HNHacker News
TopNewBestAskShowJobs

AlanSE

547 karma · joined February 6, 2014

submissionscomments
AlanSE··on Avoiding hallucinations in LLM-powered applications
Funny enough, in 2023, the option most likely to be viable for the impossible task of categorizing the training set would be to use a LLM. While I don't "trust" these AIs to give accurate information, it's probably within their capabilities to categorize by the above mentioned categories... then feed that back into another (very expensive) round of training, along with some theoretical developments to boot. I do think this is within the realm of possibility in the next ~1 year, but would be hard.
AlanSE··on Almost three quarters of the golden age of Hollywood has been lost
Perhaps non-human entities will lament that. Or perhaps humans who are leveraging more advanced search/discovery tools to look for specific things.

Otherwise, there is too much content for anyone to view.

AlanSE··on Vast new stores of water reported on the moon
The problem with asteroids is the population dynamics and time required with these heliocentric orbits. Yes, there are asteroids at a lower Delta-V than the surface of the moon. There might be only a handful that are even worth capturing and mining. If this scaled up, however, you would quickly deplete this population of NEOs.

While the Delta-V might be less, it's not less by MUCH because you spend most of your fuel getting to Earth's escape velocity in the first place. The Delta-V benefit is also balanced against a massive time lag involved in sending a robotic spacecraft to capture and boost it to an orbit where we can better access and refine it. Or you send the refinery to the asteroid for larger asteroids. Problem there - you'll basically have to build a new refinery for each asteroid. Even for larger NEOs at higher Delta-Vs, the mass is not all that much. It's also not clear that microgravity will be any better than 1/6th gravity.

There are very strong arguments for Lunar mining. We're talking about ~1 week for transit time from Earth to there. The moon can build vast Earth-like industry. You get to cherry-pick a wide variety of asteroid material types (except for volatiles) without building a new factory for each. Surface transportation isn't particularly hard. Materials to build solar panels are widely available.

Asteroids will be useful at some point, but only after we can reach the frost line. Otherwise, the asteroids we get will be parched and not offer much beyond what's on the moon, or what we can lift from Earth. After we get to the frost line, we will get vast troves of organic material, but this is at tremendous Delta-V cost and outrageously remote and time consuming to access.

AlanSE··on Give babies peanut butter to cut peanut allergies, study says
From 2005 until now, rates of peanut allergies have increased 2-fold or more in the US. Similar for tree nuts. Anecdotally, it feels like more. There's a huge generational shift even among recent parents, where it feels like a problem just didn't have very long ago, but we do now.

It's true that we had bad advice to not introduce peanuts which provably made the problem worse (and has now been reversed), but this is only part of the problem. Papers on the subject make clear that are other countries that are not doing intentional early-introduction (which we are now in the US) and don't have the problem.

It's also abundantly clear from the research that the population-scale problem is 100% environmental... as you would expect since the SAME POPULATION saw rates go through the roof in a decade. It's something we did, environmentally, and we don't know what.

It also feels frustrating that the medical profession seems overwhelmingly incurious about discovering whatever it is we're doing that is literally harming children.

AlanSE··on Give babies peanut butter to cut peanut allergies, study says
I'm in the same boat, and it's true, even doctors don't get it. Gone through several pediatricians and got a blank stare back, mouth-agape.

Children are getting older, and thankfully, are no afflicted with the same allergy as their mother. But as time wears on, it is truly emotionally draining. The nuts held in a hermetically sealed container are compared to a caged viper - as they can easily cause life-threatening anaphylaxis for exactly 1 person in the household. We have to follow a super-clean process outside-and-inside to give an introduction, but you know what, kids don't always want to do what they're told. It sucks, and it's a problem more and more people will have in the future.

AlanSE··on NIO reveals aggressive plan to add 1,000 swap stations in 2023
Some form of this thought is irresistible, but we shouldn't hesitate to point out the obvious flaws. You load up your car with more people and more stuff when you go on a trip, and more batteries would add MORE load on the frame. This doesn't work. Ideally, you would add another axle.

Suspend disbelief and follow that thought to its logical conclusion.

