jonex

As people have mentioned, it isn't that surprising that a specially trained model beats a general purpose model.

However, I think it kinda misses the potential of llms for diagnosis. What happens if we also include journal notes? What if we let the model generate questions to ask the patient and include those? What if the images aren't perfect examples and happen to contain out of distribution content like a tattoo?

There's no mention of the reason for the removal from what I can see, do you have a link that provides more context?

Did you dissolve it in hot water before consumption?

For my master thesis I did a lot of digging into the research on success factors for startups. And at least based on what I found back then, you're right. The correlation is weak between any factor like this and success. Generally positive though, so it does make sense to try doing some planning. But it's definitely not a requirement.

In general, observable factors doesn't strongly correlate with success. Probably both because there are so many to choose from and because the real world is complex and typically doesn't align well with any predefined plan.

Looks like https://clackernews.com/ is leaking!

Feature request: Make it default behavior on phones that you can have multiple passwords, connected to different profiles. With no way to determine how many profiles a phone have.

I'm sure there's some people here working on mobile operating systems, might be worth considering?

Didn't it start with asm.js, a subset of javascript serving as a way to compile C code to be run in a browser? Then browser makers figured that it'd be better to have a dedicated target for this. So while it can be used to achieve performance in specific scenarions, it's largely designed with the goal in mind to be able to run non js code in a browser. The wasm toolchain Emscripten encompasses this notion quite clearly as it emulates things like filesystems etc. If the main goal was faster execution, they would probably have gone a different route. PRobably even gone for a new language altogether.

I'd like a toolchain better targeted for the pure acceleration use case though. Emscripten adds a lot of bloat and edges just to serve out of the box posix compatibility. Which is nice for quick demos of "look I can run Doom in the browser"-kind. But less useful for advanced web app usage, where you anyways will want to keep control of such behavior and interact with the browser apis more directly.

It's the difference between proportional voting vs winner takes it all. In the latter case you can't really hold politicians accountable, as you will have to choose between effectively throwing your vote away or voting for the one opposition candidate, that often will be just as bad.

While the UK have some level of representativeness, each circuit has a winner takes it all structure, making change quite hard to achieve on a larger scale.

Good to be aware: Full face helmets restrict your view more, so there is a safety trade-off to be made.

Maybe he's actually wearing protective gear? Jeans with slide tolerant layers and padding are pretty common as warm weather gear and there are jeans style jackets as well. If you don't know what to look for it can be hard to tell the difference to regular clothing.

This isn't just about perception. If you:

* Have a license * Wear appropriate gear * Follow speed limits * Don't drink and ride * Aged somewhere 30-50 * Have more than a few months experience on your bike * etc.

You are statistically "one of the safer ones". Not safe, you are never truly safe when in traffic.

I find that annoying too. In case you haven't seen, there's the sponsor block extension for that, which is not limited by anti-adblock measures.

It takes away the ability to know what it does though, which is also often considered an important aspect. By not publishing details on how to train the model, there's no way to know if they have included intentional misbehavior in the training. If they'd provide everything needed to train your own model, you could ensure that it's not by choosing your own data using the same methodology.

IMO it should be considered freeware, and only partially open. It's like releasing an open source program with a part of it delivered as a binary.

To me it sounded like click bait. So I checked comments and concluded that it was indeed about the very visible RDS signal and not some hidden channel used by some secretive agency that would indeed be somewhat mysterious.

I don't think the fact that it worked in generating clicks is really an argument for bait titles. Given the positive comments about the content I think some editorializing could have been helpful to focus on the hacking journey aspect though, which seems to be the point rather the specifics of RDS itself.

I sure hope this does not mean a reduction in quality, one of Legos big benefits has been the longevity, lego pieces from the 70s are still in my family's rotation. Reuse should always come before recycle.

They didn't specify what they qualified as a multivitamin, my guess is that the definition is rather loose due to the scale of the study. There's a lot of different mixes of multivitamins in the store. You would buy the one that fit's the efficiency pattern you want to prevent. If your goal is to specifically reduce mortality rather than prevent short term fatigue, you would need to pick the right one.

From what I can tell, they haven't ensure that the long term consumption pattern was consistent either, so they may be mixing people who took multivitamins for a year around the initial study with those who took it every day for their whole life. That would reduce the effect size significantly.

