Artificial Intelligence
-
The biggest issue in Veterinary Radiology is that there are very few people doing the job in relation to the demand. Just like anything, the turnaround time from case submission to a report being generated can be anywhere from a day to a week depending on that caseload.
People demand things now and are unwilling to wait, so AI is a perfect tool to speed up that process. GP's who are already under a lot of stress from their normal caseloads, don't need owner's of pets phoning every hour asking if the report is done yet and are using that tool to speed up that process.
The problem is, things get missed...
We do our best to turn reports around as quickly as possible, but sometimes caseloads, life, illnesses, etc. get in the way. I would rather have a human check my X-rays and imaging everyday of the week, but that's not what society wants.
They want that information now and there will be compromises until technology catches up with that want and the ability to produce reports that are accurate.
On the other hand there are reports/studies out there that AI has caught tumours at early stages that Radiologists can't see yet. There is potential there and hopefully there will be some cohesion between the two.
"I don't give a shit what anyone else is doing, we will do what is best for us and our customers" - Giles P. :)
-
@WhiskeySandwich I literally loled at that
️WTB
IHSH-IHG-BLK XXL
Sugar Cane Coke Stripe SS L charcoal -
@goosehd i get you. We do our best and its not a perfect world. Shit happens and stuff gets missed, even by humans. You can't expect a product to be less flawless than it's creator, i guess. Also, its true that if it can save some time, even while it waits for a human to get to it, then it is still helping.
-
@popvulture The thing that gets me about that selection of a name, is that it kinda implies that data is just suspended in space. Like a cloud isn't really a vessel for information. Misrepresents that a record must have a tangible place to be stored.....except for the Akashic Record, of course...
-
Ha right, and not the reality of server farms devouring tons of energy. That aspect itself will be an interesting crisis point for AI to eventually reach.
-
@Matt It's funny that you bring that up as I thought the same thing once or twice myself. I think it's a natural progression for all people to look back at the good ol' days and worry about the future. Your grandparents did it, your parents, and you...
We keep moving forward for better or worse and only time will tell which side of the equation we and our offspring are on.
-
@Matt I've gone down that road of thought before, but I always end up back at: The whole point was to give them the tools to do better than me, so they can take the keys one day.
-
-
@louisbosco hahaha, classics! and movies like those both guided the reality and set the tone for our reception of it
-
I suppose the downfall will come when we decide how much we rely on AI.. with machine learning, would they learn that we are the cause of all problems? sounds like a familiar plot to me.. haha
-
@WhiskeySandwich I find your signature equally poignant!
Mad how former emperors of Rome, and dump truck graffiti scrawlers can touch a man's soul in similar ways...
-
@popvulture said in Artificial Intelligence:
I’ve been subscribing to the mindset that it’ll be less of a job replacer and more of a job augmenter/assistant in the future. There are all kinds of things that stuff like Midjourney, Stable Diffusion, etc already do to make my life a shitload easier.
This. I (a lawyer) have been dabbling with AI for the last few weeks, and so far it's really good at some things, and then hits a wall.
For instance, if an interesting case gets a judgment, I can ask it to summarise the case, and it will do a very decent job. If I ask it to weigh the impact of that case in the broader corpus of caselaw in that area, it throws a wobbly. So far, at least, it looks like it will help us with understanding an area of law, but won't replace the ability to think creatively about how to apply that law to our clients' issues.
As a general point, I find there is an art to crafting a good prompt too. If I just say "summarise case [X]" then it does a passable but not too impressive job. But if I say "summarise the legal arguments, judicial reasoning, and broader implications of case [X]" then it is much more impressive. You basically need to treat it like an intelligent ignoramus: capable of doing a lot but unaware of any context. So if you tell it (i) what you want; (ii) why you want it; (iii) what you expect it's answer to include; and (iv) how you want it presented, it does a lot more for you than otherwise.
The scary thing is that this tech is as bad as its ever going to be right now. That chap from one of these AI companies had serious egg on his face just a couple of weeks ago when he said that he didn't think an AI would ever be able to generate video, and then OpenAI announced Sora literally the next day. I've been trying to write a Powerpoint to educate my colleagues about all this stuff and how to use it, and it's been a really hard project to complete because there's constant development and news in this field and I have to keep updating the damned thing!
Take the dive...
-
Yep, post ChatGPT advancements in this field is measured in days unlike in decades it was before.
-
@EdH “ You basically need to treat it like an intelligent ignoramus: capable of doing a lot but unaware of any context.” This bit for sure. It’s smart as hell and can do the task, but you have to be very clear in defining the task. If you want a specific answer, ask a specific question.
-
@WhiskeySandwich said in Artificial Intelligence:
If you want a specific answer, ask a specific question.
To which I'd add, "and direct it to the resource". I've had a few examples of it referencing older law - presumably because there will be more written about old law - which has been superseded. But when I've specifically told it "make sure you refer to this" and then given it a link or copy-pasted the updated legislation into the chat box, it has been able to incorporate that into its response.
Take the dive...
-
@EdH said in Artificial Intelligence:
So if you tell it (i) what you want; (ii) why you want it; (iii) what you expect it's answer to include; and (iv) how you want it presented, it does a lot more for you than otherwise.
I find that a tad scary in that your request predisposes the answer and effectively reinforces any preconceived notions you may have had in asking the question in the first place.
In science we are taught to avoid bias in the question at all costs. While the bias is still inevitable, the following steps (research, hypothesis, experiment, data analysis) help minimize the bias in the conclusion.
Does the scientific method or a similar construct exist in law?
"I don't give a shit what anyone else is doing, we will do what is best for us and our customers" - Giles P. :)
-
@EdH yeah that could definitely be tricky. I find search engines doing that in general and it’s very frustrating. Especially if it finds false or speculative information, or opinions even, but it presents them as results all the same. There’s just so much junk data out there…. You have to wade thought an ocean of bs just to get a little nugget of truth or fact.
-
@goosehd yes, in a way, it’s just telling you what you want to hear. Lol
