96
Do you feel that use of generative AI and LLMs is ever justifiable? If so, when and where? If not, why not? (lemmy.world)
submitted 5 months ago* (last edited 5 months ago) by to c/asklemmy@lemmy.world
159 comments fedilink hide all child comments
 

You can take "justifiable" to mean whatever you feel it means in this context. e.g. Morally, artistically, environmentally, etc.

you are viewing a single comment's thread
view the rest of the comments
[–] 29 points 5 months ago (2 children)

Medicine.

Evidence shows that some highly specialised models are better at things like detecting breast cancer in scans than human doctors.

Properly anonymised automatic second scans by an AI to catch the markers that human doctors miss for another review by a specialist is an excellent potential use case for an LLM AI.

Transcription services can save doctors huge amounts of admin time and allows them to focus on the patient if they know there's a reliable system in place for typing up notes for a consultation. As long as it's treated as a "please review these notes are accurate" rather than treated as a gospel recording and the data is destroyed once it's job is complete and the patient has been able to give informed consent.

The way these things are being used in actual medical contexts right now is frankly terrifying.

  • permalink
  • fedilink
  • source
  • hideshow 4 child comments
    •
  • [–] 24 points 5 months ago* (last edited 5 months ago) (1 child)

    I had a colonoscopy last year (such fun!) and there was an 'AI' monitoring the camera feed to detect anomalies. If it spotted something it just drew the doctor's attention to it for his expert, human review. I was ok with that. Effectively an extra pair of eyes that can look everywhere on the screen all at once and never blink.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 3 points 5 months ago

    That's how AI systems should be used. A "heads up, something weird here" system.

    I could also see it being used well like this for patient history analysis. Often a doctor is treating 1 symptom of something larger. They can't see the wood for the trees. An LLM could pick out oddities and flag them. The doctor can then filter out the mistakes and hallucinations, but be alerted to rare or unusual conditions that match the patient's symptoms and history.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 5 points 5 months ago*

    Yeah the sciences in general I'd say. There's a project aiming to translate the tens of thousands of cuneiform clay tablets that sit in storage all because there's like a handful of people in the world that can read them- AI is an amazing way to mass translate them and unlocking vast troves of hitherto completely unknown ancient knowledge.

    The problem is not even the AI, but the scientists themselves who guard the tablets jealously because they don't want anyone else to translate "their" tablets that they dug up, even though they are incapable of possibly make a dent in the sheer volume in their collected lifetimes.

    Imagine, so much information encoded, from thousands of years ago that could reveal so much about the origins of our culture and civilization!

  • permalink
  • fedilink
  • source
  • parent