Hey, an area I actually work on and use in production! MAB in our case has shown improvements in our A/B metrics. My general takeaways for my experience with MAB:
* we have a single optimization goal (e.g the metric to minimize or maximize). The hard part isnt defining a optimization goal. The hard part is identifying what stakeholders actually want the tradeoff between different metrics. If you have goal A and goal B, where B is more aggressive than A, then getting agreement on which is better is hard. This is a people problem.
* MAB seems to be a good proof of concept that something that can be optimized but isnt an "end game" optimization path.
* MAB for A/B testing is going to mess up your AB data and make everything more difficult. You should have a treatment that uses the MAB algorithm and a treatment that doesnt.
All of the above is for non contextual MAB. I am currently learning about different MAB algorithms although each of them are pretty similar. The ones ive read about are all effectively linear/logistic regression and tbe differences come from exploration mechanism and how uncertainty is represented. Epsilon greedy has no uncertainty, exploration just happens a fixed percentage of time. UCB is optimistic about uncertainty, and the amount of optimism controls exploration. Thompson sampling uses statistical (beta) distributions about uncertainty, exploration happens less as confidence about a particular set of options increases.
Overall its a fun area to work in that is quite different from your typical CRUD development which is a nice change of pace.
No public write ups. What information would you want to see? I think there's two categories: infra ops info, and then product insights and "gotchas"/unintuitive but valid results.
It's complicated. Tsunami forecasting is a very inexact science and "3m" means "very large".
The average actual height in eastern Japan (Tohoku) was 4-6m, but there were peaks up to 20m in places like Ofunato where the local geography funneled all the water upwards.
I'm surprised so many people don't understand what tsunamis are. It's a "wave" created by a sudden shift in the Earth's crust. Imagine, suddenly, water on each of side of that split is now at different heights and has to equalize. It's much closer to just removing a dam that is holding back water equal in height to the new difference between the sea floors.
> What you get is not a "wave" but a wall of water.
Its a wave (or series of waves) with a large wavelength and speed in deep ocean, that becomes a shorter wavelength and very large amplitude by shoaling as it hits shallow water.
Its different from typical wind-driven ocean waves for a lot of reasons; but a big indicator is wavelength -- wind-driven ocean waves have wavelengths up to hundreds of meters, tsunamis have wavelengths (in deep ocean) of hundreds of kilometers.
More like tides than waves, as has been stated elsewhere in the thread, is both technically wrong but substantively (with the caveat that "waves" really means "typical wind-drive waves") correct, in that tides are also manifested through waves, but waves which have wavelengths of thousands of kilometers, and so tsunamis are waves more similar to those making up tides (hence the old colloquial use of "tidal waves", which properly refers to the waves manifesting tides, to refer to tsunamis) than to wind-driven waves.
Not true. As the news reporters here in Japan are repeating every few minutes, there will be many waves and they can get bigger over time. They already have, 20-30cm initial waves had 40-60cm later waves.
Waves can get bigger due to earthquakes not being instantaneous or necessarily a single movement, due to amplification by geography, by reflections, by aftershocks, and many other things. The news is suggesting waves lasted about a day for a previous event in a similar area.
> I'm surprised so many people don't understand what tsunamis are.
“I’m
Surprised so many people don’t know what ‘X’ is/are isn’t a very nice thing to say. Your comment could have done without that, the rest of it would have been fine.
I don’t take offence. I’m not the most educated, and I don’t live in or near a tsunami prone area, I know about other natural disasters that are relevant to where I live though, maybe more than the parent poster.
I don't think he's even right. Like what he is saying is in actuality wrong. He's surprised because he's ignorant. I'm all for people saying stuff the way he says it. He believes it's true, then he should stand behind. But then the consequence is that he needs to be accepting of when people call him out for being utterly wrong.
The difference is that people know what 2m (wind driven) waves look like at their cities seawall. A 2m tsunami is a -completely- different phenomenon, because of its length. Depending upon the underwater geography, a 2m tsunami might flood right over their 3m seawall, and wipe out entire parts of the city, sweeping hundreds of people out to sea. A 2m wind wave will get saltwater spray on cars driving by. They are both waves, but they share very little in characteristics other than their fundamental physics. It’s like saying that a slingshot fires a 12mm projectile, and so does a 50 caliber anti material rifle. The fact that they are both projectiles, of the same size, is much, much less informative than other facts about their nature.
