781 karma · joined June 2, 2011
Would you tell Terrance Tao that they didn't produce [this conversation](https://chatgpt.com/share/6a5fdc7a-d6f8-83e8-bbea-8deb42cfed...), the model did?
Jeff was a ACM Fellow in 2009 and published the massively influential MapReduce paper in 2004.
Documentation errors have always been an issue. They were when there were paper charts, or human transcriptionists, or when manually typing into the EMR, or when using speech recognition (which is AI/ML!) to do the typing for you.
Not all e-scribes use LLMs, but most of them do rely on ambient audio recordings for speech recognition, which nowadays runs entirely locally. That text then needs to be processed into your clinical documentation, and there are tons of ways to do that (including LLM processing).
The author has obviously never talked to clinicians or hospital administrators about the challenges of maintaining clinical documentation, and knows little to nothing about the reality of software that runs in clinical contexts.
Have you seen what that looks like in a hospital system?
Autopilot can be thought of better as "auto-aviate". That is to say, if there is already a navigation plan, the aircraft can follow that plan. Simple autopilots just keep the wings level, others can hold an altitude and change heading. More sophisticated ones can change altitude or even fully land the plane.
All of those things, however, require people to manage the "Navigate" part. "Aviate" is a deterministically solved problem, at least in normal flight operations. As you point out we trust autopilots today, including on (nearly) every single commercial flight.
LLMs are a poor alternative to "aviate", but they could be part of a better flight management automation package. The parent article tries to use the LLM to aviate, with predictable results.
If paired with a capable auto-pilot (not the relatively basic one on that C-172), the LLM could figure out how to operate the FMS and take you from post take-off to final approach and aid in situational awareness.
Currently, I don't think there is a commercial solution for GA aircraft that could say, "Ok, I'm 20NM from KVNY, but there are three people ahead of me in the pattern, so I have to do a right 360 before descending and joining downwind on 34L".
Having an LLM propose that course of action and tell the autopilot to execute on it definitely would be an improvement to GA safety.
You've heard of the saying that extraordinary claims require extraordinary evidence? Holding a hypothesis of water-vapor from air travel being the primary driver of warming trends is extraordinary.
Invoking the oft-repeated "do your own research" rhetorical crutch and referring to scientific consensus as "hype" doesn't help your case.
It's not even worth it to say why or how, since not even doing rudimentary research means that you aren't interested in developing a well-informed opinion.
Maybe we're of a different age to remember when Scott was super influential as a blogger / conference speaker, but even now he's not some random VP.
Mid-day sun in a clear sky is very white, in the 5k-6k color temperature range. It's hard to get a sense of how white it is because of how bright it is. In fact, the color temperature on the surface can be even higher than in outer-space!
Compare this to a "warm" light bulb, which is around 2.5K. Sunrise/sunset is also around that range.
Perhaps the "warm color" sun mindset comes from the only times that people can look directly at it. That is to say, around sunrise or sunset.
The officer chose to engage and close on the vehicle and chose to circle from the front. If the officer was concerned about being run over, they shouldn't have stood right at the bumper. The car was clearly in gear, moving forward was an obvious expectation.
Did the officer have an escape route? Obviously yes, since they only had to side step to avoid the car. Was there an exigent circumstance? No, the officer could have retreated and nobody else was clearly in harm's way. Was the driver clearly a threat? Again, no.
No, this was straight-up murder from a trigger happy psychopath.
I think you conflated flights (several 10Ks per day) with passengers (several million per day).
One in a million flights is one accident every few decades.
> at least in the US. Engines will fail
As per the report, this appears to be a structural failure, not an engine failure.
OP is asking what are the line-drawing characters encoded as e.g: "┌" and "┐".
Since the charset returned by the app is UTF-8, these will be interpreted and encoded as UTF-8 and not whatever "ASCII - Extended" means.
What about something like mdocx?
If that happens once and the restaurant makes good on the mistake, I wouldn't hold it against them.
Sending something back to the kitchen is a way better product analogy than food poisoning.
If you look at various definitions of what facism means, you may see something like: "characterized by severe economic and social regimentation and by forcible suppression of opposition" (from M-W).
A "loyalty rating" implements both economic regimentation (the insinuation that higher scoring companies have better favor) and suppression of opposition (that companies actively avoid being seen as opposition).
So this is text-book fascist behavior.
It's not hyperbole to envision the justice department looking the other way for high-scoring companies, and actively persecuting low-scoring companies. You're right in that this is already happening (like with e.g. Harvard), but implementing a score in the open makes it shockingly easy to carry out fascist directives across the government bureaucracy.
It seems unfair to move to "not recommended" due to a single instance of a hardware failure, especially if the manufacturer made it right. And repair-ability is one of their core values!
At most this should've triggered a "this happened to me, keep an eye out if this seems to be a thing." note in the review instead of moving to not recommended.
