Cloud Hosting

You hear that? That's the sound of people everywhere saying 'they can do that?' after reading how researchers used imaging, computer vision, and machine learning to read ancient Roman text within what most would think amounts to little more than a stick of charcoal by looking at it.

The 2000-year-old scroll is one of more than 800 discovered in the ruins of Herculaneum, a Roman city buried by the eruption of Mount Vesuvius in AD 79. Carbonised by the eruption, each scroll is incredibly fragile, and any attempt to open one up to see if it's in any way legible has ended in a crumbling mess. Yet researchers have been keen to unlock the secrets stored in these scrolls, and it looks like machine learning might have proven the key to doing so.

Firstly, the scroll read by researchers was unwrapped, virtually. This complex process involves scanning the scroll, which took place at a particle accelerator near Oxford, UK. Then, these crumpled layers were unfurled into what amounts to flat layers of papyrusstill virtually, of course.

"The X-ray photos are turned into a 3D volume of voxels using tomographic reconstruction algorithms, resulting in a stack of slice images," the Vesuvius Challenge webpage says.

Then a step called Ink Detection was carried out on the unfurled layers, and this uses a machine learning model to identify the inked regions of the papyrus.

Various teams of researchers have then been trying to extract text from the scrolls as a part of something called the Vesuvius Challenge. Each team used various methods, models, and improvements to the previously developed methods to try and uncover more of what these scrolls are hiding. Each team then had until midnight on January 1, 2024 to submit their results to enter the challenge before a team of "eminent papyrologists" reviewed each entry to verify the results. The winning team won $700,000.

The winning teamcomprised of Youssef Nader, Luke Farritor, and Julian Schillingerentered a submission that was deemed by all the judges as being the most readable of the lot. Here's how they did it, according to a non-eminent, non-papyrologist (me).

The team's approach built on a previous discovery of a crackle pattern, which was discovered by Casey Handmer last year. This was later developed on by Luke Farritor, who used a GTX 1070 as a part of his more recent prize-winning efforts. Other teams of researchers, including another member of the winning team, Youssef Nader, had also built excellent machine learning models for detecting ink using fragments of the scrolls that had fallen off, though these only appeared to work well on fragments, not rolled up papyrus.

The winning team used the combined knowledge of these approaches to produce the clearest results with the rolled up scroll remains.

"The submission contains results from three different model architectures, each supporting the findings of the others, with the strongest images often coming from a TimeSformer-based model In addition to unparalleled ink detection, the winning submission contained the strongest auto-segmentation approach we have seen to date."

Around 5% of the first scroll has now been read as a result of all the work from all the teams of researchers working on this crumbly subject. And that's why the 2024 Vesuvius Challenge Grand Prize has now been announced, with the ambitious aim of going from the 5% now known up to 90% of all four scrolls that have been scanned.

So, there's definitely work to be done to understand more of it, but even today the researchers have some idea of what it says: "as too in the case of food, we do not right away believe things that are scarce to be absolutely more pleasant than those which are abundant."

Yep, it's all about the pleasures of life, or what's known as Epicureanism. Namely, for this section at least, the pleasures of stuffing your face.

See the article here:
Machine learning means we can read a 2000-year-old carbonized scroll buried by Mount Vesuvius and turns out it's all ... - PC Gamer

Photo by Emile Perron on Unsplash

Create a prediction pipeline using a Flask web app, project deployment in the AWS cloud using the CI-CD pipeline

Introduction

This part is a continuation of Part 2 where we did Data Ingestion, Data Transformation, Model Training, and Model Evaluation, Model Hyperparameter Tuning. In this part, we will develop a Flask web application, deploy the project in the AWS cloud using the CI-CD pipeline.

We need to know some Flask before proceeding with building the application. Also, create the templates folder in the main project structure.

Create 2 files inside the templates folder: home.html and index.html.

These 2 files are responsible for the frontend design of the application.

Create the predict_pipeline.py inside the pipeline folder that is inside the src folder.

Explain the app.py and predict_pipeline.py in detail.

Run app.py on the terminal (of course inside the conda environment)

Open a new window and run on the search bar.

Now add /predictdata (from app.py) to http://127.0.0.1:5000/

Looks like this

First, create the folder .ebextensions and create the file python.config inside the folder. This instance is created to deploy the project on Amazon Web Services (AWS) Elastic BeanTalk (EBS). Then create the python file application.py to deploy the project. Copy the contents of app.py and paste in application.py. Then commit into GitHub using the previous steps.

Better to use application.py instead of app.py to avoid dependency issues.

The contents of python.config

We are using the elasticbeanstalk AWS (to deploy the application) instance and the Python container.

For AWS deployment, we need an AWS account.

Go to the search bar and search Elastic Beanstalk

We will use a Codepipeline to integrate the GitHub we updated earlier with the AWS Elastic BeanStalk.