Saying that tsunamis are waves is easy to equivocate into tsunamis are waves, like other waves. This is an equivocation that is very misleading and can get people killed.
Insofar as the goal of communication is to communicate meaningful information, it is less accurate to say “tsunamis are waves” than it is to say “tsunamis are nothing like normal waves”, or to say “tsunamis are like a wall of water, not like a wave” or “tsunamis are more like tides than waves”.
So yes, tsunamis are waves, but insisting that tsunamis are waves without qualification that their effective characteristics are fundamentally much different and more dangerous than a regular wave is misleading through omission in a way that could directly put people’s lives in jeopardy.
Being pedantic about definitions and being accurate in conveying meaning are not the same thing, and communicating in good faith normally is about conveying meaning in an accurate manner, not just using words in an accurate manner.
FWIW I also believe that meanings are important, but there is a point where pedantry falls into bad-faith territory.
I appreciate your effort to provide an understandable explanation.
That said, in context the original statement was so extremely misrepresentative of the reality that I felt it left the realm of "inaccurate but effective for communication". I certainly didn't see the objections as pedantic.
A clarification was appropriate because it really did miss the physics, but doubling down on the definition of wave without talking about speed, length, and volume (which is what had confused OP in the first place) was not only suboptimal in teaching useful knowledge to OP, it was also misleading in a way that could (and did, in at least one case of a commenter in this thread) lead to a dangerous misconception about characterising tsunamis.
Perhaps it wasn’t intentionally pedantic, but the way that it was doubled down on later makes me suspect an argument in bad faith, or at least an epic case of missing the opportunity to usefully inform.
I value this site for the general character of people trying to educate more than just troll, and I think it’s important to try to educate trolls as well to understand a more constructive and respectful way to interact here. Ostensibly, we take off our clown shoes and leave them at the door.
OTOH I may have read multiple comments in similar tone that were not all attributed to one poster , giving me a mistaken impression of the intent. In that case, I owe an apology for perhaps overreacting.
Nitpick: while what you say is generally true, there are several scenarios that can create true dramatic “wall of water” tsunami waves that have leading slopes of 45-90 degrees and heights in the tens of meters.
The most obvious (but relatively rare) are tsunamis amplified by submarine canyons and other coastal bathymetry like the Nazare submarine canyon famous for the biggest waves on the planet (50+ footers are common in season). If an earthquake directs a tsunami at that canyon, the resulting waves will be spectacular and probably drown everything north of the cliffs. Unfortunately we don’t have any historical records about what happened at Nazare after the 1755 Lisbon earthquake so we don’t know just how big those waves can get.
Then there’s landslides like the one that caused the 1958 tsunami in Lituya Bay, Alaska which creates a much more sudden displacement than an earthquake. Based on the surrounding mountainsides the wave created from that landslide might have peaked at ~500 meters without the 100+ mile wavelength you’d see in a normal tsunami wave.
The most common however are tidal bores, which can send a 30+ foot vertical wave down rivers and narrow channels. This phenomenon shows up relatively frequently in earthquake youtube videos near rivers, though the wall is usually only 5-10 ft tall.
Oh yes! It’s absolutely true that underwater geography can steepen the wave front and amplify the height, sometimes by orders of magnitude. The deep water height of a tsunami wave is often a lot different from what you will see at the coastline.
An entertaining anecdote from the pre- smartphone era:
I sailed to the site near Chenega bay where the earthquake wiped out the village in 1958. We got permission from the elders at the Chenega bay village to land at the island, and it was extremely humbling to see the high water mark from the coastline, and to see the wreckage of boats far, far up on mountainsides.
I’m not a big believer in supernatural stuff, and there are plenty of alternative explanations, but it still freaks me out a bit that the photos we took (aside from the digital ones) did not develop any images of the village site. It was white as if it had been overexposed, even in the case of 1/2 frames. On both disposable cameras. Other photos from the same day, taken in other directions, turned out fine. The digital camera fell overboard in 500 fathoms, so we lost those photos the next day.
As for how exactly this could happen in any reasonable version of events, I’ve got nothing. I guess sometimes chance events line up just right to make for a good story.
Interestingly, there are big tidal bores frequently in Turnagain arm, with 1-3m being common. I’ve seen people surfing it with wetsuits in the ice cold water, getting a run of several miles lol.