You have almost certainly PLANNED your trip. You know what time you will depart, and the trip may take 4 to 10 hours, done in a single day. Additionally, you will spend 30 minutes packing your car, and more space is a premium. Keep following me, just a little bit longer..

The day before you depart at, say 3pm, a worker from an established car rental service pulls up in front of your house and parks on your street. They unhitch a trailer from their car, do some technical checks, and are gone in 5 minutes.

You arrive home from work, and have a trailer that will add 400 miles of range to your car and offer extra packing space (for things). You pack up the trailer & car and go to bed. In the morning, as you pull out, you get out and hook up the trailer.

Depending on your trip, you may charge up the car & trailer once, or not at all. I'm guessing that it will be able to connect two chargers at once.

This seems like not-a-bad-idea, and compatible with current infrastructure. While it incurs a cost, and requires EVs to be built for the purpose, it's a much more minor change than battery swapping. The cost could be quite small compared to the other solutions for the same problem, and it frankly sounds more direct.

AlanSE··on I changed my mind about nuclear waste
If you take the climate science seriously, then you really must accept net-zero by 2050. After that, maybe net-negative.

Assuming you're an educated person, the question we must ask is if natural gas is viable to fill in the gaps from renewables in 2050. Given the net-zero requirement, that means a natural gas plant that puts its CO2 emissions back in the ground. That can make sense with direct carbon-capture and storage, but the problem is high capital costs totally undermine the low capacity factor which is why we use it today.

If we go natural gas CCS vs nuclear in 2050, maybe it's a tossup, but they're BOTH poorly suited for the job, if we're talking about solar at >50%. Daily storage will be solved fairly well by then, and seasonal energy storage will be the challenge. But nuclear is baseload.

There are interesting ideas like running nuclear to heat a vat of salt which is used as a means of energy storage so that it would work in a grid with lots of renewables. But that's peak smoothing over several days. Seasonal thermal storage has been written about but is basically crackpot stuff.

So I've kind of come around to seasonal storage of green hydrogen. The technology is at least focused on the right problem. Duration of storage and energy density arguments favor chemical energy, and low capital costs per nameplate capacity are the primary driver.

AlanSE··on I changed my mind about nuclear waste
I don't know if I agree with the "organized civilian protestation" entirely. While NIMBY did play a role historically, this seems out-of-touch when I consider the mothballed & troubled builds which are currently in South Carolina.

I do believe regulators have their foot on the brakes, and there are some really telling stories from people in the actual industry. The level of paperwork and scrutiny are hard to fathom for people in other industries. But right now, in the US, that is more of a Federal political issue. The AP1000 lost crucial time due to added missile shield requirements. That doesn't scream local on-the-ground public resistance. You can pin some of this nearly directly to anti-nuclear views from (let's be honestly) Democratic senators (not from the south). Federal policy both encourages and discourages nuclear, but the domestic industry was too weak, and the regulatory policy took critical hits during a formative time.

I care very deeply about addressing carbon emissions. I want nuclear to work, but that's not an argumentative hill I should die on when solar has seen costs go bananas. Maybe another human generation will change things, but from lived experience, I would bet not.

As an engineer, I admit that either nuclear or solar COULD HAVE become the backbone of our energy system. But time is out for climate action. Solar is modular, proven, and roaring. Further deployment will create a new engineering problem of storage, but we don't have time to turn that into a delay tactic. There will be solutions to energy storage, just as there are solutions to nuclear waste. Neither are cheap, but either are a drop in the bucket compared to the climate damage we face.

AlanSE··on Photoprism – open-source Google Photos Alternative
That's awesome! Thanks for sharing. I had a go at all the contenders a while ago when Google Photos changed their storage policy and did not see that one.

I see a sharing menu in one of their screenshots. Considering that it's self-hosted, how does sharing work in that system?

AlanSE··on Heat pumps of the 1800s are becoming the technology of the future
The fridge isn't the biggest energy hog in the house, but I'm very sympathetic to the absurdity of heat levels in a house.

I often see my HVAC cooling when the set-point temperature is actually _higher_ than the outside temperature. Logically, the house is a heat generator, it makes sense physically. The roof is black, etc.