The general advice on vitamin supplements is usually to take the ones you have explicit reasons for taking and to focus more on food in general. This study, while possible useful as a way to debunk highly unrealistic claims of multivitamin effectiveness, doesn't really change the picture here.

tsan will catch a bunch of potential race conditions for you, under the condition that you run it somehow. How to make sure it's run? Well, add a test for the relevant code and add it to your tsan run in your CI and you'll certainly catch a bunch of race conditions over time.

This has saved me a bunch of times when I've be doing work in code with proneness to those kind of issues. Sometimes it will just lead to a flaky test, but the investigation of the flake will usually find the root cause in the end.

I like the idea of focusing on the key notion of wether the task will succeed or not. A challenge in these things are to design things such that users understand what is communicated, and you also want something that fits the style of the app overall.

Personally I'm a fan of progress bars that indicates actual progress rather than any estimates the developer though would be good to provide. But if getting-stuck probability could be hinted at it would be great.

Maybe, whenever the bar gets stuck, it can slowly start greying out based on the probability that it's completely dead? That way an attentive user can use this as a clue for when to reload and try again, while not breaking the previous behaviour or introducing too much confusing information.

It's clearly not properly designed to be for swedish, this is noted in the reply, apart from some letters being off, it lacks åäö. But it's still not clear why they would use "u-boat" for u. While apparently correct english (never heard it), it being Swedish ubåt seems higher likelihood to me.

I'm wondering if they might have combined mixed two designs, as a lot of the letters works in both. Maybe they looked at two pictures and didn't realise one was for another language? It seems possible to me that they'd design the ball based on product pictures, and those are often not showing all angles of a product.

Don't think it's easy to say one answer or

Weight gain does often happen a few years post treatment. Even then it's beneficial compared to no intervention, which would most likely have them be even worse off five years later. Long term health outcomes are better for those who get treatment than those who don't.

The dichotomy is relevant if you are of the opinion that some of the foundational design choices of Carbon makes it less viable as a C++ successor based on the similarity of those choices to other projects that were struggling due to this. The talk is arguing for the validity of the construct with various examples, not just Dart.

The construct can be useful without being the end-all answer to what should be done and I would definitely advice you to watch the full talk if you haven't yet. And the consider how each point may apply or not apply to Carbon. Don't just dismiss it on the notion that Carbon is different from Dart.

Actually, I'd encourage you to reach out to Herb Sutter and ideally meet up in person to discuss the matter. Your goals are aligned in many ways and while you have different approaches a lot of good can come out of sharing ideas.

It's not at all hard to build a sane codebase in C++ if you are a group of experienced C++ devs with an eye for detail. Not saying mistakes never happen, but most of the time the tools available are sufficient to build correct code if you know how to use them right.

However, the problems abound when you have less experienced devs in a less structured environment and working with legacy C or C-style code, the latter often being the reason for why C++ was used in the first place. For this, something like cpp2 can be revolutionising as it makes it easier to write correct code and provides mechanisms to disable many of the unsafe patterns.

If Kagi were as popular as Google, you'd quickly see pages appear "without ads", that looks suspiciously like what an ad would say. The problem is SEO, and it affects the mainstream search engine because that's what's worth optimising for.

For instance, Google product search results would likely be greatly improved by always showing relevant reddit posts first. But this would last a very short amount of time until reddit would be completely drowned in junk. (I've seen people argue that this is already the case, but searching reddit still tends to work well for the products I search for)

We can be happy that Google search hasn't gone this way, and as a Kagi user you should be glad that it's not a more popular engine, as the quality of results likely highly depend on that just fact.

In a way the ads-filled product search results are better, because at least they're honestly trying to sell you the product you are searching for rather than making fake comparison sites pretending to be genuine content.

There's certainly a lot of ways food substances can interact with bodily systems. However it can be hard to discern the exact source as different substances are correlated. Like you mention, issues with FODMAP is common, and those are often correlated with carbs. So there's a significant risk that a real effect is misattributed. Ideally you'd test if by blinding the intake, but that's often impossible or at least infeasible, so in the end we tend to latch on to some hypothesis without too much evidence.

It's also good to be aware of how strong the nocebo effect can be in this context. If you are convinced of the danger of something it can lead to significant issues. For an extreme example, severe airborne peanut allergy seems to be hard to be reproduce in controlled studies, yet there's people having significant symptoms from this, like lack of breathing etc.

