First question. What happens when the old cohort who don’t use AI die out? We are not seeing a decrease in adoption of AI use in these fields but an increase. And that increase is compounded by the people who never learn such skills in the first place because they use AI to do the work for them that gets them through the schooling that would teach them such skills.
Second question did you read the parts about how news media is portraying studies, or the parts about how studies are using miniscule (entirely too small) sample sizes, or the parts where the studies aren’t being peer reviewed before the articles relating to them spread misinformation about them?
The tools aren’t ready for prime time use, but they are being used in medicine.
You seem to have glossed right over the detriments that doctors and researchers are already experiencing with Generative AI LLM’S (you keep saying ML, and that’s not exactly the subject we’re talking about here), And the fact that it takes extensive experience, and a knowlegable expert to fix, in a world where the AI LLM’S are contributing to a significant decline in the number of people who can do that, meaning that correcting LLM outputs will happen less and less over time because they require people to correct them, people to create the data sets, and people to understand and have expert knowledge in the data sets/subjects in order to verify the outputs and fix them.
I can appreciate you not wanting to speak on a hypothetical but that just doesn’t ring true to me either because it means you haven’t thought about the implications of this tech and it’s effect on the industry being discussed or you have and you are ignoring it.
Not weighing the huge benefits of a tech against its detriments is dangerous and a very naive way to look at the world.
For your first question, what you’re describing is a problem with education and staffing, not a problem of the tool itself. I’m not suggesting you keep around ‘one old man who hates AI’, my pitch you bar the use of AI for human-level checks.
For your second, yes I saw the part about how news and media are representing AI in healthcare, but I don’t really see how news or media are relevant here. Could you explain this a bit for me?
I don’t intend to gloss over the issues with Generative AI/LLMs, I tried to be specific in my separation of ML from them in my original comment where I said LLMs in their public facing version (ChatGPT, Claude, whatever) aren’t very useful.
The original comment I replied to asked “is “AI” even useful (etc)” but also mentioned LLMs. I was trying to make the point that LLMs aren’t the only type of AI and that others can be employed to great effect. If that was unclear, that’s my bad but that was my intention.
The reason I don’t want to engage with a hypothetical is because I could just as easily counter with “what if it diagnoses at a 100% success rate? What if fear of losing skills results in doctors never wanting to use AI, resulting in more deaths?” Neither hypothetical argument is really very helpful for the discussion. I promise you I’ve thought about this a lot (but again, I’m not an expert, nor am I in the field), but more importantly I have friends finishing doctorates in the bioinformatics field whom I get some insight from, and I’m, at least at this point, convinced of the benefits.
First question. What happens when the old cohort who don’t use AI die out? We are not seeing a decrease in adoption of AI use in these fields but an increase. And that increase is compounded by the people who never learn such skills in the first place because they use AI to do the work for them that gets them through the schooling that would teach them such skills.
Second question did you read the parts about how news media is portraying studies, or the parts about how studies are using miniscule (entirely too small) sample sizes, or the parts where the studies aren’t being peer reviewed before the articles relating to them spread misinformation about them?
The tools aren’t ready for prime time use, but they are being used in medicine.
You seem to have glossed right over the detriments that doctors and researchers are already experiencing with Generative AI LLM’S (you keep saying ML, and that’s not exactly the subject we’re talking about here), And the fact that it takes extensive experience, and a knowlegable expert to fix, in a world where the AI LLM’S are contributing to a significant decline in the number of people who can do that, meaning that correcting LLM outputs will happen less and less over time because they require people to correct them, people to create the data sets, and people to understand and have expert knowledge in the data sets/subjects in order to verify the outputs and fix them.
I can appreciate you not wanting to speak on a hypothetical but that just doesn’t ring true to me either because it means you haven’t thought about the implications of this tech and it’s effect on the industry being discussed or you have and you are ignoring it.
Not weighing the huge benefits of a tech against its detriments is dangerous and a very naive way to look at the world.
For your first question, what you’re describing is a problem with education and staffing, not a problem of the tool itself. I’m not suggesting you keep around ‘one old man who hates AI’, my pitch you bar the use of AI for human-level checks.
For your second, yes I saw the part about how news and media are representing AI in healthcare, but I don’t really see how news or media are relevant here. Could you explain this a bit for me?
I don’t intend to gloss over the issues with Generative AI/LLMs, I tried to be specific in my separation of ML from them in my original comment where I said LLMs in their public facing version (ChatGPT, Claude, whatever) aren’t very useful.
The original comment I replied to asked “is “AI” even useful (etc)” but also mentioned LLMs. I was trying to make the point that LLMs aren’t the only type of AI and that others can be employed to great effect. If that was unclear, that’s my bad but that was my intention.
The reason I don’t want to engage with a hypothetical is because I could just as easily counter with “what if it diagnoses at a 100% success rate? What if fear of losing skills results in doctors never wanting to use AI, resulting in more deaths?” Neither hypothetical argument is really very helpful for the discussion. I promise you I’ve thought about this a lot (but again, I’m not an expert, nor am I in the field), but more importantly I have friends finishing doctorates in the bioinformatics field whom I get some insight from, and I’m, at least at this point, convinced of the benefits.