Java Data Structures
C# Data Structures
Gov & Misc Docs
Data Sci (old)
App Dev (old)
|Welcome to www.theparticle.com.
It's the newest pre-IPO dot bomb that's taking the world by storm.
Now is a perfect time to buy lots of worthless and overpriced shares!|
Internet is becoming more and more polluted with
junk-mail, people selling crap, and businesses which don't know their place on the net.
They're all trying to make this wonderful place (i.e.: the net) in to hell (i.e.: real
world). Internet should be viewed as a place of imagination, creativity, and most of all:
fun. Internet is not some really advanced tool for searching for people to rip-off. It's
about searching, and finding, things which are useful, helpful, and promote the sharing of
ideas. This is what this site is striving to become.
News, Updates, & Rants...
About 90% of the folks at JFK are wearing face-masks.
- Alex; 20200318
NJ implements a voluntary curfew. Stay in-doors after 8pm.
- Alex; 20200316
Visiting Washington, D.C. It seems everything in-doors is closed. Walked by Lincoln memorial, Washington monument, White House, and Capitol. Lots of free parking everywhere.
...and back to NYC :-)
- Alex; 20200315
Happy Pi Day!
...And another anniversary :-)
Driving to Luray Caverns.
- Alex; 20200314
Trump announced ``we'll be doing more testing'' and suddenly the markets go up up up.
- Alex; 20200313
Brooklyn College Update: COVID-19: Brooklyn College is Closed on Friday, March 13 ``Dear Brooklyn College community, I write to inform you that a member of the Brooklyn College community tested positive for the coronavirus (COVID-19) this evening. ... The student takes two classes on Tuesday evenings and she was last on campus on March 3. She did not develop symptoms until Thursday, March 5, and tested positive this evening.''
The market dropped again, down 9.99% in a day. My prediction: this will get much worse before it gets better.
Went to BJs, and folks have gone crazy... Huge queues. Empty shelves all over. Boring stuff like macaroni is sold out, etc. No flour. Cleaning products, anti-bacterial hand soap, etc., all gone. Wonder if that's happening in other places. Last weekend at Walmart was also very crowded (and no hand-sanitizer).
- Alex; 20200312
Ok, it seems stuff is getting serious. CUNY is moving all classes to `distance learning'. My work already encourages folks to work-from-home, but now mostly telling everyone to do so. Trump suspends travel from Europe (not including UK) to US for 30 days. The markets went into bear territory---and probably go much lower in the next few weeks.
In other news, happy anniversary to yours truly :-)
- Alex; 20200311
The coronavirus news is ramping up quickly...
- Alex; 20200310
Driving to Niagara Falls. The place looks deserted. Almost everything in the area is closed---hotels are cheap, etc.
- Alex; 20200307
To continue with the "physical universe is essentially memory" idea (2020-02-28): How do planets move around the sun? Planets have a sort-of memory of their state (location in the gravitational field)---that state evolves in such a way that the planet appears to move around the sun. Everything about their ``memory'' is encoded in that state.
Why should we expect the universe to have any sense of "physics" as we know it? As opposed to a random mess? Here's my guess: Randomness is very elusive. So any non-random system will exhibit some form of structure, at all levels. In other words, "non-random" is simpler than random. Non-random state will evolve via some non-random process---leading to what we percieve as the "laws of nature" (?).
More thinking ahead.
- Alex; 20200304
The Federal Reserve slashed interest rates by half a percentage point. Hmm... Now I'm wondering what they're not telling us.
The markets by themselves are already flying sky-high... stimulating them like this is just crazy---unless they're expecting some crazy sh*t to go down. (like... coronavirus' 6% death rate to hit most of the world population... 1918-flu killed ~50m world-wide, and that one had 2.5% mortality rate)
- Alex; 20200303
Here's what Bernie Sanders' 'Medicare for All' proposal actually says.
I must say I'm impressed. This is definitely better than the current system (Obamacare, or what came before). But I'd go a bit farther, I'd change: ``Every individual who is a resident of the United States is entitled to benefits for health care'' to ``Every individual is entitled to health care.'' My version is simpler :-). And percentage spending wise, it would be marginal.
This bill should have a similar wording to ``We hold these truths to be self-evident''... applying to everyone. After all, if someone is having flu-like-symptoms, I really don't want them to be thinking ``oh, I'll wait it out and save a few dollars by not seeing a doctor.'' That's like a recipe for coronavirus epidemic. Maybe this coronavirus will wake up everyone to the fact that our collective healthcare is not about individual coverage---if we selfishly want to maintain good health, then we have to ensure everyone around us is also in good health.
e.g. you could have the most expensive insurance plan, and some poor dude in the subway could sneeze coronavirus into your face... if we want a healthy society, everyone's gotta be treated.
A few sneaky things in the bill that may not be ``obvious''. The bill makes it illegal to duplicate coverage (so private insurance corps cannot cover already covered things). Now, everything gets coverage by default---so just about all medical coverage is "already covered", making it illegal for other corps to spring up and offer duplicate coverage. Which is where the sneaky thing comes in: if a doctor wants to be "outside the system" (not participate in this thing), they'd have to have patients who are willing to pay cash---because no insurance company would be able to duplicate-cover a patient just so the doctor can charge the private insurance company more for something that the patient is already covered for under this bill. But I guess that's the point: to eliminate private insurance as an industry, which IMHO is a "good thing".
This has about 0 chance of passing though, even if Sanders gets elected.
- Alex; 20200301
Been having interesting thoughts about the universe; mostly a continuation of what I was thinking around 2019-08-15 regarding the nature of time.
