Synthesised conversations and secure chatbots at the GenAI Academic Forum

• AI
• chatbots

Generative AI was front and centre at the recent Academic Forum, with a lens on where we are now and where we should be heading from a wide range of presenters. Faculties revealed how GenAI is currently being used in the classroom, the UTS Library explained how they have adapted to AI, Kylie Readman looked ahead to assessment opportunities in a landscape where GenAI will not be banned, and Jan McLean gave an overview of the UTS response to the TEQSA request for information. 

Below we summarise the new GenAI capabilities that were presented by Simon Buckingham Shum at the forum. You can also view the full presentation:

GenAI is increasingly multimodal and mobile 

The latest tools, such as ChatGPT4-0, are now multimodal. New functionality allows AI to look through the camera and interpret the visual scene while conversing with the student. This means it can function as a useful accessibility aid.  

We’ve moved a long way from texting a chatbot to speaking with it naturalistically and having it respond in a very realistic voice. The results are increasingly mobile, more efficient and located on devices that will soon be in every student’s pocket. 

Secure and customisable UTS chatbots  

While curious students will always be playing with new AI apps, we can’t require them to register with random AI vendor websites to conduct work considered vital to their studies. GenAI must become part of the secure UTS learning technologies ecosystem. In addition to providing secure Microsoft Copilot integrated with Bing search, the key development in the last quarter has been the design, testing and piloting of secure, customised chatbots for our students. Simon demo’d a couple of these (jump to video demo at 9:08).  

Make it a podcast with NotebookLM 

Remixing the TEQSA response 

While we consider our blog post is a pithy distillation of a dense TEQSA response, you can also address TL;DR by using something that has recently come from Google. You can now generate a synthesized conversation about your source material so it’s adapted into a podcast-style audio. This will be particularly useful for lengthy documents with a dense amount of information, like the TEQSA report.  

In this example, we have an upbeat ‘talk radio’ chat between a male and a female host, which gives us a conversational, accessible introduction. The synthesised voices are indistinguishable from humans (notwithstanding the occasional glitch), and the presenters joke, laugh, change tone and interact quite compellingly. This easy listening material could be extremely useful for a wide range of students depending on their preferred modes of engagement. 

Until now, the presenters have never challenged the content — they just describe it in mostly helpful accessible language. Just as with sycophantic chatbots trained to please, it requires explicit prompting that gives the AI permission to push back.  So, for this next version, I have added a prompt: I want the female presenter to argue persuasively for the need to change our assessment, while the male presenter needs to be curious but sceptical. So, we can now shape the conversation in whatever way we consider pedagogically useful. Here’s a snippet where the responses are more sceptical than the original version (jump to 1:18).

“But I’m sure there are some people who think that they’re being a tad, you know…” (her) , “Overly ambitious.” (him)

“And let’s be honest most academics I know are stretched pretty thin.” (her), “Tell me about it: grading, research, admin, it never ends.” (him), “Right — so where are they going to find the time to redesign entire courses?”  (her) 

Now, I was impressed with this. These are exactly the sorts of reasonable reactions that universities the world over are engaging with — after all, systemic transitions such as this are far from simple. The AI presenters are voicing precisely the doubts and worries that sit at the heart of our assessment crisis, and all this with a minimal prompt.  

Algenie unbottled 

The next example is from UTS start-up Algenie, who pioneer breakthrough technology to unlock the potential of algae. This conversation (jump to 6:30) was tuned for a fictional scenario where we want to help first-year undergraduate Biology students understand the startup. The prompt sets up the female presenter to ask questions encouraging students to connect what they’re learning in their courses. The male presenter welcomes the questions but lacks confidence, and suggests they discuss these ideas further in their next tutorial group.

The risk of disinformation 

The ability to generate easily digested and accessible information is exciting, as is the ability to tune the conversation. The unsettling flipside of this is that we now have the ability to generate completely biased dialogue with the superficial appearance of a ‘deep dive report’, which undermines analysis that others would consider authoritative (see the final example in this more detailed blog post). This is dual-use technology for sure – and looks like a gift to disinformation campaigns. 

Share

Twitter Facebook