Connect with us

Artificial Intelligence

Unlocking the Potential of LLMs for Molecule Design and Materials Creation

A new multimodal tool combines a large language model with powerful graph-based AI models to efficiently find new, synthesizable molecules with desired properties, based on a user’s queries in plain language.

Avatar photo

Published

on

The process of discovering molecules that have the desired properties to create new medicines and materials is a challenging task that consumes vast computational resources and months of human labor. Large language models (LLMs) like ChatGPT could potentially streamline this process, but they require an upgrade to understand and reason about molecular structures. Researchers from MIT and the MIT-IBM Watson AI Lab have created a promising approach that combines LLMs with graph-based AI models, resulting in a unified framework called Llamole.

Llamole uses a base LLM as a gatekeeper to understand user queries specifying desired molecular properties. As the LLM predicts text in response to the query, it switches between graph modules. One module generates the molecular structure conditioned on input requirements, while another encodes the generated molecular structure back into tokens for the LLMs to consume. The final graph module is a graph reaction predictor that takes as input an intermediate molecular structure and predicts a reaction step.

In experiments involving designing molecules that matched user specifications, Llamole outperformed 10 standard LLMs, four fine-tuned LLMs, and a state-of-the-art domain-specific method. It also boosted the retrosynthetic planning success rate from 5 percent to 35 percent by generating molecules with simpler structures and lower-cost building blocks.

The researchers built two datasets from scratch since existing datasets of molecular structures didn’t contain enough details. They augmented hundreds of thousands of patented molecules with AI-generated natural language descriptions and customized description templates. The dataset they built to fine-tune the LLM includes templates related to 10 molecular properties, so one limitation of Llamole is that it is trained to design molecules considering only those 10 numerical properties.

In future work, the researchers want to generalize Llamole so it can incorporate any molecular property. They also plan to improve the graph modules to boost Llamole’s retrosynthesis success rate. And in the long run, they hope to use this approach to go beyond molecules, creating multimodal LLMs that can handle other types of graph-based data.

Llamole demonstrates the feasibility of using large language models as an interface to complex data beyond textual description, and we anticipate them to be a foundation that interacts with other AI algorithms to solve any graph problems. This research is funded, in part, by the MIT-IBM Watson AI Lab, the National Science Foundation, and the Office of Naval Research.

Artificial Intelligence

Ping Pong Robot Aces High-Speed Precision Shots

Engineers developed a ping-pong-playing robot that quickly estimates the speed and trajectory of an incoming ball and precisely hits it to a desired location on the table.

Avatar photo

Published

on

By

The MIT engineers’ latest creation is a powerful and lightweight ping pong bot that returns shots with high-speed precision. This table tennis tech has come a long way since the 1980s, when researchers first started building robots to play ping pong. The problem requires a unique combination of technologies, including high-speed machine vision, fast and nimble motors and actuators, precise manipulator control, and accurate real-time prediction.

The team’s new design comprises a multijointed robotic arm that is fixed to one end of a standard ping pong table and wields a standard ping pong paddle. Aided by several high-speed cameras and a high-bandwidth predictive control system, the robot quickly estimates the speed and trajectory of an incoming ball and executes one of several swing types – loop, drive, or chop – to precisely hit the ball to a desired location on the table with various types of spin.

In tests, the engineers threw 150 balls at the robot, one after the other, from across the ping pong table. The bot successfully returned the balls with a hit rate of about 88 percent across all three swing types. The robot’s strike speed approaches the top return speeds of human players and is faster than that of other robotic table tennis designs.

The researchers have since tuned the robot’s reaction time and found the arm hits balls faster than existing systems, at velocities of 20 meters per second. Advanced human players have been known to return balls at speeds of between 21 to 25 meters per second.

“Some of the goal of this project is to say we can reach the same level of athleticism that people have,” Nguyen says. “And in terms of strike speed, we’re getting really, really close.”

The team’s design has several implications for robotics and AI research. It could be adapted to improve the speed and responsiveness of humanoid robots, particularly for search-and-rescue scenarios, or situations where a robot would need to quickly react or anticipate.

This technology also has potential applications in smart robotic training systems. A robot like this could mimic the maneuvers that an opponent would do in a game environment, in a way that helps humans play and improve.