The more I think about it, the more it seems memory plays a central role to our perception of the world. Observations by themselves aren't that meaningful unless we compare them to things we've observed before.
For example, the way we know something is moving is by making 2 observations; and recalling the first observation when we make the second.
It all starts with observations, and memory of such observations. Suppose we remember estimates of probability distributions, then every event/observation will only make sense in that context (an event/observation that is either likely or not given our estimate of the distribution).
Replace "estimate of probability distribution" with a more general ``model'', and suddenly we're talking about `laws of physics'. Note that there's no ultimate truth being derived or observed---just observations that we either recognize or not given our model (which was built using past observations).
At this point, we can ``observe'' things such as objects falling down and building a model of their speed as they're falling... leaving the details of defining ``objects'' themselves (e.g. objects are mostly empty space---made of atoms that we can't see---and those themselves made out of quarks that we currently have no way of ``observing'' direclty---meaning we can't form models of observing them).
So objects are collections of tinier objects that we cannot observe... yet we can observe objects themselves.
Now, if we stick with the idea of working at the same granularity as the observations (forgetting the ``inner structure'' that may or may not be observable), then we're OK. For example, we can model ocean waves... where individual water molecules are irrelevant, etc. There's no "more detail" down there---it's just the level of our observations that we are modeling.
In other words, our laws of nature are just observations mapping to future observations... (via some model that we learn from observations). It's a function... and our memory lets us build estimates of such observation to observation mappings.
(the observations themselves could be subjective, or of poorly defined aggregates, or averages---e.g. temperature.)
Quantum mechanics takes this "observation" idea quite literally. Everything starts with an observation---there is no existence without an observation. But there is still an assumption that "stuff exists"; that there's a truth to the universe out there... perhaps there isn't.
Imagine if WE had no memory at all. Then every event will appear random (no concept of physics is possible). We can't even note that time is passing. Would that mean that planets do not go around the sun, or that atoms do not exist?
All it would mean is that WE cannot model what we are observing.
Now suppose our existing senses (and limited brain capacity) is making us blind to some part of the universe... Does that mean it doesn't exist, or in some sense not real?
(It's quite arrogant of humans to think we can observe everything that is observable...).
Now, lets explore the other direction: of remembering as much as we can. We make observations, and use the best modeling and data compression mechanisms we can imagine---to remember as much detail as we can.
The primary goal of such detailed observations is so we can distinguish a similar but different observation sequences. So detail has to be "at the same level" as other potential observation sequences to enable that.
Anyway, the more and more we remember, the more our memory is being consumed... even if we just remember the count of observations---an incrementing integer---it is still consuming memory (perhaps we only need to distinguish lenght of sequences).
This ever-increasing-memory requirement is similar to Entropy. It cannot decrease. More observations "saved" means more memory consumed. Note that the observations themselves (nor the underlying processes for such observations) do not have to have increasing ``entropy'' or disorder... the entropy is entirely in our minds---if we started remembering things earlier, it would've been higher, etc.
Entropy direction is often likened to direction of time. Entropy increases with time---and time perhaps is a measure of entropy increase. But if it's all in our minds, then time is simply the result of our ability to remember things (by things we mean building models of observations).
Without memory, there's no entropy---nor time (at least for beings without memory).
Which means that perahps for the universe as a whole, time does not exist---without an observer who can form mental models :-)
- Alex; 20200228
Ok, the market is going crazy. According to the news, this is due to coronavirus.
But is it really? Pretty much everyone who is buying into stocks expects a 8-10% return a year. Last year was a crazy return... but "average" (they, the experts say) is 8-10%. Now, consider that earnings yield is ~4% a year now (meaning that if you buy $100 worth of broad market index, you'll get $4 in a year in company earnings---this is not YOUR money yet, this is just "within the company", managed by folks who have the power to pay themselves bonuses). Adjusting for bonuses and inflation, and you end up with a pretty bad yield on stocks... somewhat comprable to the yield you'd get on municipal bonds.
To bring expectations of 8-10% annual return back into the picture, stocks would need to be 50% cheaper, or every company in the market-wide index would need to double their earnings (which given coronavirus is unlikely). Maintaining the "long term" 8-10% returns is just not sustainable at current levels.
- Alex; 20200227
...and back to NYC.
Kiddo played in snow on the way back.
- Alex; 20200223
Drove to Delhi.
Since there's virtually nothing there, we stayed in Oneonta, which is seemingly a much bigger place.
- Alex; 20200222
Driving back to Atlanta. Plenty of time till the flight back to NYC. Or so we thought...
By the time we realized that time is getting tight, it was VERY tight. So we drove-drove-drove to the airport, filled up car, returned car, took train to terminal, checked in car seat, ran to security checkpoint (where they had to manually check ALL our bags), ran to next train (which came just as soon as we got there), went through all terminals A-D... ours was terminal E, then ran to the gate (E34, all the way at the end)... and about 10 gates away, we heard the gate announcement that it's about to close in 1-minute... so ran there, and about 30 seconds later got on the airplane.
It was literally 30 seconds away from "doors closed" that we got into the airplane. If any step along the way took just a bit longer, we would've missed the flight (e.g. lets say TSA agent took just a bit too long to open/close our bags, etc.)
- Alex; 20200217
Visiting the Great Smoky Mountains National Park. Doing a loop around it. North to Gatlinburg via US441, then around the west side of the park via local roads, and back down to East Ellijay, GA.
- Alex; 20200216
The plan is to visit Great Smoky Mountains National Park. So driving north... to Clayton, GA.
