Lifeboat Foundation

Chinese artist responds to debate about data-scraping as he prepares for new collaboration with AI.

New sleep technology was a trend at CES 2024, and health company DeRucci showcased its new line of smart bed products — including a pillow to combat snoring.

At CES 2024, a Chinese-based company presented what it claims is the first smart, antisnoring pillow.

There is a growing need to develop methods capable of efficiently processing and interpreting data from various document formats. This challenge is particularly pronounced in handling visually rich documents (VrDs), such as business forms, receipts, and invoices. These documents, often in PDF or image formats, present a complex interplay of text, layout, and visual elements, necessitating innovative approaches for accurate information extraction.

Traditionally, approaches to tackle this issue have leaned on two architectural types: transformer-based models inspired by Large Language Models (LLMs) and Graph Neural Networks (GNNs). These methodologies have been instrumental in encoding text, layout, and image features to improve document interpretation. However, they often need help representing spatially distant semantics essential for understanding complex document layouts. This challenge stems from the difficulty in capturing the relationships between elements like table cells and their headers or text across line breaks.

Researchers at JPMorgan AI Research and the Dartmouth College Hanover have innovated a novel framework named ‘DocGraphLM’ to bridge this gap. This framework synergizes graph semantics with pre-trained language models to overcome the limitations of current methods. The essence of DocGraphLM lies in its ability to integrate the strengths of language models with the structural insights provided by GNNs, thus offering a more robust document representation. This integration is crucial for accurately modeling visually rich documents’ intricate relationships and structures.

In today’s column, I will examine closely the recent launch of the OpenAI ChatGPT online GPT store that allows users to post GPTs or chatbots for ready use by others, including and somewhat alarmingly a spate of such chatbots intended for mental health advisory purposes.

OpenAI has launched their awaited GPT Store. This is great news. But there are also mental health GPTs that are less than stellar. I take a close look at the issue.

From blanket bans to specific prohibitions

Previously, OpenAI had a strict ban on using its technology for any “activity that has high risk of physical harm, including” “weapons development” and “military and warfare.” This would prevent any government or military agency from using OpenAI’s services for defense or security purposes. However, the new policy has removed the general ban on “military and warfare” use. Instead, it has listed some specific examples of prohibited use cases, such as “develop or use weapons” or “harm yourself or others.”

PLA scientists are reportedly using AI and large language models like Baidu’s Ernie to train a military AI system that can better predict the behavior of human adversaries.

Chinese scientists have allegedly combined AI and LLM to enhance the accuracy of predicting human behavior during military conflicts.

And that, according to the researchers, is exactly what the AI did, identifying whether prints from different types of fingers came from the same person with 75 to 90 percent accuracy, the BBC reported.

“It is clear that it isn’t using traditional markers that forensics have been using for decades,” study co-author Hod Lipson, a roboticist at Columbia University, told the broadcaster.

The researchers trained their AI model on a database of 60,000 fingerprints. Lead author Gabe Guo, a senior undergrad at Columbia University, told CNN that the AI was able to look beyond finger features known as “minutiae” that detectives have relied on for centuries.

From Ola’s Krutrim to Reliance, Indian companies are racing to push the bar on generative AI with a focus on Indic language and cultures.

ICFO and Qurv researchers have fabricated a new high-performance shortwave infrared (SWIR) image sensor based on non-toxic colloidal quantum dots. In their study published in Nature Photonics, they report on a new method for synthesizing functional high-quality non-toxic colloidal quantum dots integrable with complementary metal-oxide-semiconductor (CMOS) technology.

Invisible to our eyes, shortwave infrared (SWIR) light can enable unprecedented reliability, function and performance in high-volume, computer vision first applications in service robotics, automotive and consumer electronics markets. Image sensors with SWIR sensitivity can operate reliably under adverse conditions such as bright sunlight, fog, haze and smoke. Furthermore, the SWIR range provides eye-safe illumination sources and opens up the possibility of detecting material properties through molecular imaging.

Colloidal quantum dots (CQD) based image sensor technology offers a promising technology platform to enable high-volume compatible image sensors in the SWIR. CQDs, nanometric semiconductor crystals, are a solution-processed material platform that can be integrated with CMOS and enables accessing the SWIR range. However, a fundamental roadblock exists in translating SWIR-sensitive quantum dots into key enabling technology for mass-market applications, as they often contain heavy metals like lead or mercury (IV-VI Pb, Hg-chalcogenide semiconductors). These materials are subject to regulations by the Restriction of Hazardous Substances (RoHS), a European directive that regulates their use in commercial consumer electronic applications.