Lifeboat Foundation

One of the most actively debated questions about human and nonhuman culture is this: Under what circumstances might we expect culture, in particular the ability to learn from one another, to be favored by natural selection?

Researchers at the Max Planck Institute for Evolutionary Anthropology in Leipzig, Germany, have developed a simulation model of the evolution of social learning. They showed that the interplay between learning, memory and forgetting broadens the conditions under which we expect to see social learning to evolve.

Social learning is typically thought to be most beneficial when the environments in which individuals live change quite slowly—they can safely learn tried and tested information from one another and it does not go out of date quickly. Innovating brand-new information, on the other hand, is thought to be useful in dynamic and rapidly changing environments.

To effectively assist humans in real-world settings, robots should be able to learn new skills and adapt their actions based on what users require them to do at different times. One way to achieve this would be to design computational approaches that allow robots to learn from human demonstrations, for instance observing videos of a person washing dishes and learning to repeat the same sequence of actions.

Researchers at University of British Columbia, Carnegie Mellon University, Monash University and University of Victoria recently set out to gather more reliable data to train robots via demonstrations. Their paper, posted to the arXiv preprint server, shows that the data they gathered can significantly improve the efficiency with which robots learn from the demonstrations of human users.

“Robots can build cars, gather the items for shopping orders in busy warehouses, vacuum floors, and keep the hospital shelves stocked with supplies,” Maram Sakr, one of the researchers who carried out the study, told Tech Xplore. “Traditional robot programming systems require an expert programmer to develop a robot controller that is capable of such tasks while responding to any situation the robot may face.”

Large language models (LLMs) are advanced deep learning algorithms that can process written or spoken prompts and generate texts in response to these prompts. These models have recently become increasingly popular and are now helping many users to create summaries of long documents, gain inspiration for brand names, find quick answers to simple queries, and generate various other types of texts.

Researchers at the University of Georgia and Mayo Clinic recently set out to assess the biological knowledge and reasoning skills of different LLMs. Their paper, pre-published on the arXiv server, suggests that OpenAI’s model GPT-4 performs better than the other predominant LLMs on the market on reasoning biology problems.

“Our recent publication is a testament to the significant impact of AI on biological research,” Zhengliang Liu, co-author of the recent paper, told Tech Xplore. “This study was born out of the rapid adoption and evolution of LLMs, especially following the notable introduction of ChatGPT in November 2022. These advancements, perceived as critical steps towards Artificial General Intelligence (AGI), marked a shift from traditional biotechnological approaches to an AI-focused methodology in the realm of biology.”

New research into large language models shows that they repeat conspiracy theories, harmful stereotypes, and other forms of misinformation.

In a recent study, researchers at the University of Waterloo systematically tested an early version of ChatGPT’s understanding of statements in six categories: facts, conspiracies, controversies, misconceptions, stereotypes, and fiction. This was part of Waterloo researchers’ efforts to investigate human-technology interactions and explore how to mitigate risks.

They discovered that GPT-3 frequently made mistakes, contradicted itself within the course of a single answer, and repeated harmful misinformation. The study, “Reliability Check: An Analysis of GPT-3’s Response to Sensitive Topics and Prompt Wording,” was published in Proceedings of the 3rd Workshop on Trustworthy Natural Language Processing.

Researchers at NASA’s Langley Research Center in Hampton, Virginia recently flew multiple drones beyond visual line of sight with no visual observer. The drones successfully flew around obstacles and each other during takeoff, along a planned route, and upon landing, all autonomously without a pilot controlling the flight. This test marks an important step towards advancing self-flying capabilities for air taxis.

“Flying the vehicles beyond visual line of sight, where neither the vehicle nor the airspace is monitored using direct human observation, demonstrates years of research into automation and safety systems, and required specific approval from the Federal Aviation Administration and NASA to complete,” said Lou Glaab, branch head for the aeronautics systems engineering branch at NASA Langley.

It is safer and more cost-effective to test self-flying technology meant for larger, passenger carrying air taxis on smaller drones to observe how they avoid each other and other obstacles.

Researchers develop an AI technique called Material Transformer Generator that integrates composition generation, structure prediction, and stability analysis to automatically design promising new two-dimensional materials.

Optomechanical coupling enables an on-chip frequency comb and optical time-lens for 70-fold optical pulse compression.

Carbon nanotubes have long tantalized researchers with their extraordinary mechanical and electronic properties. As one-dimensional nanostructures with remarkable mechanical strength and electrical conductivity, CNTs have been eyed for next-generation composites, energy storage devices, sensors and more. Yet realizing their promise has proven an enduring challenge.

CNTs have ultra-high surface energy and readily form large bundles rather than remaining as individualized tubes, severely compromising resultant material properties. Exfoliating these bundles, particularly in solution, has remained an immense difficulty despite intense R&D efforts over 30+ years employing covalent and noncovalent functionalization strategies.

Covalent approaches disrupt the CNTs’ pristine sp2 carbon networks, damaging their intrinsic properties. Noncovalent methods like surfactants and polymers have had limited success in debundling smaller diameter single-wall CNTs (SWCNTs), especially longer high aspect ratio tubes preferred for optimal conductivity and strength. And virtually all tactics have struggled to exfoliate specific SWCNT types, hindering enrichment in metallic SWCNTs boasting far higher conductance than their semiconducting counterparts.

A research team led by Professor Jong-min Choi of the Department of Energy Engineering has developed a technology that can significantly improve the efficiency of quantum dot photovoltaic cells by introducing organic solvent dispersible MXene.

The findings were published in Advanced Energy Materials (“Organic solvent dispersible MXene integrated colloidal quantum dot photovoltaics”).

Comparison of the dispersibility of quantum dot solar cell ink organic solvent according to surface modification of MXene. (Image: DGIST)