- Alex; 20200215
Flying out to Atlanta :-)
- Alex; 20200214
...and back to NYC.
- Alex; 20200209
Driving out to Concord, MA. No particular reason, it's just a place we haven't been to yet.
- Alex; 20200208
And driving back home to NYC.
- Alex; 20200202
Driving out to Ticonderoga, NY (that's the north point of Lake George). Nice drive in wintry conditions.
- Alex; 20200201
Finished reading The Doomsday Calculation: How an Equation that Predicts the Future is Transforming Everything We Know about life and the universe by William Poundstone.
I really didn't know what to expect when buying this book. It seemed to do with Bayes theorem, but it actually is quite a bit more. The basic idea is that self-sampling may often work.
For example, lets say you find a random widget, with a serial number 98. (small number for this exmaple). What you could think is that you're extraordinary lucky, and found the 1st percentile of the widgets ever produced... therefore there are actually a total of 98*100 or so widgets out there in the world, or somewhere aorund 10k.
Similarly, you could assume that you're extraordinary lucky, and you got the 99th percentile (the last 1%) o the widgets ever produced, in which case there are perhaps about 100 or so of such widgets.
Both guesses would be unlikely (you got 1 in 100 to get it "right"). what about over 50% chance? Well... you're found widget 98, which with 50% certainty falls within the middle 50% of the widget production cycle... in other words, 98 could be anywhere between 25th percentile and 75th percentile of production widgets. So with 50% certainty, you can claim that there are ~130 to ~400 of such widgets ever produced.
All that from finding just 1 widget!
I was reading about this concept before. Apparently allies during WW2 used serial number of tank parts to estimate the german tank production capacity.
Anyways, the reason the book has doomsday in the title is because we find ourselves among many humans who have ever lived. So just pretend that every human has a serial number (order of their birth on this planet). We can apply similar logic to guestimate (with say 90% certainty) how many humans there will ever be.... unless we're extraordinary lucky, we are not in the first 1% of the humans that will ever be born (e.g. what's so social about us?). Similarly, we are not in the last 1% of the humans that will ever be born... (again, what's so special about us?). So with any confidence level, we can create bounds on number of humans that will ever exist... and we're somewhere in that range (unless we're extraodinary).
Yes, it's kinda a silly argument, but the numbers are real. If true, it's unlikely human beings will venture out to the stars... which would explain why we haven't been contacted by beings from other stars.
- Alex; 20200127
First day of classes for Spring 2020 semester :-)
- Alex; 20200127
The hotel in Hartford had a pool, but by the time we ended up going, it was closed for maintenance :-/
- Alex; 20200126
Heading upstate to Albany. Ended up in Hartford, CT.
Rain rain go away come again another day...
- Alex; 20200125
Visiting the Smithsonian National Air and Space Museum in DC. Half the buildin is closed due to construction (and will apparently be closed for 3 years). Then the other half will be closed for another 3 years. So it's gonna be 6 years until the whole thing is open again.
Neat fact: parking nearby is FREE on sundays and holidays.
- Alex; 20200120
Visiting the Wright Brothers National Memorial. This place is much bigger than I imaigned. I also thought it was mostly on the beach (video of first flight seems to be on sandy terrain), but it's all grassy now.
- Alex; 20200119
Going on a road trip to North Carolina, the Wright Brothers National Memorial Visitor Center. That's the place the first airplane flew.
On way, stopped by Wallops Flight Facility visitor center (the place NASA launches rockets from). It's pretty neat :-)
- Alex; 20200118
Finished Artificial Intelligence: What Everyone Needs to Know by Jerry Kaplan. Pretty good overview. Not really revolutionary, just a good overview. Speculates on what economy would look like once tech takes over most jobs. Suggest stupid things like investment accounts for kids---where they can invest in whatever they want. This is similar to converting social security into an investment account like thing. The problem with those schemes is that if folks manage to lose everything, what then? No safety net? If there is a safety net, then there's no reason to be safe with investing---tails I win, heads you lose situation.
- Alex; 20200114
Visiting White Sands National Park. It just became a national park on December 11, 2019.
This is an awesome place---and definitely deserves to be a national park. The sand is like snow---and folks slide down hills. Our kiddo really enjoyed this place---drawing stuff in sand, running around sandy hills, etc.
There are picnic tables with bbq grills---so next time we're coming there we'll be prepared for a bbq.
- Alex; 20200112
As we were landing, we decided to visit Grand Canyon National Park. Apparently ABQ is only about 5 hours drive away from the place. So that's where we went. Took the shuttle bus to visit a few attractions.
Then headed to Walnut Canyon, an unforgettable trip.
- Alex; 20200111
Flying out to ABQ (Albuquerque). Gonna visit White Sands National Park.
- Alex; 20200110
Went on a trip to Atlantic City. Stayed in The Claridge. Worst hotel we've stayed at so far. In fact, most of Atlantic City is a pretty ugly place to visit. Very unlike Vegas.
- Alex; 20200104
Finished The Case Against Reality: Why Evolution Hid the Truth from Our Eyes by Donald Hoffman.
Ok, I'm not sure about this book. Some concepts are neat, others are just dumb. Here's a very brief summary of the book: since quantum mechanics says state doesn't exist until it is measured (e.g. Bell's experiment), that means that nothing exists until it is observed. e.g. There's no moon. There's no universe. There is no objective anything. Sciences that rely on observations are missing the big picture. Every observation is very personal, and is in your mind---and the mind just consumes the signals that are fed by our sensors. Our sensors are kinda like the desktop in the computer---we see what we wanna see to get through evolution (e.g. our perception of distance is directly related to how much energy it costs to catch something or run away from something). We see things that are relevant for evolution, not what's really out there.
