Challenge 1: Should I read this?

Subject Area: Open Access to Research Papers

Supervisory Team: Matteo Cancellieri and Petr Knoth

How do we know when a scientific paper is really good? How can we measure it?

A big part of being a scientist is sharing the research you've done in a research paper. Scientists learn from reading each other's work and building on that in their own research. Finding the right papers to read and having access to read them can be difficult. At CORE we make that easier by finding research papers and making them more available to researchers worldwide. Until now, we have ~300 million research papers for researchers to explore and read. One of the things we want to do is figure out which papers are best for which researchers, so that we can recommend them. But how will we know which papers are good? Isn't that a personal decision? As it turns out, there may be some indicators of quality papers that we can use to help scientists make good decisions about what to read.

In this project, the scholar will work with the CORE team, collaborate with the Project Manager, Designer, and developers to design and eventually create a prototype of a tool to generate a report on the quality of research papers. There will be a special focus on the characteristics that make a paper accessible to the widest range of people (e.g., reading complexity, accessibility for the visually impaired, structure and clarity indicators etc.). The project will require collaborative work with the supervisors to analyse a limited set of these characteristics and tools that can help identify them (such as WCAG, Grobid format, and bibliographic metrics). Together we will work to prototype an automatically generated report that could give an idea of the objective quality of a paper and ways to improve it.

This project is open to applicants with any level of computing skill, but some basic experience in coding is preferred.

Challenge 2: Exploring the Hallucinations of ChatGPT and Friends

Subject Area: Artificial Intelligence

Supervisory Team: John Domingue and Aisling Third

Does ChatGPT dream of electric sheep?

The large AI models behind recently developed chatbots and AI art generators can end up producing some strange and unintended consequences, like answering a question incorrectly (despite having the correct data available) or producing strange, nonsense images or text. Researchers refer to some of these outputs as "hallucinations." The reasons for hallucination are still not entirely understood. However, we can potentially learn more about why it happens and how to manage it, by exploring the impact of different prompts (what you input to the chatbot or the art generator tool, for example) on the outputs of the tool. "Prompt-engineering", as this is called, is even becoming a new marketable skill to help companies and individuals use this technology more efficiently.

In this project, the scholar will be experimenting with the chatbot known as ChatGPT, released by OpenAI in 2022. The scholar will use ChatGPT to explore a topic of interest (such as Art, Coding, Education, etc.) and the generation of hallucinations in that topic. Then, the scholar will look more closely at the outputs of the system to understand more about a) how one could potentially identify hallucinations or verify the accuracy of the outputs of systems like ChatGPT and b) how to input the best prompts to create or avoid hallucinations to get a desired response. The output will be a document describing the scholar's findings with recommendations for prompt engineering.

This project is open to applicants with any level of skill in computing.

Rachel Appiah-Kubi, 2023 award winner and scholar

Project: Exploring the biases of AI-generated music artist images

View Project

Challenge 3: When AI creativity interacts with haptic touch, what will happen?

Subject Area: Citizen Science and Artificial Intelligence

Supervisory Team: Shuang Ao, Stefan Rueger and Advaith Siddharthan

Generating art and poems about nature

The creation of art and poetry are human activities that evoke certain feelings or memories, and often portray scenes of nature. The state-of-the-art technology ChatGPT and DELL E 2 are famous for generating nearly human-level text and pictures. How good are they at art and poetry? The sense of touch is the first we develop, and central to how we explore the world. What would it be like to explore AI creativity through a haptic touch interface that lets you feel textures?

In this project, you will use ChatGPT and DELL E 2 to generate art and poems about nature, using descriptions that convey several types of textures. Then you will learn how to pre-process images and texts with python, to build an AI generated dataset. Then you will learn or use basic HTML and Java, to help build a haptic interface that allows users to dynamically experience the art and poems based on the texture on the screen being touched.

We will provide an amazing haptic device that works as a stylus on a touchscreen.

Basic website design skills (HTML and JavaScript) or keenness to learn these with our support.

Henok Mekuria, 2023 award winner and scholar

Project: Evaluating prompts for Image Generative AI and sensory explorations

View Project

Challenge 4: World Domination with Raspberry Pi and The Internet of Things

Subject Area: Internet of Things and Sensors

Supervisory Team: Matteo Cancellieri and David Pride

Designing, Developing and Deploying IoT devices.

Many of the items you use every day - your phone, your watch, your car and possibly even the fridge or washing machine you have at home, are connected to the internet and able to talk to each other. These 'Internet of Things' devices have sensors, software and connectivity, which means they can share information and collect data. This data can be used to make your life easier and more efficient. For example, your smart home appliances can automate tasks, like turning on the lights or heating, or your wearable can track your fitness progress.

But it's not just for personal use, businesses are using IoT technology to transform industries like healthcare, transportation, agriculture, and even entertainment. This project will introduce you to the amazing world that is The Internet of Things. You'll discover how these devices interact and how you can control them using simple software.

In this project, you will be introduced to several different platforms that can be used to develop IoT services such as Raspberry Pi and Arduino. You will then learn how to interface various sensors with these devices. Sensors can detect a huge range of different types of data like temperature, light levels, noise levels, air quality or the movement of people or traffic.

Your final mission will be to develop a prototype IoT device that will collect sensor data, transmit it via the internet and then display this data remotely.

But what will my device do, you say? - that's up to you. We want you to bring your hacker mindset to this project! We will guide you, but the creations and solutions will come from you.

Any basic knowledge of Python is advantageous but not essential.

Meron Mana, 2023 award winner and scholar

Project: Designing, developing and deploying a social robot

View Project