In the general case it's usually not possible to accurately review an individual physician's performance. The software developers here on HN like to think in simplistic binary terms but in the real world of clinical care there is usually no reliable source of truth to evaluate against. Occasionally we see egregious cases of malpractice or failure to follow established clinical practice guidelines but below that there's a huge gray area.
If you look at online reviews, doctors are mostly rated based on being "nice" but that has little bearing on patient outcomes.
A friend of mine had such a bad experience with _multiple_ American doctors missing a major issue that nearly ended up killing her that she decided that, were she to have kids, she would go back to Russia rather than be pregnant in the American medical system.
Now, I don't agree that this is a good decision, but the point is, human doctors also often miss major problems.
Amazing how you can just deflect any criticism of LLMs here by going “but humans suck too!” And the misanthropic HN userbase eats it up every time.
We live during the healthiest period in human history due to the fact that doctors are highly reliable and well-trained. You simply would not be able to replace a real doctor with an LLM and get desirable results.
> Amazing how you can just deflect any criticism of LLMs here by going “but humans suck too!” And the misanthropic HN userbase eats it up every time.
I think it's rather people trying to keep grounded and suggest that it's not just the hallucination machine that's bad, but also that many doctors in real life also suck - in part because of the domain being complex, but also due to a plethora of human reasons, such as not listening to your patients properly or disregarding their experiences and being dismissive (seems to happen to women more for some reason), or sometimes just being overworked.
> You simply would not be able to replace a real doctor with an LLM and get desirable results.
I don't think people should be replaced with LLMs, but we should benchmark the relative performance of various approaches:
A) the performance of doctors alone, no LLMs
B) the performance of LLMs alone, no human in the loop
C) the performance of doctors, using LLMs
Problem is that historical cases where humans resolved the issue and not the ones where the patient died (or suffered in general as a consequence of the wrong calls being made) would be pre-selecting for the stuff that humans might be good at, and sometimes wouldn't even properly be known due to some of those being straight up malpractice on the behalf of humans, whereas benchmarking just LLMs against stuff like that wouldn't give enough visibility in the failings of humans either.
Ideally you'd assess the weaknesses and utility of both at a meaningfully large scale, in search of blind spots and systemic issues, the problem being that benchmarking that in a vacuum without involving real cases might prove to be difficult and doing that on real cases would be unethical and a non-starter. And you'd also get issues with finding the truly shitty doctors to include in the sample set, sometimes even ones with good intentions but really overworked (other times because their results would suggest they shouldn't be practicing healthcare), otherwise you're skewing towards only the competent ones which is a misrepresentation of reality.
The fact that someone would say stuff like "Doctors are more like machines." implies failure before we even get to basic medical competency. People willingly misdirect themselves and risk getting horrible advice because humans will not give better advice and the sycophantic machine is just nicer.
> I think it's rather people trying to keep grounded and suggest that it's not just the hallucination machine that's bad, but also that many doctors in real life also suck
No, you see this line or argumentation on every post critical of LLM's deficiencies. "Humans also produce bad code", "Humans also make mistakes" etc etc.
> No, you see this line or argumentation on every post critical of LLM's deficiencies. "Humans also produce bad code", "Humans also make mistakes" etc etc.
So your reading of this is that it's a deflection of the shortcomings?
My reading of it is that both humans and LLMs suck at all sorts of tasks, often in slightly different ways.
One being bad at something doesn't immediately make the other good if it also sucks - it might, however, suggest that there are issues with the task itself (e.g. in regards to code: no proper tests and harnesses of various scripts that push whoever is writing new code in the direction of being correct and successful).
Even in medicine, often the difference between drug A and drug B is the difference between the two in statistical terms. If drugs were held to the standard "works 100% of the time", no drug would ever be cleared for use. Feelings about AI and this administration are influencing this conversation far too much.
It's like people want to remove the physician or current care from the discussion. It's weird because care is already too expensive and too error prone for the cost.
Medical errors are one of the leading causes of death. It's a real catch-22. If you're under medical care for something serious, there's a real chance that someone will make a mistake that kills you.