Basically, when it comes to food, a lot of skepticism, including to our own experience, is often required. The risk is that we otherwise end up with unnecessary restrictive eating.

Personally I have a lot of things I "can't" eat, but even then, I try to challenge it now and then, sometimes I find something in a category I can actually eat and sometimes it turns out that I can eat it in low amounts or slowly learn to handle it.

The problem with the typical CT isn't that they call out bad things. The problem is that they tend to use the rumours of bad stuff and build complex narratives around it that ends up much further than the original events would ever warrant.

It starts off with something like "CIA was behind the coup in country X". Which might have been true, but lacking concrete evidence it opens the door for other things without concrete evidence. Next claim will be something like media is intentionally not reporting about this. Then this is explained by "globalists" controlling the media and working with the CIA. Then it's "the Jews" are actually the ones working with the globalists here. Or it's the globalists trying to push vaccines even though they are apparently unsafe. Why, because they make mind control easier? Or maybe it's a way to reduce population.

This kind path into insanity is sadly not that rare. And while there's certainly people at the top of the funnel that has some reasonable skepticism while reporting somewhat unsubstantiated facts. There are also plenty who will start off relatively reasonable, but then take their followers through a journey down into conspiracy land.

A notable harm this had was indeed the covid vaccines. Some influencers in the health space, used to light weight CT content like, "the sugar industry is behind todays health recommendations" and "big pharma tries to push dangerous blood pressure medicine even though it doesn't help". So when vaccines came, they were primed towards looking for stuff like that and decided that vaccines were bad and presented this idea using biased statistics and CT ideas.

A lot of people are dead or were unnecessary Covid complications due to this movement. And arguably, the CT culture was a big contributor to this.

It's important to call out the actual bad stuff that's being done in various contexts, wether it's in the name of national security, medicine or even just industry causing pollution. It's also natural that people build up a sense of distrust and skepticism against unsubstantiated claims in this area given the potentially severe consequences downstream.

Interesting experiment and a good start at evaluating the concept, in particular, the fact that they compared to other methods makes it way stronger than similar concepts that forget to even compare to index. However, even though they clearly tried to avoid obvious pitfalls of backtesting, they also left lots of potential sources for bias.

Their method of collecting news is based on what popped up in their post-facto web search, it also relied on them classifying them for relevance. This can generate bias compared to, for instance, feeding all articles tweeted from a specific news source irregardless of "relevance". I also wonder what assumptions they've made about timing, slight errors in timing estimated could have huge impact on results.

The fact that the shorting seemed to make for the most part of the gains (although the long strategy was not terrible at 50% gain vs index going down) might also be something somewhat biased towards the fact that this was a downturn year and it might not fare as well in a more regular year.

The condition, according to the article, was not having taken pills and being dying, but "if a call-attendant believed a person might be at “imminent risk” of taking their own life in the next few hours, days, or week".

This is weak enough that it will happen many times for people who are not in a significant risk, but for whom such an intervention can be seriously detrimental to their life. Note that suicidal thoughts are most of the time just thoughts and not acted upon.

If nothing else, it should be made clear to the person calling what risks they are taking. Many people might believe that they have confidentiality and that they can talk freely without risk of consequence.

In Sweden they can certainly do so, if a person is a danger to themselves or others (as declared by two doctors) they can be put into closed mental care with police escort and drugged to the extent that they can't speak for themselves without any right to legal protection. The protection of their life is considered to be more important. I don't know about US law, but probably other countries has similar laws.

Linked by michaelwww in a previous thread. The video presents two very different approaches to dealing with the legacy of Javascript. It is particularly interesting to the see the pragmatic views of Hejlsberg contrasted with Bak's focus on getting the language fundamentals right.

For a class of problems, there's basically two options of language to use that would fit the requirements, C++ and Rust. While Rust is promising and getting some traction, saying that it's the clear winner in all aspects would be a stretch. So therefore there's a lot of people who end up with your first statement. And given the kinds of problem where it's the best tool, a lot of those would end up actually interested in the second statement. However, of course most people wouldn't have the opportunity to do so. We're lucky that there are people that end up with all three like Herb Sutter, IMO, or we'd be stuck with 98 semantics... (I guess it would have helped Rust getting popular faster though...)