It would offer a good number of benefits if the system could outright open a duct to the outdoor air, and suck it in whenever the local outside temperature is within the range requested by the user. People who are into optimizing energy use (they exist) can go even further and pre-cool their house during the night in summer.

For this to work, all you need a pusher fan, no refrigeration at all. There might be some pressurization problems, like, you may need a duct both for the intake and outlet. Also might require another filter... but air quality would improve significantly.

This is a really "dumb" idea, but it's perfectly in-line with all the new ideas being thrown out there. The new ideas just tend to throw in an additional heat storage mechanism, like a water tank (in the article). You can get a lot more efficiency gains by saving the night's cold in a tank and using it through the day. But on a more basic level, you can pump straight into the house when the conditions are right.

AlanSE··on Eggs are 60% more expensive than last year in the US
Inflation is always permanent.
AlanSE··on The CPSC is on a mission that has nothing to do with safety
The corrected numbers still look like the 3%-5% range, which is still probably enough of a reason to switch stove types... if all the other reasons weren't enough already.
AlanSE··on Antidepressants are over-prescribed, but genuinely help some patients
I'm picking up a theme about the US health care system - that doctors respond to conditions caused by environment with a drug.

Our lifestyle (activity) is unhealthy, our food is unhealthy, our houses are nice but we don't get out of them, our relationships are unhealthy, even our culture creates anxiety. I speak for myself, but also everyone I know.

AlanSE··on Names should be cute, not descriptive
This reminds me (from a long time ago) of seeing IT department stick names of jungle animals on computers so they can be recognized on the network.

These days, there are many cases where _instances_ of something are given random "cute" names. A particular server will have a suitable-for-work but fun name slapped on, like "grumpy goat".

Hover, I can't imagine _classes_ of something following that kind of naming convention. This post seems to be more about _queue_ names or _service_ names. Software DOES intentionally use cute and random names for big-ticket items (see "git"). It would be confusing to apply this practice to more minor software, because it would be too many to remember.

If I have a service as a part of an app that consumes data from a redis queue and sends it to a log collector, a fully descriptive name is distasteful, but I wouldn't want to name it "anteater" because even if my mental picture is vivid, other people will... not get it. I'd call it "log-passer" or something.

AlanSE··on The Fermi paradox revisited: Technosignatures and the contact era
Yes, this is true but only from the reference frame of the cosmic background radiation. It is still possible that you can get O(n^3) from the reference frame of the aliens themselves, because time dilation makes travel fairly fast. You get into some spacetime philosophy about what metrics are worth evaluating O against in the first place.
AlanSE··on January 1, 2023 is Public Domain Day: Works from 1927 are open to all
Those dates will be just in time for people to read the earliest wave cli-fi books as they're living through it!
AlanSE··on The Fermi paradox revisited: Technosignatures and the contact era
That argument seems to violate the basic rules of exponentials. If an advanced species did settle on an stasis, then that will only last until some subset of the population decides to go beyond the stasis. Particularly for a space-faring civilization, the time scales and resources are so fast, that any multiplication factor beyond unity will eventually go bananas. Any individuals with the prerogative to multiply will swamp those who lack it, same with Earth evolution. The selfish gene will prevail unless there is centralized control.

I don't have any problem with this reading - advanced space-fairing beings have never existed in our light cone. SETI won't find anything because there's nothing there. We have a good chance of making it ourselves, but it's 50/50 because we have no priors, our decisions actually matter.

AlanSE··on Nuclear physicist explains why fusion ignition is hailed as a major breakthrough
Pacer took thinking about nuclear fuels in a lot of new and interesting directions. For the use case of interest, they really wanted to maximize the fusion yield fraction, minimize the radioactive products, and minimize the total yield. Had it ever been allowed to proceed to a commercial scale, these would have been very tight, very challenging, technical constraints.

If you relax one of those requirements, then it becomes instantly clear that limitless cheap energy is possible. For a somewhat absurd example, imagine that we build Pacer plants in space instead of on Earth. There's no longer any size constraint on the blast chamber. You can go to triple-stage (like Tzar Bomba) or more designs, at which point the fusion fraction gets extremely close to 100%, and the cost per unit energy becomes somewhat hard to fathom. But on Earth, it still would have been expensive.