The last chapter tries to merge religion and science... religion is the study of stuff that is not part of the physical world. Science is all about stuff that can be observed, etc. Since nothing really exists---perhaps the two should merge, keeping the scientific method (religion becomes rational), but discarding reality (science is no longer about observations).
- Alex; 20200101
Happy New Year!
- Alex; 20200101
This sure was an eventful year.
- Alex; 20191231
Microsoft's Andrew Shuman on the Cortana app's death, natural language, and Alexa.
The problem with Google, Amazon, (and I guess Apple, though I've never tried Siri), and Cortana, is that they aren't very useful when it comes to general AI queries, they always-listen and send back everything they hear (perhaps not Cortana, but who knows), and their existing use-cases are not that useful. Their primary use (for me, at least) is to: check time, check weather, play music or search something on the web, schedule alarms and timers. Anything outside of these limited functions have met with various failure modes. This is not that far from Prof.Phreak, and I wrote that in 1996!
Recently google has started to screen phone calls (automatically!); so that's something new, and actually useful.
I guess my primary gripe is that these agents should be MY agents. They shouldn't require internet access (unless I ask them to search the web). They should not be optimizing engagement or conversion rate (to turn me into a paying customer), but MY learning rate. Think The Diamond Age's Illustrated Primer, or the Librarian from Snow Crash.
In other news, spending weekend near Albany :-)
- Alex; 20191228
Finally tried out DOOM, the 2016 version. This game is awesome. It runs just fine on Linux via Steam (with all settings set to max).
The graphics are amazing, and mechanics are awesome---especially the gore parts. Ripping up monsters to shreds, etc., definitely more fun than DOOM3. Music is also good, kinda similar to Quake1 music.
- Alex; 20191221
Hmm... Robinhood trading app fined for 'failures'. Ok, this news story has very vague wording. It seems there's no definite proof that the routed orders really did get worse price (``may not have'' is not proof), and certainly no proof that orders got price outside the NBBO (national best bid and offer---that would've been a big violation for everyone involved). Also, it appears Robinhood doesn't actually do execution themselves, so best-ex is only peripherally applicable to them. The key seems to be: ``FINRA said that Robinhood did not "reasonably consider" factors such as "price improvement" that it could obtain from alternative markets.''
In other words, you submit an order to buy XYZ at price $10.00 or better, and you get an execution at $10.00... but had Robinhood tried their best, they may have found a place where your order may have gotten an execution at $9.99 (often against hidden interest that nobody knows about, not even Robinhood). But it seems Robinhood just routed the order to whoever paid them the most money. In other words, Robinhood should have known better, and should have been on the side of their customers, not taking advantage of them.
This is serious, and something that Reg BI would fix (once it goes into effect in 2020).
- Alex; 20191220
Played a bit with the Amazon Alexa... and on first impressions, it's much dumber than `Ok Google' assistant. e.g. ``sing something for my 2-year old'' doesn't work on Alexa, but works on `Ok google'. Both devices are stupid, e.g. asking ``what's your maximum volume level'' didn't return a descired response in either, and google actually set itself to the maximum volume level.
- Alex; 20191207
Couldn't resist the Black Friday deal on Google Pixel Slate. It's surprisingly big. I somehow thought it would be the size of a Surface Go, but it's actually about the size of my X260 ThinkPad (about 0.5 inch less wide), and about the same weight (with keyboard attached). This ``tablet'' has the same portability profile as a laptop.
The good news is that aparently Linux stuff runs right out of the box, without any unlocking or fiddling. Stuff just works. Can apt-get anything you want, and it's there. Very impressive machine. Also been reading up on how security is handled in ChromeOS, and apparently it's ``fairly secure'' (each user account is encrypted, etc.,), though I'm not sure I'm ready to trust Google on this yet. Is it ``secure'' in a sense that nobody can get in, or is it ``nobody except Google'' can get in... them encrypting files using your Google account credentials seems like it wouldn't be very secure from google itself.
- Alex; 20191201
What would an atom of hydrogen look like right next to a black hole? Specifically, how would the electron orbital look like?
Here's a guess: the orbital will be deformed by gravity---instead of being a perfect sphere, it will look like tear-drop with the pointy end towards the black hole. Due to this deformation, the proton (and subsequently the atom) will be accelerated towards that ponity bit (towards the black hole).
All this should also happen on earth---just not as extreme. That means that atoms on earth should not be perfectly spherical, with the exact shape depending on their location.
- Alex; 20191125
So around 2005, I bought $1k of DIA and $1k of QQQ, and mostly ignored those positions, thinking they'd be more or less the same.
The other day the position were: DIA ~$2.79k, and QQQ $5.05k. That's quite a difference! Adjusting for dividends, that would've been $3.5k and $5.5k respectively. Keeping money in the bank with the best savings account interest rate would've resulted in $1250 balance (gain of $250 over 15 years). Similar bad results for government bonds.
- Alex; 20191124
Uh, oh! Hungarian scientists may have found a fifth force of nature. Unless you're a Jedi, there is no such thing as ``force''. No, the `force' of gravity is not a force. What we observe as force is simply effects of a quantum field. This research appears to have found a new particle, one that they are labeling as a `force carrier'.
- Alex; 20191123
If Electromagnetic radiation includes light and magnetism, why don't we have magnetic lasers?
...and back on the Internet via Verizon Fios :-)
- Alex; 20191116
I've recently pontificated on the notion of similarity in class; might as well write down the important points.
Similarity is impossible to define objectively. For example, how would we compare thing A to thing B?