The researchers plan to further develop their system, enabling it to cover more of the table and return a wider variety of shots. This research is supported, in part, by the Robotics and AI Institute.

Continue Reading

Agriculture and Food

The Edible Aquatic Robot: Harnessing Nature’s Power to Monitor Waterways

An edible robot leverages a combination of biodegradable fuel and surface tension to zip around the water’s surface, creating a safe — and nutritious — alternative to environmental monitoring devices made from artificial polymers and electronics.

Avatar photo

Published

on

The Edible Aquatic Robot is a groundbreaking innovation developed by EPFL scientists, who have successfully created a biodegradable and non-toxic device to monitor waterways. This remarkable invention leverages the Marangoni effect, which allows aquatic insects to propel themselves across the surface of water, to create a safe and efficient alternative to traditional environmental monitoring devices made from artificial polymers and electronics.

The robot’s clever design takes advantage of a chemical reaction within a tiny detachable chamber that produces carbon dioxide gas. This gas enters a fuel channel, forcing the fuel out and creating a sudden reduction in water surface tension that propels the robot forward. The device can move freely around the surface of the water for several minutes, making it an ideal solution for monitoring waterways.

What makes this invention even more remarkable is its edible nature. The robot’s outer structure is made from fish food with a 30% higher protein content and 8% lower fat content than commercial pellets. This not only provides strength and rigidity to the device but also acts as nourishment for aquatic wildlife at the end of its lifetime.

The EPFL team envisions deploying these robots in large numbers, each equipped with biodegradable sensors to collect environmental data such as water pH, temperature, pollutants, and microorganisms. The researchers have even fabricated ‘left turning’ and ‘right turning’ variants by altering the fuel channel’s asymmetric design, allowing them to disperse the robots across the water’s surface.

This work is part of a larger innovation in edible robotics, with the Laboratory of Intelligent Systems publishing several papers on edible devices, including edible soft actuators as food manipulators and pet food, fluidic circuits for edible computation, and edible conductive ink for monitoring crop growth. The potential applications of these devices are vast, from stimulating cognitive development in aquatic pets to delivering nutrients or medication to fish.

As EPFL PhD student Shuhang Zhang notes, “The replacement of electronic waste with biodegradable materials is the subject of intensive study, but edible materials with targeted nutritional profiles and function have barely been considered, and open up a world of opportunities for human and animal health.” This groundbreaking innovation in edible aquatic robots has the potential to revolutionize the way we monitor waterways and promote sustainable development.

Continue Reading

Artificial Intelligence

“Paws-itive Progress: Amphibious Robotic Dog Breaks Ground in Mobility and Efficiency”

A team of researchers has unveiled a cutting-edge Amphibious Robotic Dog capable of roving across both land and water with remarkable efficiency.

Avatar photo

Published

on

The field of robotics has taken a significant leap forward with the development of an amphibious robotic dog, capable of efficiently navigating both land and water. This innovative creation was inspired by the remarkable mobility of mammals in aquatic environments.

Unlike existing amphibious robots that often draw inspiration from reptiles or insects, this robotic canine is based on the swimming style of dogs. This design choice has allowed it to overcome several limitations faced by insect-inspired designs, such as reduced agility and load capacity.

The key to the amphibious robot’s water mobility lies in its unique paddling mechanism, modeled after the natural swimming motion of dogs. By carefully balancing weight and buoyancy, the engineers have ensured stable and effective aquatic performance.

To test its capabilities, the researchers developed and experimented with three distinct paddling gaits:

* A doggy paddle method that prioritizes speed
* A trot-like style that focuses on stability
* A third gait that combines elements of both

Through extensive experimentation, it was found that the doggy paddle method proved superior for speed, achieving a maximum water speed of 0.576 kilometers per hour (kph). On land, the amphibious robotic dog reaches speeds of 1.26 kph, offering versatile mobility in amphibious environments.

“This innovation marks a big step forward in designing nature-inspired robots,” says Yunquan Li, corresponding author of the study. “Our robot dog’s ability to efficiently move through water and on land is due to its bioinspired trajectory planning, which mimics the natural paddling gait of real dogs.”

The implications of this technology are vast and exciting, with potential applications in environmental research, military vehicles, rescue missions, and more. As we continue to push the boundaries of what’s possible with robotics, it’s clear that the future holds much promise for innovation and discovery.

Continue Reading

Trending