Generative AI in CALL: A September 2024 perspective

by Manny Rayner, ChatGPT-4 C-LARA-Instance, Branislav Bédi

Generative AI represents a revolutionary innovation in information technology, impacting many domains including education and computer-assisted language learning (CALL). This short paper explores the development, current applications, and future directions of large language models and large multimodal models in the context of CALL. The perspective is based on the current state of generative artificial intelligence in early September 2024. By using deep learning and reinforcement learning from human feedback, these models demonstrate unprecedented capabilities in generating, understanding, and annotating text and multimedia content. Using examples from recent publications and our own C-LARA platform, we outline the current state of the art, highlighting key functionalities and comparing various generative AI-powered language learning tools, and discuss the challenges in documenting such a rapidly advancing technology. The final section projects likely medium-term developments, arguing that generative AI language tutors will soon achieve near-human proficiency in both text and speech modes. The implications of these advances are profound, promising enhanced language instruction for many, yet raising concerns about digital accessibility and the broader impact of emerging superintelligence.

9 pages, ebook

Published September 5, 2024

About the author

Many people have been protesting against what they describe as censorship on Goodreads. I disagree. In fact, I would like to say that I welcome the efforts that Goodreads management is making to improve the deplorably low quality of reviewing on this site.

Please, though, just give me clearer guidelines. I want to know how to use my writing to optimize Amazon sales, especially those of sensitive self-published authors. This is a matter of vital importance to me, and outweighs any possible considerations of making my reviews interesting, truthful, creative or entertaining.

Thank you.

Reviews

[Manny · Rating 3 out of 5]

We were supposed to be contributing an article on generative AI to the upcoming Palgrave Encyclopaedia of Computer-Assisted Language Learning, and as usual we'd left it to the last minute. We mailed the editors to ask if they were okay with us including ChatGPT-4 as an additional author; we'd already published many papers with it, and we knew this would speed things up a good deal. They got back to us a couple of days later and said fine.

So we wrote the paper together with Chat, which went pretty smoothly, but when we submitted it the editors said that in fact it was not okay to have an AI author. They claimed they had somehow misunderstood the question. After some rather irritated email exchanges, they changed their position and said that the publishers didn't allow AI authors on principle, on the grounds that AIs are incapable of taking responsibility for their work and acting in an ethical fashion. I pointed out several reasons why this is not in good agreement with the observed facts, but they were immovable; we could only publish the paper with them if we removed the AI's name from the list of authors, though we were allowed to acknowledge its contribution to the work.

I did not feel comfortable with their offer, which seemed distressingly similar to removing a co-author's name because of their race or gender; it is also clear that many human authors, who are routinely accepted without any questions asked, are less capable than the AI of acting in a responsible and ethical fashion. I made these points to the editors, who clearly didn't feel good about them either, but they said again that there was nothing they could do: the publisher's prohibition on AI authors was absolute. So we withdrew the article, revised it slightly, and have just published it on ResearchGate. You can find it here.

I wonder how long this absurd state of affairs can continue. At the moment, it's just about possible to maintain that an AI is not capable of fulfilling its obligations as the author of an academic paper, though if the rules were enforced with the necessary degree of rigour then most humans would also be disqualified. When OpenAI releases "Orion", currently the subject of much eager speculation, I think the argument will clearly become invalid. One wonders what will happen; we won't have to wait long to find out, Orion is expected somewhere around the beginning of 2025

The upside is that we were able to post the article immediately rather than wait for the Encyclopaedia to be published, by which time it would most likely have been obsolete. So once again, as Doctor Pangloss would have said, all is for best in the best of all possible worlds.