health. Fecal samples can indicate whether people have eaten individial foods, and studies to improve nutrition research that try to figure out how certain foods affect our health usually rely on volunteers to keep food diary which is inaccurate and incomplete. This study showed that microbes found in feces can indicate food consumption with 80-87% accuracy.

hashtags

The Issue of Moderation in the Metaverse

Category Technology

tldr

Ravi Yekkanti is an employee of WebPurify, a content moderation services provider to internet companies such as Microsoft and Play Lab. His job is to make sure everyone in the metaverse is safe and having a good time, as reports of sexual assaults, bullying, and child grooming in the metaverse arise. His job is complex, murky and in-demand, as the burden of safety in the metaverse is left to private security agents such as Yekkanti to intervene.

hashtags

Engaging Users with AI Chatbots for Improved Critical Thinking

Category Artificial Intelligence

tldr

A new study from MIT and Columbia University has found that providing AI chatbot users with questions as opposed to ready-made answers can help improve their critical thinking skills. This could be beneficial in ensuring that AI-generated information is accurate and not overdependent. These new methods also offer users insight into an AI system’s decision-making process.

hashtags

The Paradox of Generative AI: Is AI-Generated Content an Artistic Revolution or a Form of Cheating?

Category Artificial Intelligence

tldr

The heyday of generative AI has been ushering in a debate over what constitutes art, with AI-generated images and text winning awards and prizes. This raises serious questions about privacy, creativity, and the validity of AI-generated content as an art form. As technology advances, chances of AI-generated content becoming an impediment to creativity or a tool unwittingly used for nefarious purposes continue to increase.

hashtags

Exploring the Use of AI in Fashion Design

Category Computer Science

tldr

Researchers at University of Florence, University of Modena and Reggio Emilia and University of Pisa recently introduced a new computer vision framework that could help fashion designers to visualize their designs, using AI models in fashion design. This new framework could help fashion designers find new inspiration and identify potential issues. The models used in this area were based on latent diffusion models, which could produce higher quality images, rather than generative adversarial networks (GANs). The researchers also created new datasets for training the model and showed good performance in initial evaluations.

hashtags

Aging Molecular Process Unlocks a Piece of the Puzzle

Category Science

tldr

A new study in Nature has found a critical molecular process in five species that powers every single body cell that degrades with age. With two interventions, scientists were able to slow down transcription in multiple species, including mice, thus potentially increasing lifespan.

hashtags

Evolution at Warp Speed: Building Proteins with AI

Category Artificial Intelligence

tldr

Building proteins with AI has two main strategies - the IKEA approach and starting with a vision and design tailored to individual needs. The University of Washington used a top-down approach grounded in AI reinforced learning to design structures, such as a 20-sided shell that can be used in medicine and gene therapies, that outperform the latest vaccine candidates in clinical trials.

hashtags

Survey Says: Americans Care About Climate Change But Don't Want to Pay

Category Science

tldr

A survey by the Energy Policy Institute at the University of Chicago in anticipation of Earth Day found that Americans believe climate change is real, but they're mostly not willing to spend money or go out of their way to help fix it because of economic considerations.

hashtags

Exploring the Booming Market of Electric Vehicles

Category Science

tldr

The demand for electric vehicles has been booming over the past decade, with more than 10 million sold in 2022 and a 35 percent growth predicted for this year. EV market share more than tripled from 2020 to 2022, with China leading the global charge due to small, cheap models of EVs. Europe is the second-largest EV market, followed by the US with a 55 percent growth. Reducing reliance on combustion engine vehicles is an important part of the transition to renewable energy, though EVs also come with their own set of challenges and drawbacks.

hashtags

Self-Driving Cars Should Learn to Communicate with Cyclists

Category Computer Science

tldr

In order to reduce the number of cyclist accidents on the roads, researchers from the University of Glasgow are suggesting that future generations of self-driving cars should learn to interact with cyclists. This involves the AVs having displays to signal their intentions, and cyclists wearing smart glasses that also indicate what the vehicle is doing.

hashtags

Pagination: page = 120, postsCount = 2323, postsPerPage = 19, totalPages = 123