> Interestingly, there are big tidal bores frequently in Turnagain arm, with 1-3m being common. I’ve seen people surfing it with wetsuits in the ice cold water, getting a run of several miles lol.
That’s going on the bucket list! How dangerous is it? If you biff it and get caught by the wave can you just dive under the wavefront and come up behind it like on a beach?
I don’t know much about the dangers involved, but it isn’t (or wasn’t, AFAIK) a big attraction… so maybe it’s crazy or dangerous? I’m not sure. Or it might be because the water is so cold that hypothermia and death is inevitable in less than an hour? Or because there probably is no realistic chance of rescue if something goes wrong?
At any rate, there was enough people willing to try that I saw it a couple of times, and TBF I think windsurfing isn’t too unusual in the arm now. But there is also trees and stuff in the tidal bores sometimes, so maybe debris is a big problem or the area is too shallow to be safe.
But, if it’s doable, I’m sure it would be memorable!
I think you’re out of touch. A tsunami is a wave both from a pedantic perspective and an intuitive one and most people aren’t deceived into thinking that tsunamis aren’t dangerous at all because it’s a wave. That’s just made up garbage.
You’re like coming up to me and saying hurricane is not wind because it’s dangerous to think of a hurricane as only wind.
Dude. Nobody is thinking hurricanes are just chill just because hurricanes are wind. This is a fucking non-issue.
I think what you’re trying to say is that the wave length of a tsunami is much longer than the amplitude even though the amplitude is still epically high. But don’t try to conflate this with a safety issue of people dying because somebody called it a “wave” that’s just garbage.
So, I’m sorry that I evidently didn’t manage to convey my point effectively. The problem is that wave is accompanied by a measurement that deceptively buries the lede.
Everyone knows hurricanes are wind. So they look for the wind speed to understand the threat. And it’s effective at characterizing the threat. A 100mph wind is going to be similarly destructive as any other 100mph wind. It works and is semantically and linguistically accurate.
Everyone knows a tsunami is a wave, and it is a strong intuition to believe that a wave is defined by its height. , and the height of the tsunami is actually one of the most widely reported metrics. But intuition about the effects of a tsunami by wave height is dangerously wrong. A tsunami is not at all similar to the vast, vast majority of waves in character and effect. Its speed and length at way, way out of band, and are seldom reported.
My understanding is the most deadly/destructive parts of hurricanes are usually:
1. the storm surge, the potential wall of water brought by the continuous winds and waves near the shore, followed by
2. the flooding from heavy rains, then
3. followed by the wind.
So your example might also be hitting the same issue you're trying to avoid.
Note, the worst storm surge is from the eye towards the side where the winds are blowing in the direction of the shore. That's only part of the area with the peak winds.
Good points. Where I am at it’s mostly the wind because I am a well drained higher elevation, so I’m sure that coloured my perception. But you are right, the storm surge and flooding also do a great deal of damage.
Idk. I don't live anywhere where tsunamis are an issue but seeing measurements like a 1m wave does make me wonder about waves I see at the beach that are that high regularly. I find myself going "oh not so bad then" only to read about thousands of people being evacuated and major damage
This doesn't make sense to me intuitively. It must be a wave.
Imagine you have a fault line. There is a left side and a right side to the fault line. If the left side lowers with a shift then that shift MUST be localized to the area around the fault. Because if it wasn't then that means there's an elevation change across the board for everything to the left of the fault. You see how that doesn't make sense? So if the entire country of japan was on the left side of the fault then the entire country of japan shifts in elevation which is unrealistic.
So that means, if what you say is semi-true then the shift in elevation is localized to the area left along the fault but the elevation further left remains the same. It's like a slight dip or bump along the fault line. It must be like this because the alternative is just unrealistic. This MUST be what happens when tectonic plates "shift". You won't see the ENTIRE plate shifting in elevation.
With naive logic, one would think that the water simply fills the localized gap but given how deep the ocean is relative to the actual shift way down in the abyss I'm betting if you were on a boat on top of the fault you wouldn't notice anything. But the movement does create a slight imperceptible "filling" that you don't notice. This is a "wave" but it's invisible.
The wave will translate leftward if the movement of the "shift" was sort of in that direction, but you don't see it. BUT as the sea floor gets nearer and nearer to the surface of the ocean the energy of the wave gets squuezed into less and less ocean water mass (i'm remembering how tsunamis work now) and THEN it becomes visible. Right? Just imagine a sideways cross section. As the tiny wave travels from big ocean with huge depth to coastline with no depth the energy of the wave gets concentrated into a thinner and thinner layer of water.