You also don't sue for malpractice unless something goes catastrophically wrong. I've had doctors make ludicrously bad diagnoses, and while it sucked until I found a competent doctor and got proper treatment, it wasn't something I was going to go to court over.
We do in fact prosecute drug dealers. In fact, if anything China's opium problems of the 19th century were ended by communist cadres mass executing drug dealers (and some users) in the 20th century.
Many countries have successfully cut down sugary drink consumption through regulation.
Where to draw the line is a matter of policymaking decisions and its not black and white. Its a very gray area that needs to be tuned carefully by society.
Still more laws and regulations. We're already drowning in regulations.
If you start to require such things, then you should also require that labels declare whether the artist indeed sung him/herself, and whether it was their real voice or some autotuned stuff.
> If you start to require such things, then you should also require that labels declare whether the artist indeed sung him/herself, and whether it was their real voice or some autotuned stuff
How do you complain about regulations, and then insist on more regulations? These seem like two completely orthogonal things to me. One is to prevent unmitigated spam on music platforms. The other corrects your pitch slightly.
"If *you* start to require such things, then *you* should.."
> prevent unmitigated spam on music platforms
That's not the true reason. The reason is that some clever people have found out how to earn money with fake musicians. As a musician myself I can tell you, that such operations make the already completely unjust compensation scheme even worse for real musicians. But even if this scam was avoided, we still would suffer from the exploitative and abhorrent compensation scheme imposed by the record monopolists (who simultaneously claim to represent the interests of artists while primarily lining their own pockets).
If you are interested instead whether it was a real artist or just an AI (or a "stunt performer" pretending to be an artist, which includes all people who can't sing in tune without autotune), then you should be as consistent as the food industry. Personally, I'm more interested in the music itself. It's nice to know who played it, but that doesn't change the musical quality.
No that is the reason. The fact that AI music can be created, marketed, and flooded off of prompts that, themselves, can be automated, is remarkably unhealthy for the music space. And saying the music industry is fucked anyway is not very convincing.
> or a "stunt performer" pretending to be an artist, which includes all people who can't sing in tune without autotune
Really? Sorry but I don't consider people like that stunt musicians. There's a wide range of human expression to be shared beyond singing technically well. A stunt musician would be someone doing nothing. Like prompting an AI.
Since you're keen on gatekeeping, I'm a musician. My whole family is. You may very well play music, but are you a musician? Your comment is genuinely more offended with auto tune than AI music. I can't imagine working myself to that conclusion, and I think that says a lot about your values, and they're not in line with any musician I know. They're more in line with a developer or A&R
The GP never explicitly mentioned law. Now's would be a great time for the recording industry to prove their value by instituting a framework for self-regulation. They could actually differentiate themselves from things like Bandcamp if they had some proper enforcement on this topic. But of course, we know that they're purely extractive, so that won't happen.
The fellow said "It should be required to include the AI model as a featured artist"; the common means is laws and regulations, like in the food industry.
> They could actually differentiate themselves...
Instead, it will go in exactly the opposite direction: the record monopolists will use AI themselves to further improve their already high margins. They have already secured the technology for this, if you have been following the news about Suno and Udio.
My main critique is that it's not clear where the loss of identity came from?
Is it that large corporations ruined being a programmer? Don't work for one.
Is it that silicon valley culture ruined being a programmer? Don't live on the west coast.
Is it that LLMs ruined being a programmer? Don't use them.
Or is it just that you're getting older, and it's not so much about programming as it is about the world not being the same one you grew up in. The inevitable alienation of aging.
Personally, I think programming is just as rad as it always was. More so even. It's never been easier to learn, there's more cool languages than ever, hardware is cheap, we just invented alien technology, basically every person has a computer in their pocket, basically every company needs programmers function, etc, etc.
Everybody is so down about the current moment in time, and no doubt there's plenty to be down about, but a quick read of history will tell you that it's always like this. Nothing to be down about. It's just business as usual. Social and news media have wormed their way into everyone's vibes and feeling bad is addictive.