One possibility is to check if thing A is actually renamed thing B (identical). Kind of like a symbolic link A pointing to B. That would give us a measure of either 1 or 0: identical or not. But then this ``identical'' measure isn't exactly perfect, as one thing is a symbolic link and the other is not.
Now, if the things are obviously different, then the measure returns a 0. At this point we can either accept it, or redefine similarity on a scale of 0 to 1.
One common similarity measure is to extract features from thing A, and thing B, and then perform some sort of distance measure between the respective feature vectors.
Another possibility is that the scale is the fraction of features that are shared (not distance, but count of shared features).
Note that both these approaches, we are extracting features: what kind of features do we extract?
Consider if we're comparing image A to image B. Do we extract the pixels? The size of the image? The average brightness of the images? The choices are limitless (well, finite world and all that)---and our choice of features will depend on what we consider to be important features for the task at hand.
Other researchers doing other things with the same thing A and thing B will be extracting different features.
- Alex; 20191115
Still no internet :-/
This is actually quite nice. No internet TV. No YouTube. No internet browsing. It's awesome---from a certain point of view.
- Alex; 20191111
...and very abruptly we're without Internet service :-/
- Alex; 20191108
Happy Birthday kiddo!
- Alex; 20191105
During the hike, stopped (quite often) to just sit and gaze at the stars. Saw 5 shooting "stars", and 1 sattelite. The dark skies there are awesome.
After the grand canyon hike, passed out for a few hours in the car.
Chatted with some random stranger, and learned that the record for rim-to-rim-to-rim run is apparently 5:55:20, set on Oct. 4. 2016 by Jim Walmsley, just 2 days after I hiked that trail :-). That's an amazing time. Just crossing from south-to-north in under 6-hours (running) is a huge achievement---this dude did the whole round trip in that time. Consider that he'd have to ``average 8:27 per mile over a trail route that includes more than 22,000 feet of vertical change'' (quote from article). That's damn impressive.
Not wanting to do anything too physical the rest of the day, decided to drive to Wupatki and Sunset Crater. On the way there, noticed a `Red Mountain Trailhead', so pulled in. Went on a 3-mile hike of that crater---and it was awesome.
Then drove down the Wupatki and Sunset crater road. Then went to Walnut Canyon---but was late by 5-minutes. Apparently they don't let folks enter the `island' (the main attraction) past 4pm, and I arrived at 4:05. the no-fun-park-lady wouldn't budge on letting me in. Anyways, ended up doin the `rim' hike that I've never done before. Not terrible, just not as fun as walking around in the island.
...and then got stuck in traffic on way to Phoenix. There's like 1 highway in the whole region, so once it got "closed" (due to accident), there was a traffic jam that spanned hours. I tried everything, Waze, maps, etc., was going from one place to anoter, but the ugly thing was that the accident was right after the last exit where you could change route options... beyond that point, there was only 1 option, the highway.
Seriously considered going back to Sedona, and trying another route to Phoenix... but that would've been 3-hours and I would've missed flight.
Lucikly, the police opened 1-lane, and the traffic started moving. And I didn't miss the flight! :-)
Android Auto: This was the first car rental that had Android Auto feature. It was a Dodge Ram 1500 pickup truck. Awesome truck, but Android Auto was very unreliable. At first it connected, but then it kept getting frozen---to the point of being unusable. Ended up navigating via the phone (which was the original plan).
- Alex; 20191103
The flight was late. Apparently. I passed out as soon as I got into the seat, but apparently I slept through a lot of delays. Anwyays, got to Phoenix a bit later than usual, and then while driving to Grand Canyon stopped at a rest stop and napped some more. Got to grand canyon around 7am.
Stated down the South Kaibab trail. There are a LOT of people here. A lot more than usual---a lot more than I expected. In fact, a large fraction of the people are doing rim-to-rim-to-rim, except most of them are running. Don't think I've ever seen this many people doing this. I was hoping for this to be a quiet time, but there are a ton of hikers on the north rim as well (though everything there is closed).
Anyways, I'm terribly out of shape for this. I haven't done this in 2 years, and these 2 years haven't been kind to my fitness. I really struggled. Barely got to the summit of north rim, and on the way back, barely got to the summit of south kaibab. Every step was a huge struggle. When I was doing this hike twice a year, it was much easier...
The weather was... all over. It was about 80 degrees in some spots (with the sun, probably hotter). Then it was probably close to 20-degrees in some places at night.
On the way back, about 10 or so minutes before Phantom ranch, slipped on a lose rock and fell badly on my arm. Used up all the bandaids. Wasn't fun.
- Alex; 20191102
Flying out to Phoenix, Arizona.
- Alex; 20191101
In other news, changed car oil. It's been 20 months and almost 20k miles since the last change (I'm not very proactive with these things). Surprisingly, the car sound changed after the oil change---it was very deep ggrrr engine sound, now it's not as deep.
- Alex; 20191027
Got Surface Go restored---Microsoft sent me a Windows 10 recovery USB, so just insert and reboot (while holding down volume button). Back to factory settings. There's a program there I didn't notice: whiteboard. Pretty neat with a pen.
- Alex; 20191023
- Alex; 20191021
What's the best way to sort data?
If the data can fit in RAM, we use one of the methods we've learned in school: bubble-sort (eh, just kidding). Likely Quicksort, or heap-sort.
Actually, quicksort may not be the best way: there's such a thing as radix sort, which may be faster. e.g. quicksort is generally O(n log n), while radix sort is often linear, O(k*N) where k is some constant.
