Our smart devices take voice commands from us, check our heartbeats, track our sleep, translate text, send us reminders, capture photos and movies, and let us talk to family and friends continents away.

Now imagine turbocharging those capabilities. Holding in-depth, natural language exchanges on academic or personal queries; running our vital signs through a global database to check on imminent health issues; packing massive databases to provide comprehensive real-time translation among two or more parties speaking different languages; and conversing with GPS software providing details on the best burgers, movies, hotels or people-watching spots trending along your route.

Tapping into the seductive power of large language models and natural language processing, we’ve witnessed tremendous progress in communications between us and technology that we increasingly rely on in our daily lives.

But there’s been a stumbling block when it comes to AI and our portable devices. Researchers at Apple say they are ready to do something about it.

The issue is memory. Large language models need lots of it. With models demanding storage of potentially hundreds of billions of parameters, commonly used smartphones such as Apple’s iPhone 15 with a scant 8GB of memory will fall far short of the task.

In a paper uploaded to the pre-print server arXiv on Dec. 12, Apple announced it had developed a method that utilizes transfers of data between flash memory and DRAM that will allow a smart device to run a powerful AI system.

The researchers say their process can run AI programs twice the size of a device’s DRAM capacity and speed up CPU operations by up to 500%. GPU processes, they say, can be sped up to 25 times current approaches.

“Our method involves constructing an inference cost model that harmonizes with the flash memory behavior, guiding us to optimize in two critical areas: reducing the volume of data transferred from flash and reading data in larger, more contiguous chunks,” the researchers said in their paper titled, “LLM in a flash: Efficient Large Language Model Inference with Limited Memory.”

The two techniques they used were:

1. Windowing, which slashes the amount of data that needs to be exchanged between flash memory and RAM. This is accomplished by reusing results from recent calculations, minimizing IO requests and saving energy and time.
2. Row column bundling, which achieves greater efficiency by digesting larger chunks of data at a time from flash memory.

The two processes, say the researchers, “collectively contribute to a significant reduction in the data load and an increase in the efficiency of memory usage.”

They added, “This breakthrough is particularly crucial for deploying advanced LLMs in resource-limited environments, thereby expanding their applicability and accessibility.”

In another recent breakthrough, Apple announced that it had designed a program called HUGS that can create animated avatars from just a few seconds worth of video captured from a single lens. Current avatar creation programs require multiple camera views. The report, “HUGS: Human Gaussian Splats,” was uploaded to arXiv Nov. 29.

Their program can create realistic dancing avatars in as little as 30 minutes, far shorter than the two days required for current popular approaches, according to Apple.

© 2023 Science X Network

In recent years, engineers have developed a wide range of robotic systems that could soon assist humans with various everyday tasks. Rather than assisting with chores or other manual jobs, some of these robots could merely act as companions, helping older adults or individuals with different disabilities to practice skills that typically entail interacting with another human.

Researchers at Nara Institute of Science and Technology in Japan recently developed a new robot that can play video games with a human user. This robot, introduced in a paper presented at the 11th International Conference on Human-Agent Interaction, can play games with users while communicating with them.

“We have been developing robots that can chat while watching TV together, and interaction technology that creates empathy, in order to realize a partner robot that can live together with people in their daily life,” Masayuki Kanbara, one of the researchers who carried out the study, told Tech Xplore. “In this paper, we developed a robot that plays TV games together to provide opportunities for people to interact with the robot in their daily lives.”

The new robotic system created by Kanbara and his colleagues is based on a dialogue system that supports smooth conversations with human users. This system processes what users say while playing games with the robot and produces suitable responses.

The researchers also developed a system that adapts the content of the game that a human user and robot and playing, for instance by changing the game characters and what they are doing. As the game progresses, the robot can say things that are aligned with what is happening at the time, while retaining a friendly tone.

“The proposed robot dialogues with a human while playing a competitive TV game together to realize a continuous interaction between the human and the robot in daily life,” Kanbara explained. “The robot is happy when it wins and frustrated when it loses.”

To assess the potential of their proposed robotic system, Kanbara and his colleagues tested it in an experiment involving 30 human participants. These participants were asked to play the same videogame alone and in conjunction with the team’s robot, then share their feedback in a simple questionnaire.

In this questionnaire, most participants said that they found playing with the robot more enjoyable than playing alone. This highlights the promise of the team’s companion robot for entertaining users and improving their gaming experience.

“In recent years, some have been exploring the potential of TV games as tools to help the elderly to practice their mental skills and remain sharp,” Kanbara said. “Therefore, our robot can be used as one of the functions to continue to use partner robots that watch over and support the lives of the elderly.”

The robot developed by this team of researchers could soon be improved further and tested in additional experimental trials, to further explore its potential. In addition, this recent study could inspire the development of other interactive, instructive and robot-assisted gaming experiences targeted toward specific segments of the population, such as older adults or children diagnosed with autism spectrum disorder (ASD).

“The evaluation in this paper showed the effectiveness of the proposed robot during a short game time of about 15 minutes,” Kanbara added. “In the future, we aim to design a robot behavior that continues to motivate interaction over a long period of time, in terms of months or years.”

© 2024 Science X Network