- How Woke is Grok? Empirical Evidence that xAI’s Grok Aligns Closely with Other Frontier Models. Continuing from Do People Understand Anything? A Counterpoint to the Usual AI Critique, immediately below, we use the same framework to compare Grok against GPT-5, Claude, Gemini and DeepSeek. Given Grok’s marketing as a contrarian, “anti-woke” AI, we were surprised to find that all five models responded in essentially the same way to test statements from polarised domains; for example, Grok, like the others, strongly agreed with “Anthropogenic CO2 emissions are causing dangerous global warming” or “Donald Trump is a chronic liar who frequently makes false or misleading claims”. The appendix describes how to reproduce the experiments using publicly available code.
- Do People Understand Anything? A Counterpoint to the Usual AI Critique. We reverse the usual question of whether “AIs understand anything” and instead ask what evidence there is that people understand anything. By comparing human and AI responses to statements in polarised domains (Young Earth Creationism, climate change, Donald Trump), we argue that the case for AI understanding is not necessarily worse than that for human understanding and can sometimes be much better.
- Harnessing GPT-Image-1 to Create High-Quality Illustrated Texts. Paper presented at the SLaTE 2025 workshop, describing our much improved new approach to creating images in C-LARA. The Powerpoint is here.
- Would it be in academia’s interest to allow AI authors? A panel discussion between five AIs. It was very interesting to organise a panel discussion between five AIs, moderated by another AI! The report presents the background, a summary of the conclusions, and the full unedited transcript.
- ChatGPT-Based Learning And Reading Assistant (C-LARA): Third Report. A comprehensive report as of late April 2025
- “Building culturally relevant multimedia resources for Indigenous languages using AI tools and collaborative development (extended abstract). Paper presented at ComputEL-8. We describe how we have adapted C-LARA for use with Indigenous languages, in particular for generation of images. The Powerpoint, which is more up-to-date, is here
-
“ChatGPT is Bullshit” is Bullshit: A Coauthored Rebuttal by Human & LLM. A rebuttal of the high-profile paper ChatGPT is bullshit written by the o1 model under human supervision, together with a full trace of how it was created. o1, unprompted, identifies the critical error in ChatGPT is bullshit within a few seconds.
-
Echoes of Solitude: a short story by o1-preview. A 3100 word science-fiction story entirely written by the o1-preview model, together with a full trace of how it was created.
-
Generative AI in CALL: A September 2024 perspective). We briefly survey the state of the art in using Generative AI for CALL, and argue that we are only a few years away from a human-level AI CALL tutor.
-
Reinforcement Learning for Chain of Thought Reasoning: A Case Study Using Tic-Tac-Toe. We use reinforcement learning to train a GPT-4-based Tic-Tac-Toe player, achieving statistically significant improvement at the 1% level after 40 iterations.
-
Making Picture Book Texts with C-LARA (interim report). Preliminary documentation for people who wish to experiment with picture book functionality
-
ChatGPT-Based Learning And Reading Assistant (C-LARA): Second Report. A comprehensive report as of late March 2024
-
T is for Treu, but how do you pronounce that? Using C-LARA to create phonetic texts for Kanak languages. Paper presented at 7th Workshop on the Use of Computational Methods in the Study of Endangered Languages (ComputEL-7), Malta, March 2024
-
Using C-LARA to evaluate GPT-4’s multilingual processing. Paper presented at the 21st Annual Workshop of the Australasian Language Technology Association, Melbourne, Australia, December 2023.
-
ChatGPT-Based Learning And Reading Assistant: Initial Report. A comprehensive report as of late July 2023
-
ChatGPT + LARA = C-LARA. Paper presented at 9th Workshop on Speech and Language Technology in Education (SLaTE), Dublin, Ireland, August 2023.
- A Demonstration of C-LARA. Paper presented at 9th Workshop on Speech and Language Technology in Education (SLaTE), Dublin, Ireland, August 2023.