If the data doesn't fit in RAM (or is too big for one CPU to chug through), we have no choice but to cut it into chunks: use merge-sort. The basic idea is to cut the dataset into two, recursively repeat this cutting until we can fit the chunk into RAM. Once in RAM, we use quicksort. Then we ``merge'' sorted chunks---since we always split the original file into 2 (recursively), the merge operation merges two files at a time.
A more generalized version of this is to cut the file into N chunks (e.g. 1000gig file cut into 1gig chunks). Each of the N chunks is sorted independently. Then the question is how to ``merge'' 1000 such files.
One possibility is to do pair-wise joins (ending up with 500 sorted files after first iteration, then 250 files on next iteration, etc.). This method forces us to rewrite the entire dataset multiple times.
The other possiblity is to open all 1000 files, read first record from each file, and do a search to find the minimum element. Output that element, and advance to next record from that file. Then repeat sequential search again to find the next minimum. Repeat until all the 1000 files are done. Problem is how we do this search for the minimum: if we do a linear search through all 1000 open files, then we're actually doing O(N * 1000) number of comparisons (where N is number of records being sorted).
If we want one process to merge 1000 sorted files, the proper way to do this is via a heap. Open each file, read first record, and insert it into a heap. O(log 1000) speed. Pick the top element from the heap, output it. Advance pointer, re-insert the value back into heap. Repeat. This heap based merge has cost O(N * log M) where N is number of records being sorted, and M is number of files. In the case of 1000 files, that's O(N * 10)
Ok, but the bottleneck problem still remains. We did a good job of cutting the dataset into N chunks, but the merge operation funnels everything into one CPU. (the pair-wise merge can be parallelized, but in the end, the entire dataset has to go through one CPU).
Can we do better?
Of course we can! Lets address why we need the merge in the first place? Because each chunk has the entire range of values. What if we can somehow ensure that each chunk only has a certain range of values? For example, if we know that the sort key is a string, perhaps we can create buckets "A" through "Z", ensuring that each file only has keys in a certain range. Sort each such chunk, and we're done: the files are collectively sorted. (e.g. we can use binary search to find each record). This is the scheme that many companies use to partition their data. The value indicates which file/partition the data is stored.
That's all good, but what about keys that can't be easily partitioned into "A" through "Z" range? For example, if the key is a timestamp? Perhaps we can ``partition'' the data by the hour? Or Minute? or Day? Each partition can be accessed (and sorted, etc.,) independently. Given a record, you can instantly figure out which partition it should go to.
What about skew? Perhaps some hours will have no data at all, and some hours will have millions of records. e.g. lets say your data is customer purchases. You partition the data by hour of the day. Very few people purchase stuff at night vs the day. So the 1-hour chunk of data is too tiny for night hours, and too big for day hours.
The solution is to sample the sort keys. Figure out how many records you'd want per data chunk (something that can comfortably fit in RAM for fast sorting); perhaps N records per file. Use N and record count to figure out how many file chunks you'd need, lets all that M chunks. Ok, so sample random M*1000 records from the dataset (often a quick scan through the data). Sort these M*1000 records, and find values for percentiles: what's the 1/M th percentile? what's the 2/M th percentile? what's the 3/M th percentile?, etc.
Create an array of these percentile values: then scan data (often in parallel), for each record idx, calculate k=lower_bound(percentilearray, recordidx). Send the record to file "k". (this is a "map" in Hadoop world). Anyways, you'll end up with M files, each with about N records, where key ranges do not overlap. You can now sort each of the M files, and declare success. (note, lower_bound step is O(log M)).
...and that's how we can sort billions of records on a cluster, without much skew screwing up things.
But that's not all! We can also use this sort mechanism for sort-merge joins! Take two big datasets, cut each one on key into M buckets (using same percentile table). then just do a sort-merge of each of the M chunks.
Because all the keys that are near by go to the same chunk, we can also use this mechanism for sequence joins (e.g. finding certain last record prior to current record, sequenced by time). The trick is to bucket the data, find the boundary records (last certain record in the chunk). Then just sort those `last' records, and re-insert them into the subsequent chunk as the first record. So the last record from previous chunk becomes first record in the next chunk. Then the compute can be done in each chunk, without shuffling the data, nor sorting the entire dataset on time (on a single node).
- Alex; 20191017
Eh. So my Surface Go did its automated updates... and suddenly became unbootable. It seems Windows 10 update has corrupted itself :-/
- Alex; 20191013
Happy b-day to yours truly :-)
- Alex; 20191005
Finished How Smart Machines Think by Sean Gerrish. An executive summary of how ML was organized in the last few years for public challenges. e.g. How the self-driving car modules are organized (sending, reasoning, etc.), or what strategies the chess playing programs used, or what approaches the folks who analized the Netflix dataset used. Not a waste of time, just not as deep as I would have liked it to be.
- Alex; 20191001
Ran up Katahdin. Awesome hike. 3-hours up, and 3-hours down. Walked the length of the Knife Edge as well. The Katahdin Stream bridge is ``closed'' (well, the bridge is just 1 stick across the stream). It's jumpable. So I jumped it... the alternative is to walk 1-mile back to the parking lot, and 1.1 miles back to the stream via the detour trail. On the way back I took the detour route (it's a much longer walk through muddy forest).
Saw lots of mushrooms, and 1 tiny snake :-)
- Alex; 20190928
Kiddo finally started to pronounce his own name. He was able to say lots of names (and name lots of things), but not his own name. Now he's saying "Ian" pretty clearly. Yey!