My intuition just sort of converged with my obscure memory of how tsunamis work so I'm pretty sure this is what's going on.
So it is indeed a "wave" that is acting on wave like phenomena beyond simply "filling a gap". In fact say there's an elevation lowering on the left side of the fault by 1 meter. The resulting wave on the coast line hundreds of miles away will be a wave that extends upward by MORE then 1 meter above sea level which is the opposite of water "filling up a gap." That's totally a wave.
Additionally water from tsunamis always recede. This wouldn't happen if the "wall of water" was simply equalizing. If that's the case the water would never recede.
Any expert who says otherwise, let me know.
edit: Actually why the fuck am I using my intuition to explain it? Just cite a source:
tsunamis are 100% waves as explained in the link. Anyone who says otherwise clearly doesn't know what they are talking about, that includes the person I'm responding to. End of story.
Yes, they are waves, but they are often very long waves. A typical 1m wave might be 20m long. A tsunami wave might be a kilometer long or longer. That is why people say they are like a tide. The wave arrives, then does not recede for several minutes. So, while a 4m wind driven wave might break over a seawall and even wash a car off the road, a 4m tsunami washes ships over that same seawall and floods the city.
It’s a wave, but it is often not at all like a regular ocean wave. I’ve been at sea when a 3m tsunami passed, we barely felt it. If it had been a 3m wind wave in that otherwise calm sea, it would have knocked dinner off the table.
We were about 50 miles offshore, off the continental shelf (in very deep water) we got the information of the wave from our regular meteorological diligence, since it was my job to get our satellite weather and any notices to mariners on a 6 hour rotation.
I saw the wave on radar first, since it lifted ships that were below our horizon up to where they could be seen again for a few sweeps. But it just felt like A gentle lifting. I didn’t even feel the subsidence of the wave. Interestingly, ships 20 miles away from us but near the edge of the shelf reported isolated severe and chaotic waves.
Here's a video of what it looks like from the 2011 event, from the POV of the coast guard approaching it. Waves don't typically look like a sheet has been flapped across one front of the entire horizon of what is visible on the ocean
Yeah that's a wave bro. Notice how the ocean rises above it's own typical sea level? That's not water "filling in a gap" the way tides do it as sea level changes.
That's a huge ass wave as it's a pulse traveling on top of the ocean, above sea level.
That's what it is like out at sea. There's a reason tsunamis are referred to as "tidal waves." For example, watch this video of a tsunami hitting a port today: https://www.youtube.com/watch?v=M1B1J6sgFxk
Yeah, they are waves I think. Just, really incredibly big waves with lots of mass behind them. I think people want to say “not a wave” to emphasize the fact that they are much bigger than the waves that the local environment is used to, so they can be really surprising.
Since we're intuiting, I'm just imagining something like quickly adding a "D.C. offset" of some given height to the crests and troughs you'd measure by sampling ocean waves.
In fact, I'm not sure I should have quotes around that. Isn't your interlocutor saying a tsunami is literally a direct current of water flowing toward the shore?
I guess it is like a step function, or at least a step function on one side and a really long decay on the other. Is a step function a wave? I’m not sure, my signal processing class was too early in the morning. Maybe it depends on who you ask, mathematicians vs engineers. I’ll go along with the ones that might make a taser or something.
Maybe the easiest way is explain it by volume of water coming at you. A 'normal' wave comes at you for maybe 2-5 seconds, then recedes. A tsunami wave might come at you for what, a few minutes? So moves more than 20x-50x the water than an equivalent 'normal' wave, which has no other way to go?
Sure it's a wave, but tides, swells and waves all oscillate just on different frequencies and amplitudes. When they all align you get rogue waves and to the casual observer of a tsunami, a wall of water coming your way.
> If the left side lowers with a shift then that shift MUST be localized to the area around the fault. Because if it wasn't then that means there's an elevation change across the board for everything to the left of the fault. You see how that doesn't make sense
Yo heard of fluid dynamics? Good luck localizing this;) maybe you can build a wall or something real quick
Obviously it is all technically waves. Even if EVERYTHING to the left lowered we would be talking about waves caused by it. But it don't need to be all lowered because waves propagate. And point is these particular waves, tsunami are not the waves you think about because you saw some on the beach. It's an ocean rising for a while. Watch some vids to get a vibe for it.
