2 November 2023
While it’s true that students have been utilising AI to generate answers for assignments and even complete entire essays, the capabilities of generative AI tools go beyond that.
Just under nine months after its initial introduction, OpenAI, the creators of ChatGPT, unveiled new voice and image capabilities for the tool, describing it as “a new, more intuitive interface allowing voice conversations and image-based interactions.” This is on top of OpenAI’s other imagery tool, DALL·E.
They are not the only ones rising in popularity in AI-generated imagery. Established tech companies like Adobe, have introduced generative AI models, Adobe Firefly, geared towards the field of design and creative production.
These innovations in text-to-image and image-to-text capabilities bring a fresh perspective to teaching and learning. While they open new doors, they also give rise to important concerns.
What would you get when you feed ChatGPT an image?
According to OpenAI, it’s as simple as taking a picture of a problem and asking how to solve it. Powered by multimodal models like GPT-3.5 and GPT-4, these AI systems can be applied to a variety of visual content, including photographs, screenshots, and documents containing text and images.
This seemingly simple capability could potentially shake up the educational landscape. Could students access all the answers to a homework assignment with a single screenshot? Such possibilities raise new challenges for educators striving to innovate assessments while keeping pace with the widespread use of generative AI by learners.
Unlocking significant gains in creative productivity
The text-to-image potential has also transformed the development of creativity skills, allowing learners and educators alike to translate and refine ideas into images.
Lori Landay, a cultural studies professor at Berklee highlighted the “immediate potential for AI to help students close the gap between imagination and output – for example, how they can turn a sketch into an illustration using image generation AIs, even if they lack visual art skills.”
Bron Stuckey squashed fears that there’s little skill to be learned if learners were to turn to tools to produce, explaining “They’ve got to be able to critique it, to make it their own, to refine it”. She also celebrates the fact that text-to-image tools are going to “democratise game design”, individuals who can code but cannot do graphics are not left at a disadvantage because they lack illustration skills.
Yet, as AI advances and AI-generated content becomes more convincing, we must grapple with the question of whether we’re entering a world where it’s challenging to distinguish between “human-created” and “AI-generated” work. Landay expresses reservations about the authenticity of human expression when machines can convincingly impersonate people in various forms, from deepfake videos to visual art and writing.
OpenAI is well aware of these challenges and is taking steps to address them. They limit the tool’s ability to make direct statements about people and emphasize transparency regarding the tool’s limitations due to its nascent stage.
Like any other tool, generative AI requires clear guidelines and skills for responsible use. To counteract the widespread adoption of these tools, it’s crucial to reinforce integrity in academic submissions and promote healthy digital habits when using technology.
Generative AI holds immense potential in education for creativity skills development and content generation. However, its widespread use also raises concerns about authenticity and misuse. To strike a balance, we must focus on responsible AI usage in education, developing ethical guidelines and fostering transparency. This approach will ensure that AI becomes a valuable educational tool, enhancing learning experiences without compromising academic integrity.
How to navigate 4 key generative AI issues in K-12 schools
Organisation founded to create guidelines for the use of AI image generators
Pedagogical Affordances of ChatGPT-4v(ision)
Balancing promise and protection: AI safeguarding in education