- Alex; 20190926
Fixed my car's tail light. The culprit turned out to be a loose plug connection to one of these: Right Tail Lamp. After moving the plug around, the light came on---so bending the pins on the connector a bit with a screw driver, then forcing the connector in, seems to have fixed the broken light that's been on my brain since June's road trip.
- Alex; 20190924
Finished reading The AI Delusion by Gary Smith. It's much much better than I was expecting: no FUD, just very honest and true points about how AI isn't anywhere close to intelligence.
- Alex; 20190916
Read ``Exposition of a New Theory on the Measurement of Risk'' by Daniel Bernoulli. This is one good paper! Essentially risk is subjective: cannot be evaluated without taking into account of who is taking the risk.
Key quote: ``Somehow a very poor fellow obtains a lottery ticket that will yield with equal probability either nothing or twenty thousand ducats. Will this man evaluate his chance of winning at ten thousand ducats? Would he not be illadvised to sell this lottery ticket for nine thousand ducats? To me it seems that the answer is in the negative. On the other hand I am inclined to believe that a rich man would be ill-advised to refuse to buy the lottery ticket for nine thousand ducats. If I am not wrong then it seems clear that all men cannot use the same rule to evaluate the gamble.''
So lets break this down: the lottery ticket is worth either $0, or $20000. So the expectation is 0.5*$0+0.5*$20000 = $10000. So selling it for $9000 seems stupid. Or does it? Lets consider the utility from this paper.
Suppose A's total net worth is $1000, with this lottery ticket, A's net-worth will either remain $1000, or go upto $21000. Doing a geometric mean, exp( 0.5*log(1000) + 0.5*log(21000) )=4582.6. Removing our initial $1000 from it, gets us: $4582-$1000 = $3582. That's the value of the ticket to A. If A can get $4k for it, A should go for it, but shouldn't sell it for less than $3k.
Now, suppose B's total net worth is $100,000, with this lottery ticket, A's net-worth will either retain at $100,000 or go upto $120000. Doing a geometric mean, exp( 0.5*log(100000) + 0.5*log(120000) )=109544. Removing our initial $100,000 gets us: $109544-$100000=$9544. That's the value of the ticket to B.
The ``value'' never touches (or goes above) $10000, no matter how much net worth we start with. It's never a good idea to buy a ticket that's overpriced.
So we have both A and B, where A values the ticket at $3500, while the B values the ticket at $9500. That's a range of prices at which A can sell the ticket to B and both would benefit (win-win situation). Below $3500, A should hold onto the ticket (or buy it, if needed), above that it should sell it. Below $9500, B should buy (or hold onto it).
It's the same idea behind insurance. Suppose individual A has assets of $100k, and there's a 1% chance of them getting cancer within a year. Suppose cancer treatment costs $250k. The individual ends up bankrupt (going down to say $1). What's the expected net-worth after a year without insurance (again, using geometric mean): exp(0.01 * log(1) + 0.99 * log(100000))=89125.
So spending $10k or less on insurance makes sense. If insurance costs more than that (say $11k), then they might as well go without insurance.
On the flip side, the insurance company has ``unlimited'' assets (lets imagine $1-billion), how much should they charge? exp(0.01 * log(1000000000 + x - 250000) + 0.99 * log(1000000000 + x)); So for x above 2500, they're good to sell the insurance.
Now, where insurance actually falls between $2500 (the company's low limit) and $10k (the customer's high limit) is due to marketing and competition. The win-win price range. Notice that if individual has 10k in assets, then their high number drops to $800 or so, at which point they cannot get insurance.
Lets examine a ``fair'' bet: Both A and B decide to flip a coin, tails A gets $100, heads B gets $100. Lets say A has $1000 in assets, while B has $10000.
Expectation for A after the coin toss: exp(0.5*log(900)+0.5*log(1100)) = 995.00
Expectation for B after the coin toss: exp(0.5*log(9900)+0.5*log(10100)) = 9999.50
To summarize, in a fair bet, where we wouldn't expect any side to have any advantage: both sides lose (you shouldn't participate in fair or unfavorable bets; you need to have an edge). Also, the participant with most assets risks less (the more money you have, the less risky things become [even in completely symmetrical situations, such as coin flips]).
UPDATE: Kelly criterion!
Suppose we are lucky enough to find a game where we have an edge. The game goes something like this: There's a probability p of winning. We wager a certain amount. When we win, we get back the wagered amount plus fraction b of it. When we lose, we lose just the wagered amount.
Kelly started with, supose we are certain we will win. p=1. Then it makes sense for us to bet everythign we have, meaning that every time we play this game, we make 1+b return. In a simple scenario where b=1, we double our money with every round. This is the maximum return we can expect.
Now, in most scenarios we aren't certain to win. If we were to wager everything we have, we are certain to lose everything very quickly. On the other hand, we do want to take advantage of the favorable odds, which leads to betting some fraction of our wealth. Ok, so what fraction should we bet?
(another alternative to avoid ruin is to bet a small fixed amount, say $1 every round. But it turns out that is not an optimal way to grow capital).
Our return is: S * X1 * X2 * ... XN
where S is our starting wealth, and X1..XN are random varibles (our returns). We wish to maximize this product of X1..XN.