Despite the common vernacular calling them "waves" they're really more like really really high tides. You're talking about something that happens over, say, 10-90 minutes, not seconds.
This is also in many ways what makes them so deadly in places that aren't used to tsunamis. It often just looks like a regular wave or a tide that will imminently break or recede, but they never do. Here [1] is a video of one of the later waves of Thailand's 2004 tsunami.
Even worse is tsunamis are also often preceded by a 'disappearing coast' effect where the water will recede back into the ocean for hundreds of meters. This often drives tourists or locals who don't know better to go check out the sea bed and the weird behavior of the ocean, then the tsunami comes in and they're right in the middle of it.
If you're ever at a beach where the water starts rapidly disappearing, yell tsunami and get away as fast as you can. Ignore the normalcy bias, because most people, even locals, will be just standing around taking videos or even walking out into it. And don't stop running even when you're well away from the beach. It's nature's warning sign.
On 26 December 2004, at 07:58:53 local time (UTC+7), a major earthquake with a magnitude of 9.2–9.3 Mw struck with an epicentre off the west coast of Aceh in northern Sumatra, Indonesia. The undersea megathrust earthquake, known in the scientific community as the Sumatra–Andaman earthquake,[8][9] was caused by a rupture along the fault between the Burma plate and the Indian plate, and reached a Mercalli intensity of IX in some areas.
A massive tsunami with waves up to 30 m (100 ft) high, known as the Boxing Day Tsunami after the Boxing Day holiday, or as the Asian Tsunami,[10] devastated communities along the surrounding coasts of the Indian Ocean, killing an estimated 227,898 people in 14 countries, violently in Aceh (Indonesia), and severely in Sri Lanka, Tamil Nadu (India), and Khao Lak (Thailand). The direct result was major disruption to living conditions and commerce in coastal provinces of surrounding countries. It is the deadliest natural disaster of the 21st century,[11] one of the deadliest natural disasters in recorded history, and the worst tsunami disaster in history.[12] It is also the worst natural disaster in the history of Indonesia, Maldives, Sri Lanka and Thailand.[13]
It is the most powerful earthquake ever recorded in Asia, the most powerful earthquake in the 21st century, and the third or second most powerful earthquake ever recorded in the world since modern seismography began in 1900.[14][a] It had the longest fault rupture ever observed, between 1,200 km and 1,300 km (720 mi and 780 mi), and had the longest duration of faulting ever observed, at least ten minutes.[18] It caused the planet to vibrate as much as 10 mm (0.4 in),[19] and also remotely triggered earthquakes as far away as Alaska.[20] Its epicentre was between Simeulue and mainland Sumatra.[21] The plight of the affected people and countries prompted a worldwide humanitarian response, with donations totalling more than US$14 billion[22] (equivalent to US$23 billion in 2024 currency).
I was around (in India) at the time, but not near the coast, much further inland and to the north, so was not affected.
Yes, they are waves, but they are often very long waves. A typical 1m wave might be 20m long. A tsunami wave might be a kilometer long or longer. That is why people say they are like a tide. The wave arrives, then does not recede for several minutes. So, while a 4m wind driven wave might break over a seawall and even wash a car off the road, a 4m tsunami washes ships over that same seawall and floods the city.
It’s a wave, but it is often not at all like a regular ocean wave. I’ve been at sea when a 3m tsunami passed, we barely felt it. If it had been a 3m wind wave in that otherwise calm sea, it would have knocked dinner off the table.
Perhaps we can just go back to calling them tidal waves. Which is also ambiguous. I guess if I had any point it's just that it's not colloquial to call tsunami waves, its technical. If anything distinguishing based on how they feel compared to regular wind waves is more colloquial.
Honestly tidal wave, a wave that comes in like the tide, might actually be a better term than tsunami (lit Harbor Wave) A wave that destroys your harbor? A wave found in your harbor(a place where there are usually no waves)? What if you are not in a harbor, do you still get a wave?
I am just having a bit of linguistic fun. Tsunami is a great clear distinct term for what can be a very destructive event.
They are waves, but they don't behave like the sort of waves we are used to. This is the source of all the confusion.