The only thing we can do is have a good attitude, roll up our sleeves, try to fix what's broken, and keep on keeping on. The responsibility of every age.
But it will cut it, assuming you're moving your body the same amount every day. It just may take a while if you only cut a single slice of bread and you're wanting to lose a lot of pounds.
Your body mass doesn't materialize out of nothing. Food in, body mass out. Less food in, less body mass out. Simple as that. Everything else is optimization that's not really required, just eating less, patience and consistency.
His point was when you eat more calories than you need, then the body can afford to be sloppy (inefficient) about absorbing all those calories.
When you cut down a little on food but are still above or at your daily calorie requirement, the body can adapt by increasing its conversion absorption efficiency and in that case one wouldn't lose weight, because metabolically it's still absorbing the same amount of calories.
Do you have a source I can read through to understand more? I couldn't find anything supporting that idea.
Before dieting, I would expect most people to be in energy equilibrium, where their weight matches their calorie intake and calorie expenditure. Changing one side of that equation will change the equilibrium, every thing else equal.
If you eat more, you will gain weight. If you eat less... you will lose it. If you want to keep losing it, you have to keep eating less. Every target weight has an associated calorie intake / expenditure.
No doubt there are metabolic levers to pull to optimize the timeframe and that psychology and lifestyle play a big part of caloric intake, but, again, thinking about all that isn't really necessary.
Just eat less that you used to and be consistent about it. And if you're feeling spry, move more than you use to too. Keep tapering down until you're at your goal weight. This is the diet that is probably the best fit for 99% of people who are overweight. Dumb simple. No way to fail. Literally nothing to think about except the spoon, and maybe which route you're going to take walking around the block.
Those are mostly just arguing that exercise isn't a good way to lose weight. No disagreement about that, although to be super nitpicky, "not good" doesn't mean "doesn't work", they just suggest that there is a cap on how much exercise you can do before your body stops burning more calories, so there's an effective limit on that side of the energy equation.
Also, from the second one:
Long-term maintenance of weight loss requires sustained energy balance at the reduced body weight. This could be attained by coupling low total daily energy intake (TDEI) with low total daily energy expenditure (TDEE; low energy flux), or by pairing high TDEI with high TDEE (high energy flux
We’ll miss creating something we feel proud of, something true and right and good. We’ll miss the satisfaction of the artist’s signature at the bottom of the oil painting, the GitHub repo saying “I made this.”
I definitely feel this some days.
It used to be that I would nitpick code to death to get it just so, proud of the artistic decisions I made. Not only was it functional, but it was beautiful, crisp, elegant, clever.
Just the right abstractions. Complicated problems reduced to perfectly legible structures, easy to read for newcomers, easy to extend when new problems arise.
Extracting that from an LLM just doesn't produce the same feeling, even when it produces the same results. And they don't reliably produce the same results yet even, just poor imitations, however functional.
On the other hand, when they work, I am empowered to produce code I would have never thought of. I'm empowered to bounce an idea off an oracle that knows all the answers, and can tailor its responses to my exact use case. I'm coming around to finding the joy in that. I have to.
I still nitpick it to death. If anything, I have more opportunities to nitpick code. LLM code is usually filthy: terrible abstractions, no care for complexity, cohesion, and coupling metrics.
Debugging with LLMs is also a mixed experience; they can identify plenty of hypotheses very well, thanks to millions of dollars spent in RL, but they lack a more profound understanding of causal chains. They will try to change more than one thing at a time, even when doing so will completely invalidate the experiment; they will get into bizarre loops and weird tangents. They can help a lot, but if you want to have good results, it is way better if you strongly take control of the steering wheel.
LLMs are basically the average developer, although with way more breadth. But still not much depth.
For work, I regularly have 2-4 agents going simultaneously, churning on 1-3 features, bug fixes, doc updates.
I pop between them in the "down time", or am reviewing their output, or am preparing the requirements for the next thing, or am reviewing my coworkers MRs.
Out of curiousity, how do you manage the constant context switching? It's hard for me to manage the context of one coding session, let along 2-4 sessions.
reply