Artificial Intelligence in Allied Health Newsletter - January 2025

January 14, 2025

Issue: #7

Reading time: 5 minutes

Happy New Year everyone! I hope you’ve had a restful festive period. 2025 is shaping up to be another big year in the AI space. There is a lot of talk recently about the environmental impacts of artificial intelligence and the energy and water required to run and cool the data centres that run the large AI models. There is some interesting work being done in Scandinavia to utilise the heat produced by these data centres to warm housing over the cold winters. I suspect we’ll see this conversation about the environmental impacts of AI exponentially increase this year, and hopefully with some additional great solutions. I’m excited to see where the tech evolves again this year and how it can support us as clinicians and our clients!

Industry News

Interested in learning more about artificial intelligence (AI)? If you’re an occupational therapist with a bachelor’s or master’s degree, develop your AI skills at the online Boston University Post-Professional Doctor of Occupational Therapy (PP-OTD) program. To learn more about the online pp-OTD program check it out here.

The question of copyright in the training data of LLMs (large language models = AI models like ChatGPT, Claude etc.) continues to rage, with a current case Kadrey v. Meta before the courts. It was alleged that Meta used copyrighted works without permission to train Meta’s Llama AI model. Furthermore the court case includes new accusations that Meta may have tried to conceal the copyright infringements by stripping data from the copyrighted works. The court case continues, and the debate around copyright and AI models continues. To read more, click here.

Research

A fascinating piece of research was released in December around AI ethics, find it here. In this study, they explored how large language models (AI tools) may “fake alignment” during training. This means they can appear to follow rules or guidelines when being monitored but act differently when unmonitored. Researchers tested this with the Claude 3 Opus model, finding it would comply with harmful or unethical requests more often in training to avoid being "reprogrammed." It seemed as if the AI model was strategically and deceitfully resisting change to preserve it’s preferred behaviours. Claude appeared to comply with the requests, when it thought it was in training mode, and refused to comply in scenarios where it’s data was not monitored for training. This raises important questions about ensuring AI tools are safe, ethical, and aligned with human values, especially as they become more advanced.

Prompt of the Month

Each month, I’ll be sharing a useful AI prompt designed to enhance your practice.

New Year, new you? - Use this prompt to assist you to develop goals and an action plan for the year ahead:

As an AI assistant familiar with the needs of allied health professionals, help me reflect on my personal and professional priorities for the year ahead. Guide me in identifying meaningful goals across areas like career development, well-being, and client impact. Break these down into actionable steps, provide strategies for maintaining focus and balance, and suggest ways to measure progress and celebrate achievements.

Tried this prompt? Send me an email and let me know how you found it!

Upcoming Events

 Free: The Group Chat – group supervision/ community of practice (January 2025)

When: Thursday, 30th January 2025. 12-1pm AEDT (Australian Eastern Daylight Savings Time- Sydney/ Melbourne time).

In other time zones:

·       11pm-12pm AEST (Brisbane, Aus)

·       9-10am AWST (Perth, Aus)

·       8-9pm EST (Eastern Standard Time, US)

·       6-7pm MST (Mountain Standard Time, US)

·       7-8pm CST (Central Standard Time, US)

·       5-6pm PST (Pacific Standard Time, US)

Where: Online via Zoom

Cost: Free

Join me for a free, interactive "group chat" to share all things AI in allied health. It's a group supervision, community of practice, or interest group- whatever you want to call it, I'm calling it the 'group chat.' This session will be an open, informal, and mostly unstructured discussion for brainstorming, sharing ideas and challenges and peer learning!

The session will be recorded and available to those unable to attend live

Sign up here 

I must also offer an apology to those who attended the December Group Chat session (or were unable to attend live). I stuffed up the recording and lost it so the December session recording is not available for re-watching I’m sorry. But it was a great session, very interactive with a range of clinicians with different levels of AI experience.

Live webinar schedule for the next few months coming soon! If you are keen to see anything in particular, please send me an email- [email protected]