Using ChatGPT to help us understand tinnitus. Thoughts? Here's an AI Assistant I've created for understanding Tinnitus

[daverussell]

Just a topic for debate but I'm interested in your thoughts and maybe your direct experiences with AI to understand tinnitus.

I'm assuming everyone has heard about AI and ChatGPT and other Generative AI tools. GPT is short for Generative Predictive Text. It is what it states (a convincing) prediction text engine. As a teacher it's a hot topic of debate and I'm in favour of it for productivity. However, it should only be a tool and not a substitute or complete solution.. in fact I've written my own lesson planning tool that interfaces with ChatGPT. Anyway I digress!!!

A GPT is a ChatGPT feature, which in short is a predefined conversation that has been primed to perform as an assistant. I've asked this GPT to help me understand tinnitus. I'm playing around and asking it for advice as we speak - chat.openai.com/g/g-p70660x....

There are benefits to it as an on demand service. However, I'm sure we'd agree that AI cannot and should NOT or EVER replace human therapist as experts in their field, or be used as a diagnose tool (no more than Google)

It may deter people who desperately need professional help from seeking help. AI is also renowned for inaccuracies - it predicts it's responses based on large datasets, rather than having expertise.

I'd be interested to hear your thoughts. As I teach my students, use it as a tool and starting point or pointer, but do not copy it.

[AussieHeart]

The trouble with AI is it can’t distinguish between fact and fiction. Until that can be corrected (can’t see that happening), we risk being misled.

[daverussell in reply to AussieHeart]

Absolutely, though there is also much human-generated misleading information out there. The very data AI it is trained on.. The benefits of AI is that it can tailor itself, and be tailored, to it's audience. With the availability of a connected world (that I think we now take for granted) it could help to address anxieties at the touch of our finger tips.

[dutyhog in reply to daverussell]

I’m a ChatGPT user, mainly to plan lectures, and have found so far that it mainly confirms what I already know about my subjects but saves a bit of brainstorming time, and sometimes reminds me of things I wouldn’t have remembered. For learning something new it has been disappointingly vague.

I followed your link, but because I already have an account I didn’t want to continue to start a nw one, so I can’t comment on it.

[daverussell in reply to dutyhog]

I know what you mean about vague. It saves me hours of time, which I put into improving the website and how I can build and schedule lessons from the scheme of learning ,- the AI is the relative easy bit given the questions and correct data.

That applies to anything such as the GPT I've created and an Tinnitus expert knowledgeable about AI and counseling could possibly help build a useful tool.

[surreycccfan]

AI is here to stay, when I assess the applicability of AI I consider the following things:

1) What is the source data, if the AI source data is the whole of the internet (slight exaggeration) then I take results with a healthy dose of cynicism

2) What are the foundation models used to generate the sequences. I think for T data I would like to see a solid foundation model built for health data elements. Not sure I am convinced on how Chatgpt trains, validates and tunes models specific for health.

3) I want my AI to be responsible, transparent and explainable data and AI workflows. Chaptgpt has not been great with any of these elements (in my opinion)

Chatgpt is not my preferred AI tool as too much untrusted data, general models and not great transparency on governance. However, is it any better or worse than the 'wild west' of the internet, probably not. Not sure if it will put people off as most folks already search the internet when they first get T.

Where I see the power of AI for T is (1) AI helping scale workloads to enable researchers to identify meaningful results and insights on potential treatments and (2) I can see when health foundation models are tuned and robust enough with meaningful data and user friendly virtual agents, it will give users early on with their T meaningful answers on management techniques