Make artificial intelligence valuable to science

Via
Madeleine Clarke


September 4, 2023
Read 5 minutes

There has been a lot of discussion about how AI will change the way we work.

Whether you consider AI in the workforce a friend or foe, there’s no doubt this technology will change some jobs as we know them. This includes jobs in the world of science. Our researchers are exploring and applying the responsible use of AI in their work.

AI is an especially useful tool for overcoming pesky biological limitations. For example, the human eye cannot distinguish objects smaller than about 0.2mm. Traditionally, we use microscopes to overcome this. But once an object or substance is magnified, a trained professional still needs to look at it to identify and count objects of interest. Counting specimens or objects for science can be a slow, laborious, and expensive process.

Here are some examples where we’ve leveraged the impressive advances in machine learning over the past decade to make AI valuable. These projects are driven by our Machine Learning and Artificial Intelligence Future Science Platform, which unites collaborators from across the organization to advance machine learning for scientific discovery.

Count the hairs on cotton leaves

Working with experts in the field of agriculture, we created a model that can quantify the number of hairs on the back of cotton leaves.

Leaf hairiness affects insect resistance, fiber yield and value of new cotton varieties. Traditionally, it is looked at by experts in commercial breeding programs and given a score from one to five.

First, we developed AI models that could generate human-like hairiness scores with 95% accuracy.

Dr. Moshiur Farazi is one of our experts in computer vision, focusing on helping computers identify and understand objects in images and videos. He said HairNet2 is now going beyond automating current methods.

“Training models to reproduce human ratings of hairiness can increase the speed and scale of analysis,” Moshiur said. However, these models reproduce variability in the estimates of the humans they were trained on.”

“In HairNet2, we have created a new AI-enhanced scoring method that is more powerful, reliable and accurate.

“This model estimates the area of ​​a leaf covered by hairs by determining the location of all the hairs on the leaf, something that is not completely impossible for humans but is extremely difficult and waste of time.

“HairNet2 was trained using about 1000 images where humans annotated each hair. This laborious annotation process helped build an AI tool that can automate hirsutism scoring beyond human cognition,” he said.

New models are being deployed on the web interface for breeders to test during the next cotton season. You can try the demo yourself soon.

Count microscopic algae cells

Harmful algal blooms are large populations of algae that can be toxic to both humans and animals.

To identify harmful algal blooms, experts conduct extensive testing using microscopes and counting chambers (a slide with a precise grid that allows scientists to estimate the number of cells). harmful algae in a liquid sample).

Dr. Chris Jackett is an expert in object detection. He began working to augment this manual process with AI as a postdoctoral fellow with our Australian National Research Collections.

“This is an extremely time-consuming and labor-intensive task, and humans are limited in the number of samples they can process,” says Chris.

“Prolonged microscopic observation can also lead to health problems such as vision problems, poor posture, physical stress and headaches.”

To meet this challenge, we are training machine learning models to automatically detect harmful algae in images.

Our team is systematically photographing and annotating algal strain samples from the Australian National Algal Culture Collection. We’re also starting to use more AI tools to help speed up the annotation process.

With this combined human-computer effort, we have so far built an annotated dataset for 15 different algal strains, which is currently being used to train AI models. Initial testing shows that these models can successfully detect target strains with high accuracy.

Using AI to detect toxic algae faster and more accurately could have significant economic, environmental and social impacts.

“Improving the speed and accuracy of harmful algae detection could provide an early warning signal to water managers about when and where algal blooms may occur,” said Chris. .

AI-enhanced decision-making and risk management can help protect the health of the environment as well as Australia’s coastal communities, consumers and fisheries and aquaculture businesses.

How you can make AI valuable for your business

Many organizations are currently grappling with the potential of AI to transform their processes and business operations.

If your core business involves counting or identifying objects, AI can be a friend.

Moshiur said the barrier to entry for those wanting to deploy AI for object detection is low.

“Five or 10 years ago, you needed to train the model yourself and needed more data and computing power to test AI,” Moshiur said.

“If you have a very small amount of data, you can fine-tune open source models to solve your problem with just a few hundred images.”

However, he said successful application of AI depends on asking the right questions.

“Ultimately, most users want a black box where they can click buttons and get the answer they want. But we need to unpack what they want the nodes to do and prepare the data in a way that allows those nodes to give the right answer,” Moshiur said.

“Taking the time to sit down and explore the questions you want answered, while also considering the problems you haven’t been able to solve with human-driven methods or processes, is the best place to start. head.”