AlanSE··on US Department of Energy: Fusion Ignition Achieved
This seems to say to me that D-T reactions produce neutrons, and that the kinetic energy of the neutrons is smaller than what you get by hitting U with that neutron. You already have the energy from the neutron (which will land somewhere in the system eventually), and you might as well get a multiplier by putting a blanket of U-238 in front of it.

That could be carbon-copied to a fusion power plant, and indeed, there are many proposals of hybrid fusion-fission plants in the literature that only require Q values marginally greater than 1. But if you go that route, you have radiation just like a fission plant, and one starts to question why you don't just build a fission plant (indeed, why don't we?).

My personal pet theory of the future is that, one day, we'll progress so far in fusion research that we get economic energy. But at the same time, the line blurs between both fission and weapons technology, so people are unhappy with the result. This doesn't feel particularly contrarian but no one ever seems to bring it up.

AlanSE··on Secretary Granholm to announce major scientific breakthrough by DOE [video]
Huh, ok. After some quick Googling, Wikipedia paints JET has having tried, and failed, to reach Q=1, only getting 0.67, which was way back in 1997.

It's very strange to read that we get Q=0.67 in 1997 in a tokomak, and then Q=1.5 in 2022 with a totally different laser inertial confinement design. That... seems to be the written history.

I know Iter would be surpassing the record, but the timelines are long to the point that one loses hope in the project.

AlanSE··on Opening the North American charging standard
My thoughts as well, this is their business advantage to lose.

> NACS is the most common charging standard in North America: NACS vehicles outnumber CCS two-to-one, and Tesla's Supercharging network has 60% more NACS posts than all the CCS-equipped networks combined.

Yet, in your link it's complained about that the Supercharging network capacity is already strained. Bottom line is that the vast majority of the charging stations that will exist 10 years from now are still yet to be built.

Tesla may still be hoping for a regional break, as their standard wins in some geographies but not for others. This would be a loss for all of us all collectively.

Certainly feels like they should have known better. Past greed will leave them (and their customers) with a massive writeoff in the future if your prediction comes to pass.

AlanSE··on Sugar Substitutes Surprise
I think the idea is that we would keep them in a lab and feed them whatever works.
AlanSE··on How to communicate effectively as a developer
> If I get sent any messages like those examples on the left it actually upsets me.

I spend lots of time responding to issues and informal messages like that. Processing vague reports like this is nearly a daily task for me. The examples there cut deep for me.

Writing well just isn't natural. Mentally, you're on one level of the stack, and it takes work to back it out again. My forever case-in-point about this is log messages. Log messages are, by default, written terribly because the reader is not kept in mind. The log will be encountered with little context (maybe a timestamp) in the middle of a 5 MB file. If you are inside of a nested loop, your message formatting _needs to include both loop variables_, and when those are objects, there's a very non-obvious decision about what data from them is relevant to the log.

It's the same in human communication. Conversations do have context, yes. You can and should relax the rules of context as you get the ball rolling in a 1-on-1 chat with someone, but with many people in a room, or when starting a conversation, it's so important.

AlanSE··on Don’t teach kids to fear the world
> suboptimal but creative solution I saw recently was neighbors working together to remove their backyard fences to create a large and safe play area for children in the suburbs.

This is something I would love to do, but still feels so impossible. I see some homes compromise and put gates in the fence. Where I am, I see probably 70%+ of homes with backyards using them as bathrooms for their dogs, and I imagine that many people would have a problem just with the basic social contract of who cleans dog poop.

Particularly due to COVID, I've begun to appreciate just how severely the suburban structure negatively impacts all aspects of our lives. You have HOAs that invest in common playgrounds, cities that have invested in parks, and virtually everyone reports extremely high satisfaction with these things. Yet my own feelings of isolation tells me that we need much much more. I'm at the point that I would seriously seek out and move to a place with a walk-able kid-friendly setup. But it just doesn't exist.

AlanSE··on Nuclear is back on the table for a green future
Small reactors are not expected to be any cheaper than traditional reactors. They are a better bet right now because the traditional reactors keep getting mothballed in developed nations, and they have some clever technical innovations, but these are not complete game changers.