Then search for Codepipeline in the AWS search console for Continous Delivery.

Then click on Create pipeline

In the Add Source stage, use GitHub version 1 as the source, so it will integrate with your GitHub repository for this project

Next, skip the add build stage

For the add deploy stage, use AWS Beanstalk as the deploy provider, the other options are application name and environment name

The codepipeline will ultimately integrate with the created EBS application and the final deployment will be done.

The main objective of this blog was to understand the steps of deploying the project on Flask and AWS using CICD pipelines. This is the first time I am using AWS to deploy an end-to-end Machine Learning or Deep Learning project.

Hope you like the end-to-end series of learning and showcasing. I will keep doing end-to-end deployments this year and hope to make valueable projects.

GitHub Repository of the project

This marks the end of the blog.

Stay tuned for the next part and follow me here and say hi.

Twitter

References

Read more here:
End to End Machine Learning project implementation (Part 3) | by Abhinaba Banerjee | Feb, 2024 - DataDrivenInvestor

Kay Firth-Butterfield has worked on the intersection between accountability and AI for over a decade and is excited about the future. Im not an AI pessimist. I believe that if we get it right, it can open so many beneficial doors, she says. But shes still cautious. After doctors diagnosed her with breast cancer last year, she was grateful they did not rely too heavily on AI, though its increasingly used to evaluate mammograms and MRIs, and even in planning treatment. While Firth-Butterfield, who is now cured, worried less about whether a machine was reading her mammogram, she noted an over-reliance on current AI models can be problematic as sometimes they present incorrect information. Her surgeons agreed, she says.

A former judge and professor, Firth-Butterfield has emerged as one of the worlds leading experts on responsible AI, shaping efforts to ensure these systems remain accountable and transparent. Last April, she ended a five-and-a-half year stint as the head of AI and Machine Learning at the World Economic Forum, where she crafted frameworks and playbooks for companies, countries and other organizations to steer responsible development and use of AI. Her work advising the U.K. and Brazil on creating such AI systems made its way into law. If you're a government and you're using artificial intelligence with your citizens, then you have to be able to explain to your citizens how it is being used, she says. In 2016, Firth-Butterfield co-founded the Responsible AI Institute, which provides tools for organizations to build safe and reliable AI systems, and she serves on a council advising the U.S. Government Accountability Office on AI matters related to science and technology, and on an advisory board for UNESCOs International Research Centre on AI.

Nowadays, she also runs Good Tech Advisoryworking with corporations, governments, NGOs and media to implement AI responsibly. That means helping set up guidelines for the use of AI-reliant technology to minimize potential harm, while maximizing benefits and ensuring legal compliance.

As CEO of Good Tech Advisory, Firth-Butterfield has been helping hospitals in the U.S. navigate AIs potential uses, including for reading medical images and determining diagnoses. Many dont have clear guidelines about how staff can use programs like ChatGPT, even as Firth-Butterfield points out these tools can often provide inaccurate information. Those companies are wrestling with some really serious responsible AI choices, she says. Doctors using AI to efficiently type notes and handle administrative tasks can allow more time for patient care. But relying on AI to come up with a diagnosis in high pressure situations could be dangerous. And if a patient becomes sicker or dies, the question of who is liable becomes an issue.

When AI is not used responsibly, people can get hurtand its disproportionately women and people of color, Firth-Butterfield notes. Biased algorithms could prevent a worker from getting hired, unfairly reject mortgage applications or make incorrect decisions about security threats based on facial recognition, for example.

At the core of Firth-Butterfields advocacy is understanding how AI impacts the most vulnerable members of society. At the WEF, she worked with UNICEF to research the use of AI with children, and organized a Smart Toy Award that urged thoughtful implementation. We are allowing our children to play with toys that are enabled by artificial intelligence but we have no understanding of what our children are learningor where their data is going, she says.

Forbidding AI from being used in toys or classrooms as a way to protect children from its potential risks isn't the answer, says Firth-Butterfield. We do need children to be using AI in education because they're going to be using it in their work. So we have to find a responsible way of allowing that interaction between machine and human, she says. But teachers need to stay in charge. We cant just give education to AI; we need to keep humans in the loop, she says. Teachers might rely on AI for back-end administration, freeing up time to focus more on helping their students.

Its crucial to pay close attention to how the systems are constructed, but Firth-Butterfield is also concerned about who gets to participate. While more than 100 million people use ChatGPT, almost 3 billion people still lack access to the internet. We are increasing the digital divide at a huge ratenot just between the Global North and the Global South but also within countries, she says. Although AI has the potential to revolutionize teaching in schools and the treatment of medical patients, much of the world may not feel its effects. We tend to sit in our ivory towers talking about how AI is going to do everything brilliantly and we dont remember that much of the world hasnt been part of the internet revolution, she says.