I have heard description of a tsunami being "a temporary rise in sea level", which describes its behavior much more intuitively. A tsunami that tops a sea wall will flood the entire lower-lying area behind it. A usual wave, even a tall one, will only deposit some splashes of water behind the wall and go away immediately.
It's a distinction without value I think. There are waves, and many of them. There is a rise in the sea level. For anywhere affected, both certainly matter. Like you mentioned, tsunami isn't a brief event. And here in Japan, they are talking about tsunami waves, not a singular tsunami. And talking about sea level rise and checking the local power poles for sea level indicators from previous tsunami events and floods.
I is absolutely a VERY valuable distinction because the behavior as it affects humans (up to and including killing them) is VERY different.
Regular waves that are a little higher than your seawall might cause some water damage to the buildings right next to it. A tsunami that is a little higher than your seawall will flood your entire town and drown people who are caught in basements.
Sure, but if you insist it's like a tide you downplay the risk of the initial hit of the wavefronts and the potential for it to slam up the coast or a seawall becoming a larger local wave. And if you insist it's like a wave, you downplay the persistent risk of both follow-up waves and ongoing flooding that won't subside quickly.
So saying it's not waves is dangerous, and saying it's not a sea level rise is dangerous. It's not useful to try and delineate between a tsunami being one of the two when it's in reality an event that consists of both.
(Ignoring that a sea level rise and a long-wavelength wave are the same thing)
These things are 100% waves. It's not a misnomer. It fits the scientific definition of waves and it fits our intuition of what waves are. These are NOT tides.
Yes, they are waves, but they are often very long waves. A typical 1m wave might be 20m long. A tsunami wave might be a kilometer long or longer. That is why people say they are like a tide. The wave arrives, then does not recede for several minutes. So, while a 4m wind driven wave might break over a seawall and even wash a car off the road, a 4m tsunami washes ships over that same seawall and floods the city.
It’s a wave, but it is often not at all like a regular ocean wave. I’ve been at sea when a 3m tsunami passed, we barely felt it. If it had been a 3m wind wave in that otherwise calm sea, it would have knocked dinner off the table.
A tsunami absolutely does not fit our intuition of what waves are. It looks like a wave. But it does not stop. It just continues. That little wave goes on an on, farther and farther inland. After an hour it may still go on. It's a nightmare wave, because it doesn't not fit one's intuition of what waves are.
I'm reacting against the word "intuition" here. Nobody said it isn't a wave. But it's not like a wave as our intuition says.
(Though it does not _always_ look like a wave - check out the timelapse video from Kuji harbor: https://www.youtube.com/watch?v=M1B1J6sgFxk - that's why you'll sometimes see people talking about it looking like a (very) high tide.
As an example of the nightmare "wave" which doesn't stop, and is thus un-intuitive.. https://www.youtube.com/watch?v=3618dZoiaPE which is incidentally also from Kuji, but from 2011)
I remember in 2015 watching this great tsunami video at a harbor. It was about 11 minutes long.
At the start, there's just a white line at the horizon. Then the fishing boats in the harbor start rocking and jangling. Then water starts pouring over some walkways and sea walls.
Eventually the cameraman backs away and starts climbing a concrete tower; water starts to flood over the area where they had been standing. I think they climb a couple stories and are safe up there.
I haven't been able to find the video in years, but I remember being fascinated by it and I'd love to watch it again.
Edit: I never expected to find that video again, but here it is. A little more terrifying than I remember.
Ah, and here I was wondering if it would be possible to surf one of these for miles in if the timing were right. The grandparent answers that question.
I seriously wonder if people brains are being cooked these days. One of the blessing of HN used to be it was full of fairly well educated, and most importantly, curious people. Sometimes with a bit too much of a focus of the technical side of things, but at least on most technical topics the comments where a great place to get a richer understanding of a whatever was being discussed.
A tsunami is not a "bigger" wave like the ones that crash on the beach every minute. A tsunami is a single wave that crashes and crashes and adds more and more and more water for several minutes non stop, not pausing or pulling back for a single second. It is a sudden flood coming from the sea.
Depends on topography and protections in place. 10 1m waves against a sound 1.5m seawall is no big deal. 1 2m wave against the same seawall could be a problem.
Forcing reasoning is analogous to requiring a student to show their work when solving a problem if im understanding the paper correctly.