Each decade, nuclear builds also trend towards getting more expensive. Like it or not, it at least partially Fukushima Daiichi that killed the nuclear expansion US. The AP1000 had new missile shield requirements that was slowing it down before that happened, and then the Japan accident added even more regulation. There was an anti-nuclear chair of the regulatory commission at the time, and yes, this is fundamentally political, but it's hard to push back on these specifics.

AlanSE··on After millennia of expansion, the world has passed ‘peak agricultural land’
https://thehill.com/policy/equilibrium-sustainability/350461...

An interesting quote which echos what you're saying here:

> That means that table corn, carrots, cabbage, apples and broccoli are all examples of what is called, paradoxically, “specialty crops” — and they largely don’t get the same support as the heavily subsidized industrial crops.

The absurdity here is that "specialty" crops are what you might call "food". Most corn is "what my food eats".

If survival was important, people could switch trivially. I would take a 99% vegetarian diet in exchange for not dying. Problem is that the costs may be collectivized, and weighted towards future generations.

AlanSE··on 3,599,040,600 € wired by Germany to Russia since the invasion of Ukraine
As a young adult during the Bush years, I felt strongly that we needed to take aggressive action to replace foreign oil dependence with new energy sources (renewables, nuclear, whatever, sign me up). Politically, I saw both a nationalistic reason for it, as well as the obvious environmental reasons. Paul Friedman's "Real Men Tax Gas" seemed like the right angle that cut across the ENTIRE political spectrum. Really, who can disagree with this?

But I could offer a concession, at the time, that the solutions were not off-the-shelf available.

Now? The bottom has fell out of the "we can't" argument. Solar just kept going down in price along with batteries. Anyone defending coal power is hardly doing so on the price argument. A solar-based economy is not familiar, no. We will have to build infrastructure that turns on and off daily and seasonally in response to electricity supply-and-demand balance. There are large engineering changes needed... but no cultural ones.

There's no day like today. The energy transition will benefit our friends and crush our enemies.

AlanSE··on Launch HN: AirMyne (YC W22) – Capturing CO2 from air at industrial scale
> Forests only sequester carbon as they grow. After that they are carbon neutral. You end up with land that cannot be used for any economic purpose. (The creatures and plants that live in it have a value too, but that is not part of this argument).

I will be paraphrasing the science articles I have read lately...

Planting new trees on bare land does not work to capture CO2. So if you deforest an area and then re-plant, you put a lot of extra CO2 into the atmosphere. Like 30% of the trees you harvest go into durable things (like houses) and the rest will decompose and release its carbon. After you leave the ground bare, it starts spewing CO2 from the soil. This is a major carbon source. New trees you plant will EVENTUALLY soak up more CO2, but canopy closure needs to happen before that can happen. Since the young trees are planted with considerable spacing, the soil CO2 source outpaces the tree CO2 sink for many years before the balance shifts.

Mature forests throughout history probably did tend to be carbon neutral on average, yes. This is ignoring ecological changes... like... maybe forest conversion to other landscapes and fires were balanced by healthy forest uptake. I digress.

These are not normal times, and CO2 concentration is very high. Because of that, old-growth or mature forests may be significant sinks of CO2. The best strategy for us to draw down more with forests is to leave as much mature forest untouched as we can.

AlanSE··on Veloren is a multiplayer voxel RPG written in Rust
I played a bit into this game (maybe about 20 hours), and it's good. The size of the map dominates the game-play as you aim to explore all the types of destinations. It's a procedurally generated world, and some hilly towns will get a little weird. The glider is really a killer features, because the mountains are really really steep, you climb up to the top and the glide to cover a fairly large distance on the map.

It was still hard to get into the social element of the game, because of so few players and such a large map. I think other players tended to be concentrated at the main spawn point, so a lot of the rest of the world was a backwater essentially.

They also have a giant tree. It will cost you some hours, but it's quite different than anything else you'll find.

AlanSE··on Times are great for programmers now. How does it end?
I feel like this has already been happening for years... probably longer than that. I find a lot of the people I work with were pulled away from other industries. Some industries are over-represented, like science, engineering, education.
← PreviousPage 2 of 8Next →