mgraczyk

The difference in out come is

With my change: 95% of people who are shown scans have cancer and are treated earlier. 5% of people do not have cancer and get CT scans. 0.5% of people get useless biopsies Without my change: many of those 95% die, the 0.5% do not get useless biopsies

And the beauty of this is you can pick the percentage!

> If you have an growing mass in your body, then after a year, it may very well no longer make a difference whether you treat it or not. Or it may be that you would have lived another 20 years just fine

This is just wrong for many parts of the body. In your brain? Your lungs? Growing for a year between 3 scans 6 months apart? Extremely unlikely to be benign

> The same applies here: detection, even if 100% accurate, doesn't mean anything. You need to show that what you do with the result actually helps.

This is wrong. If you had a 100% accurate cancer detector, fewer people would die of cancer with no downside

I have, it's the fault of how medicine is practiced to reduce cost. It's completely avoidable, you can just not tell people their scan results if they have no symptoms and the detection is less than 95% likely to be cancer. This is strictly better than the status quo because the only difference is some people who almost certainly have cancer learn that they have cancer and nothing else changes

I mean that you should stay up to date and practiced on how to get the most out of models. Using harnesses like Claude code sure, but also knowing their strengths and weaknesses so you can learn when and how to delegate and take on more scope

You'll be behind by a few months at least, and that could be anywhere from slightly harmful to devasting to your career

Did you read my two options? Do you agree option 2 is better than option 1? If so, then scans are better than no scan

You don't need to show that it's possible to avoid false positives. That's doctors being irrational.

You only need to show that it's possible to build a diagnostic system that's better than no testing, and I have shown that already

I don't think so. That's a good point but the capability has been outpacing people's ability to use it for a while and that will continue.

Put another way, the ability to use AI became an important factor in overall software engineering ability this year, and as the year goes on the gap between the best and worst users or AI will widen faster because the models will outpace the harnesses

Even if you believe that many are too far on one side now, you have to account for the fact that AI will get better rapidly. If you're not using it now you may end up lacking preparation when it becomes more valuable

Doctors here are cognitively captured by a system designed to limit cost (and that's mostly a good thing)

But scanning frequently is overwhelmingly good for the patient. The problem is the doctors. Imagine two possibilities. 1. You scan every six months and a doctor reviews your scans but never tells you anything no matter what 2. You scan every six months and a doctor reviews your scans and only tells you results if you have an obviously growing mass that has a probability greater than 95% of being cancerous

Obviously #2 is better for the patient than #1, but #1 is equivalent to never testing if you ignore cost.

So the actual reason we don't have effect frequent scans combined with effective diagnostic techniques is cost, and doctors cope with this reality by saying clearly wrong things about "over diagnosis". It's a local minimum of the payer/provider dynamic that has nothing to do with scans per se.

And yes getting frequent full body MRIs is still overwhelming the right thing for the patient.

This entire chain of reasoning takes for granted that there won't be a singularity

If you're talking about "reforming society", you are really not getting it. There won't be society, there won't be earth, there won't be anything like what you understand today. If you believe that a singularity will happen, the only rational things to do are to stop it or make sure it somehow does not cause human extinction. "Reforming society" is not meaningful

The headline seems false, should we change it? It doesn't look like they showed any case where any autonomous car or drone obeyed prompt injections

But unfortunately it does sometimes work, for example in Yugoslavia. And it would have worked in Iraq if we hadn't dismantled the entire civilian infrastructure.

We never do anything that reduce model intelligence like that

That issue did not have any time of day dependence

It's neither of those things. The answer is in your quoted sentence. "model training"

Seems very unlikely considering the current equilibrium is for many services to offer ad-free premium versions, even recently creating these tiers. If anything the trend is moving in the other direction, more services offer paid ad free tiers.

Youtube, Spotify, Amazon Prime video, etc

The ARPU will likely remain low enough that it's not worth potential churn to competitors

Do you think this is evidence that it was designed that way, as opposed to the exact opposite?

It was not designed that way, so they had to specifically do extra things to make it not that way.

This is like saying that python is designed to be slow and linking to a post about speeding up the python interpreter as evidence

Do you have any evidence that this claim is true? That it was "designed" for this? That would be a pretty difficult conspiracy for them to keep secret

Over 50% of the German population is overweight, and over 20% is obese

I think many Germans will start using these drugs once they become cheaper

What is the mechanism that you believe would cause this to lower home prices? Early signals are that the announcement alone will increase prices

Putting aside whether this person is just lying, the actual claim doesn't seem possible if you think about it for more than a few seconds.

The claim is that the benefits fee does not go to the driver, but that is not possible in the only interpretation of the claim that matters. When the law in California was passed that caused these fees to be added. Drivers immediately started receiving benefits required by the legislation. Those benefits are paid by the delivery company. At the same time, the company added this fee. So the only question is, how much fee money is collected relative to the cost of the benefits?

From a very coarse BOTEC, it seems that the fees are probably not enough to cover the cost (for doordash), so in fact the fees do go "100%" to the drivers, in the sense that more than the amount collected by the fee is spent on the legally required benefits the the fee is based on

The incentives are that a lot of the specific claims would be bad for long term retention and would be embarrassing if made public. That doesn't imply it's fake but it's not obvious any company would be incentivized to behave this way

I think if you paste our conversation into ChatGPT it can explain the relevant upsampling algorithms. There are algorithms that will artificially enhance edges in a way that can look like "AI", for example everything done on pixel phones prior to ~2023

And to be clear, everyone including Apple has been doing this since at least 2017

The problem with what Samsung was doing is that it was moon-specific detection and replacement

> somehow seem to be defending Samsung voraciously but emptily

The first words I said were that Samsung probably did this

And you're right that I didn't read the dozens of edits which were added after the original post. I was basing my arguments off everything before the "conclusion section", which it seems the author understands was not actually conclusive.

I agree that the later experiments, particularly the "two moons" experiment were decisive.

Also to be clear, I know that Samsung was doing this, because as I said I worked at a competitor. At the time I did my own tests on Samsung devices because I was also working on moon related image quality

You're mistaken and the original experiment does not distinguish between classic edge aware upscaling/super resolution vs more problematic replacement

That is not true

For example see

https://en.wikipedia.org/wiki/Edge_enhancement

And? What algorithm was used for downsampling? What was the high frequency content of the downsampled imagine after doing a psuedo inverse with upsampling? How closely does it match the Samsung output?

My point is that there IS an experiment which would show that Samsung is doing some nonstandard processing likely involving replacement. The evidence provided is insufficient to show that

I think Samsung was doing what was alleged, but as somebody who was working on state of the art algorithms for camera processing at a competitor while this was happening, this experiment does not prove what is alleged. Gaussian blurring does not remove the information, you can deconvolve and it's possible that Samsung's pre-ML super resolution was essentially the same as inverting a gaussian convolution

Well we don't ban cigarettes

But cigarettes are harmful for everyone, social media is not