"If you have any questions or requests, please mail me at wayne@larsen.st"
I haven't seen another design system that is as comprehensive to material. Express seems like an evolutionary refresh with some things I could use right away, but otherwise most of the content is MD3. It's valuable to me as part of the larger ecosystem.
Counter-point: "That's false because Cruise had an accident for which they were at fault".
OP: "The existence of a case or some cases where a self-driving car caused injury has zero value. What matters is the rate of cases per mile driven."
You: "You do not get to counter-argue."
Yes, they do. OP's point is valid. One can't refute the original assertion by citing one accident by another company. It's a logical fallacy (statistically speaking), and a straw-man (Waymo can't be safe, because other self-driving cars have been found at fault). The validity of the original claim has nothing to do with an invalid counter-claim.
> However, that is still insufficient, even ignoring the lack of audits by non-conflicted parties, to strongly conclude Waymo is safer than a human.
When you have a large, open, peer-reviewed body of evidence, then yes, that's exactly what you get to claim. To reject those claims because Waymo was involved is ad-hominem. It's not how science works. It's not how safety regulations or government oversight works. If you think it's insufficient, you can attack their body of work, but you don't get to reject the claim because they haven't met some unspecific and imaginary burden of proof.
Waymo releases its safety data: https://waymo.com/safety/impact/, which is backed by public reporting requirements.
To say that it is wholly insufficient to make any safety claims on publicly driven 50M miles, is ridiculous. At the very least, it appears sound, robust and transparent, and able to be validated.
> https://waymo.com/blog/2024/12/new-swiss-re-study-waymo
Is Swiss Re a valid third party? They also address peer-reviewed and external validation in the above safety impact page.
I can understand being skeptical because of Cruise and especially claims made by Telsa, but there is a preponderance of supporting data for Waymo.
Given all of this evidence, you would still conclude Waymo is unsafe?
Historically speaking, if you had a 15% word error rate in speech recognition, it would generally be considered useful. 7% would be performing well, and <5% would be near the top of the market.
Typically, your error rate just needs to be below the usefulness threshold and in many cases the cost of errors is pretty small.
That's ok! Hint at it and let people discover it instead of trying to force them from the get-go. Utilize progressive complexity; start simple, from first principles, and add complexity in bite-sized chunks. Show, don't tell.
No one wants to have to learn an entire philosophy before they can start using a tool.
For inspiration perhaps review how other very deep programs represent themselves for example orgmode.org. One caution there is that orgmode itself is famously obtuse for beginners.
Lastly, it is a bold statement to say something like "we have discovered a new development methodology, and have designed this toolkit around that philosophy".
Such a statement requires a ton of evidence that such a methodology is useful, and currently there simply is not enough.
> "Glamorous Toolkit is the Moldable Development Environment"
So it's some sort of an IDE? What does moldable mean?
> "Make systems explainable through contextual micro tools"
What is a "system" in the context of an IDE? "Contextual micro tools" also sounds completely abstract.
> "Each problem about your system is special. And each problem can be explained through contextual development experiences. Glamorous Toolkit enables you to build such experiences out of micro tools. Thousands of them ... per system. It's called Moldable Development."
... this does not help at all. Just more words without meaning.
Next, there's the video. For somebody with zero context so far, why would they sit through a 46 min low quality video?
tudorgirba - if this is your project, you really need to focus on getting the top half of the page right. People won't watch your video, no one will read your book if you can't give them a hook they understand.
Use words and phrases with concrete and well understood meaning with adjectives:
* Don't say "micro tool". Like Posix utilities? What is a tool? What makes it micro?
* Don't say "contextual development". Isn't all development contextual?
* "moldable" - no one knows what this means, don't force them to try and figure it out.
* Don't say "system", it is too abstract.
What about all of the domestic producers of finished goods who can now can not afford their materials and supplies? What about the domestic supplier of that finished goods producer who has their orders cut because the finished goods producer needs to cut production? What about the domestic consumer who'd like to buy those finished goods, but they can't afford it because prices over all have increased? What about the barber shop by the finished goods plant that has to shut down because the finished goods plant cut their workforce by 50%?
100+ years of studying tariffs have shown that they are effective when very narrowly targeted. Otherwise they almost never achieve their stated goal of actually increasing domestic production.
Should we tax goods from Mississippi to NY to save jobs in NYC? After all, wages in NYC are way higher than in MS. Trade with MS pulls down wages with NYC.
If no other team sees an advantage from using torpedo bats, it would be a lot like the brotherly shove.
But first we'll have to see if this is a passing fad. In baseball, pitchers evolve pretty quickly and usually lead the batter-pitcher arms race.
I'm guessing it spread pretty quickly through the league and be used by a minority of hitters, and the advantage will flatten out. So a .210 hitter may hit .230. That is a big difference no doubt, but compare the game to when leading batters were hitting .330.