Our future is at stake in these decisions about how people use and rely on AI, she says: Its about whether we as humans build the society that we want.

This profile is published as a part of TIMEs TIME100 Impact Awards initiative, which recognizes leaders from across the world who are driving change in their communities and industries. The next TIME100 Impact Awards ceremony will be held on Feb. 11 in Dubai.

View original post here:
Kay Firth-Butterfield On Harnessing AI's Power Responsibly - TIME

Cutting-edge technology is revolutionizing the field of mental health, as a recent study unveils a groundbreaking machine learning tool that can predict the onset of psychosis. By analyzing MRI brain scans, this innovative classifier can effectively differentiate between individuals who are at risk of developing psychosis and those who are not.

The study, conducted by an international consortium of researchers including experts from the University of Tokyo, examined over 2,000 participants from various global locations. Among the participants, approximately half had been identified as clinically high-risk individuals for psychosis. The classifier demonstrated impressive accuracy, correctly distinguishing between those who would later experience overt psychotic symptoms and those who would not. During the training phase, it achieved an accuracy rate of 85%, which reduced slightly to 73% when exposed to new data. The findings have been published in the esteemed journal Molecular Psychiatry.

This groundbreaking tool could prove invaluable in clinical settings, enabling early intervention in individuals at risk of psychosis. While psychosis may encompass delusions, hallucinations, and disorganized thinking, its causes are multifaceted and varied. Factors such as illness, injury, trauma, substance abuse, medication, and genetic predisposition can all contribute to its development. By identifying those at risk, clinicians can provide timely and targeted interventions, significantly improving outcomes and minimizing the negative impact on individuals lives.

Associate Professor Shinsuke Koike from the Graduate School of Arts and Sciences at the University of Tokyo emphasized the importance of this research. He highlighted that only about 30% of high-risk individuals eventually develop psychotic symptoms, leaving the remaining 70% uncertain of their fate. To better assist clinicians in their identification process, the integration of biological markers, alongside traditional symptom evaluations, becomes vital.

As the most common age for the first episode of psychosis occurs during adolescence or early adulthood, identifying young individuals in need of help can be particularly challenging. However, with the advent of this machine learning tool, healthcare professionals are empowered to proactively intervene and provide support to those most at risk. This marks a significant step forward in mental health research and treatment.

Source: Zhu et al./Molecular Psychiatry

An FAQ section:

Q: What is the groundbreaking machine learning tool mentioned in the article? A: The article discusses a machine learning tool that can predict the onset of psychosis by analyzing MRI brain scans.

Q: How accurate is the classifier in distinguishing between individuals at risk of developing psychosis and those who are not? A: During the training phase, the classifier achieved an accuracy rate of 85%. When tested with new data, it achieved a slightly reduced accuracy rate of 73%.

Q: What was the scope of the study conducted by the international consortium of researchers? A: The study examined over 2,000 participants from various global locations and focused on individuals who were clinically identified as high-risk for psychosis.

Q: How can this tool be valuable in clinical settings? A: The tool can enable early intervention in individuals at risk of psychosis, allowing clinicians to provide timely and targeted interventions to improve outcomes and minimize the negative impact on their lives.

Q: What are some factors that can contribute to the development of psychosis? A: Factors such as illness, injury, trauma, substance abuse, medication, and genetic predisposition can all contribute to the development of psychosis.

Q: What did Associate Professor Shinsuke Koike highlight about the research? A: Associate Professor Shinsuke Koike emphasized that only about 30% of high-risk individuals eventually develop psychotic symptoms, leaving the remaining 70% uncertain of their fate. He stressed the importance of integrating biological markers with traditional symptom evaluations to assist clinicians in the identification process.

Q: Why is identifying young individuals in need of help particularly challenging? A: The most common age for the first episode of psychosis occurs during adolescence or early adulthood, making it challenging to identify young individuals in need of help.

Definitions for key terms or jargon:

Psychosis: A mental health condition characterized by a loss of contact with reality, which may include delusions, hallucinations, and disorganized thinking. Machine learning: A field of artificial intelligence where computers learn and improve from experience without being explicitly programmed. MRI brain scans: Magnetic Resonance Imaging scans of the brain, which use magnetic fields and radio waves to produce detailed images of the brains structure and function. Classifier: In machine learning, a classifier is an algorithm that categorizes or assigns labels to input data based on patterns and features. Molecular Psychiatry: A peer-reviewed scientific journal that publishes research in the field of psychiatry, focusing on the molecular and genetic aspects of psychiatric disorders.

Suggested related links: University of Tokyo National Institute of Mental Health: Schizophrenia American Psychiatric Association

Excerpt from:
Predicting Psychosis: Unlocking the Power of Machine Learning - ai2.news