> you’d have to either memorize the entire answer before speaking or come up with a simple pattern you could do while reciting that takes significantly less brainpower
This part i dont understand. Why would coming up with an algorithm (e.g. a simple pattern) and reciting it be impossible? The paper doesnt mention the models coming up with the algorithm at all AFAIK. If the model was able to come up with the pattern required to solve the puzzles and then also execute (e.g. recite) the pattern, then that'd show understanding. However the models didn't. So if the model can answer the same question for small inputs, but not for big inputs, then doesnt that imply the model is not finding a pattern for solving the answer but is more likely pulling from memory? Like, if the model could tell you fibbonaci numbers when n=5 but not when n=10, that'd imply the numbers are memorized and the pattern for generation of numbers is not understood.
> The paper doesnt mention the models coming up with the algorithm at all AFAIK.
And that's because they specifically hamstrung their tests so that the LLMs were not "allowed" to generate algorithms.
If you simply type "Give me the solution for Towers of Hanoi for 12 disks" into chatGPT it will happily give you the answer. It will write program to solve it, and then run that program to produce the answer.
But according to the skeptical community - that is "cheating" because it's using tools. Nevermind that it is the most effective way to solve the problem.
This is not about finding the most effective solution, it’s about showing that they “understand” the problem. Could they write the algorithm if it were not in their training set?
That's an interesting question. It's not the one they are trying to answer, however.
From my personal experience: yes, if you describe a problem without mentioning the name of the algorithm, an LLM will detect and apply the algorithm appropriately.
They behave exactly how a smart human would behave. In all cases.
It's hard. But usually we ask several variations and make them show their work.
But a human also isn't an LLM. It is much harder for them to just memorize a bunch of things, which makes evaluation easier. But they also get tired and hungry, which makes evaluation harder ¯\_(ツ)_/¯
If we're talking about solving an equation, for example, it's not hard to memorize. Actually, that's how most students do it, they memorize the steps and what goes where[1].
But they don't really know why the algorithm works the way it does. That's what I meant by understanding.
[1] In learning psychology there is something called the interleaving effect. What it says is that you solve several problems of the same kind, you start to do it automatically after the 2nd or the 3rd problem, so you stop really learning. That's why you should interleave problems that are solved with different approaches/algorithms, so you don't do things on autopilot.
Yes, tests fail in this method. But I think you can understand why the failure is larger when we're talking about a giant compression machine. It's not even a leap in logic. Maybe a small step
The paper doesn't mention it because either the researchers did not care to check the outputs manually, or reporting what was in the outputs would have made it obvious what their motives were.
When this research has been reproduced, the "failures" on the Tower of Hanoi are the model printing out a bunch of steps, saying there is no point in doing it thousands of times more. And they they'd either output an the algorithm for printing the rest in words or code
That seems like a complete non sequitur. This is the model explaining the rest. Obviously the explanation is not very interesting since the Towers of Hanoi is not an interesting problem. But that's on the researches for choosing something with a trivial algorithm if their goal was to test reasoning abilities.
Because that wasn't the task given to them. It's like giving a student a test and you asking them to solve an equation and they give you the general form. It's incomplete
some core GCP cloud services are down. might be a good time for GCP dependent people to go for a walk, do some stretches, and check back in a couple hours.
> I've found great success with LLMs in the research phase of coding.
This is what I've found it most helpful for. Typically I want an example specific to my scenario and use an LLM to generate the scenario that I ask questions about. It helps me go from understanding a process at a high level, to learning more about what components are involved at a lower level which let's me then go do more research on those components elsewhere.
Reinforcement learning system. Currently trying to understand how to implement contextual thompson sampling and its details after doing non contextual thompson sampling. My YouTube history is a lot of logistic regression related videos at the moment.
This is incredible work. Its jaw dropping to learn that something like this is possible at all. Sometimes I wish I could work for a company whose products make a meaningful positive contribution to the work.
Do companies like this have a need for SWEs? Are there opportunities for a backend SWE without any background in hardware or biology?
I run a small software team at a small biotech working on diseases with small patient populations, and the answer is yes x 1000. The issue is that in drug companies, software isn't the product, so SWEs will never make as much money nor be as much of a priority as in tech-proper.
There are two categories of software we need help with:
1. Salesforce for science. We don't have big data in terms of volume; we have big data in terms of heterogeneity. Tons of small data sets that need context to be interpreted, including measuring uncertainty. This software, often called an eLN or LIMS, is offered by expensive vendors who each have their custom, locked-in implementations. Every organization needs customization on top of this that can be developed and change with the changing direction of the bench scientists.
2. Informatics tools. Much of the heavier computational tools (bioinformatics, molecular dynamics, stats) were developed by academic labs, who don't have the training or incentives to create sustainable software. Alternatively, they are made by vendors who write software on short-term contracts, so they don't have expertise in house. Our mass spec vendor told us to put their analysis servers on our Citrix so employees could access it. Citrix! If you can convince those vendor to hire you and rewrite their software, please do.
Despite cool tools like alphafold making headlines, the software needs in drug development are more mundane. We need people who are excited to sit down with bench scientists and help them figure out how very normal tools can be applied to their work.
> Do companies like this have a need for SWEs? Are there opportunities for a backend SWE without any background in hardware or biology?
Of course they do. Biology and medical research can't get enough software people. But they're not as well funded as advertising or spying companies, for example, so you might have to take a significant pay cut.
I wouldn't pigeon hole yourself as a "backend engineer". Why do people do that? Software is software. The bit that matters is the core model and algorithms etc. Whether it's exposed as a web server, a CLI or just a library is a peripheral detail.
It's totally possible for a decent software engineer to learn just enough biology to get by. The limiting factor might be your interest, though. But if you have that then go for it. Get a book on genomics right now.
Yes. Aldevron and IDT are two companies (owned by danaher) that collaborated to make this happen. They have multiple authors listed in the NEJM article.
Maybe a silly question, but has anyone used these in production? Or used libraries in production which are built on these structures?
Im imagining a meeting about some project design, and thinking about how it'd go if someone suggested using parentheses to represent nodes of a tree. I imagine it'd get written off quickly. Not because it wouldn't work, but because of the complexity and learning curve involved.
Do you know the name of the problem or strategy used for solving the problem? I'd be interested in looking it up!
I own DDIA but after a few chapters of how database work behind the scenes, I begin to fall asleep. I have trouble understanding how to apply the knowledge to my work but this seems like a useful thing with a more clear application.
Software engineer here with an android phone. I've never bothered to look into Apple TV because I assumed it'd only be available on Apple devices. Similarly, I saw this post and thought there may be a reason for me to get an iPhone now as I assumed this would be available on apple devices only.
> "there may be a reason for me to get an iPhone now as I assumed this would be available on apple devices only."
That's the objective. Green text and all. To force everyone to adopt one platform because of network effects and social stigma.
These platform plays by the god tier trillion dollar companies are insidious and should be given scrutiny by the DOJ / FTC.
A breakup of these platforms would make none of this matter. You could pick and choose services across devices. We might even see some competition for Android and iPhone if the DOJ would step in and break this up.
Big tech is too big. A breakup would oxygenate the entire tech sector. It would probably even make the MAGMA stock go up because the sum of parts are being given away for free just to get eyeballs.
Billions of dollars are being given away for free to scrape in network effect advantages. It's at a level where competition from new players is virtually impossible. They can tax anything that moves. Every transaction, every relationship, every quanta of information.
I'm only aware it doesn't need an Apple device because spouse does have an iPhone and was able to set it up on our Roku that way. I still assume that someone in the household does need an iPhone in order to get a subscription, although now I think about it probably that's not true.
* we have a single optimization goal (e.g the metric to minimize or maximize). The hard part isnt defining a optimization goal. The hard part is identifying what stakeholders actually want the tradeoff between different metrics. If you have goal A and goal B, where B is more aggressive than A, then getting agreement on which is better is hard. This is a people problem.
* MAB seems to be a good proof of concept that something that can be optimized but isnt an "end game" optimization path.
* MAB for A/B testing is going to mess up your AB data and make everything more difficult. You should have a treatment that uses the MAB algorithm and a treatment that doesnt.
All of the above is for non contextual MAB. I am currently learning about different MAB algorithms although each of them are pretty similar. The ones ive read about are all effectively linear/logistic regression and tbe differences come from exploration mechanism and how uncertainty is represented. Epsilon greedy has no uncertainty, exploration just happens a fixed percentage of time. UCB is optimistic about uncertainty, and the amount of optimism controls exploration. Thompson sampling uses statistical (beta) distributions about uncertainty, exploration happens less as confidence about a particular set of options increases.
Overall its a fun area to work in that is quite different from your typical CRUD development which is a nice change of pace.