Any product can be expressed as a sum: e^(log(S) + log(X1) + ... log(XN))
Xs are random variables, and so are log(Xs). They have expectation E[log X], leading to:
e^(log(S) + N*E[log X])
It's clear that if we are to maximize the outcome, we need to maximize E[log X]. If probability of winning is p, and losing q=1-p, and we bet fraction f of our money, then the expectation per round is:
With probability p, we multiply our wealth by 1+fb, and with probability q we lose f fraction of our wealth. So how do we maximize
We take the derivative, and set it to 0, then solve for f.
p*1/(1+fb)*(b) + q*1/(1-f)*-1 = 0
pb/(1+fb) - q/(1-f) = 0
pb/(1+fb) = q/(1-f)
pb*(1-f) = q*(1+fb)
pb-pbf = q+qfb
pb-q = pbf + qbf = f*b*(p+q)
f=(pb-q) / b = (pb-(1-p)) / b
So if we come across a scenario that will double our money on a bent-coin flip (p=0.6), we should bet: p=0.6, and b=1, f=(0.6-(1-0.6)) / 1 = 0.2, or 20% of everything we have.
...if only it was that simple!
First a word on what exactly this fraction means: This fraction is the fastest way to grow capital. There is no other method that grows faster than this. Betting smaller than this fraction adds safety, but wouldn't maximize growth. Betting bigger than this fraction would likely cause ruin long term.
Now for the problems: real life doesn't present us with clear variables to stick into a function. Most bets (say the stock market) are hard to quantify in these terms. The probabilities aren't fixed: it isn't a well defined game of chance---we need to estimate probabilities, and those estimates can be off.
The other major problem is that we're maximizing long term growth, not lower volatility. With every bet being very aggressive (in relation to our wealth), our fortunes could go up and down dramatically---and if there's a minimum betting amount, we could face ruin.
More in this book: Kelly Capital Growth Investment Criterion: Theory and Practice. It's a collection of papers on this stuff. Well worth reading (well, most of them anyway).
- Alex; 20190906
Finished reading Hooked: How to Build Habit-Forming Products by Nir Eya.
Essentially there are 4 steps to building addictions (well, habit-forming products): step 1: trigger. Get something to initiate interaction. Can't go forward without some interaction---something pushing folks to do something. chill out, have a coffee. or a smoke. 2: action. Folks must do something. Press a button. light up a cigarette. anything. 3. variable reward. action must give rewards. But not just a reward (folks don't get hooked with regular rewards). it has to be variable reward... every time it's different. Keep folks guessing on what they'll get when they press that button. you can't crave something you always get. 4. investment. time. money. attention.
The more you put folks through this 4-step loop, the more they get hooked to using your product.
In other news, I've been playing World of Warcraft Classic lately. I feel chilled out in the game (trigger), I get to click buttons, run around, complete quests (actions), and I get rewarded (experience, loot, quest rewards), and I invest back into the game by playing (time, and potentially money). addictive?
- Alex; 20190904
Trying out WoW classic. So far so good.
- Alex; 20190830
...and back in NYC!
Today is also the first day of class at Brooklyn College :-)
- Alex; 20190828
Don't have anything planned for the day---so ended up driving to Pike's Peak for a picknick. Unfortunately the park is in some country that has a "fire ban", so no open fires on the mountain (though they have fire pits).
Drove through garden of the gods.
...and to Chatfield State Park (not very impressive).
...and back to Denver airport, where Suneli was padded down for trying to bring a bottle of formula milk for kiddo :-/
- Alex; 20190827
Driving through Rocky Mountain National Park.
The park rangers at the Aprine visitor center are jerks. They put a sign "no left turn" right by the entrance to the visitor center. So any east bound cars cannot turn into the visitor center---only west bound cars can turn (east bound cars apparently have to drive past to the next turnout, do a u-turn, and come back west bound). Cars that just turn left into the visitor center get a talking-to by the park ranger who very loudly says "no left turn, this is a traffic violation with a $250 fine" (didn't see anyone actually get written a ticket, but lots of drivers got that talking-to).
The parking lot had a lot of empty spots, so it wasn't like they were trying to control the crowds. No reason for that "no left turn" sign except to annoy east bound drivers.
Staying in Colorado Springs, at `the academy hotel'.
- Alex; 20190826
Long long long long drive to Rocky Mountain National Park.
Staying right outside the east entrance, in a Holiday Inn, in Fraser, CO.
- Alex; 20190825
Next stop Yellowstone. It's one of the places we've already visited this year, but I really wanted to see Artist Point again, so that's what we're doing.
Right through Yellowstone is Grand Teton National Park. Visited Jenny Lake.
What I'm amazed about Grand Tetons is that almost every time I come here, it's slightly different. Last time I was here in 2016, hiking Static Peak (via Death Canyon). Didn't make it to the summit due to snow (June). That experience/scenery was very different from the views this trip provided. This park is just amazing. Wish I could spend more time here.
- Alex; 20190824
Arrived in Glacier National Park.
Doing to the sun road, etc.
Camping right outside west entrance of Glacier NP.
- Alex; 20190823
Long long long long long drive north.
Camping in North Browning, Montana
- Alex; 20190822
First stop: Devil's Tower national monument. The right one this time. The one in Wyoming.
During our month long road trip, we ended up at "devil's tower" in South Dakota (which isn't a national monument, but a state park).
Staying in Billings, Montana
- Alex; 20190821
Flying out to Denver, for a camping trip in Glacier National Park.
- Alex; 20190820
WTF: The CEOs of nearly 200 companies just said shareholder value is no longer their main objective. ``Investing in employees, delivering value to customers, dealing ethically with suppliers and supporting outside communities are now at the forefront of American business goals.''
Ok, not sure I agree with any of this, and I'm amazed some CEOs would be on board with this... actually, probably not ``surprised'' as CEOs are employees, and ``investing in employees'' means what exactly? Probably means rewarding officers of the company at the expense of shareholders :-/
- Alex; 20190819
Concept space: how do we define things in concept space? Statements that partition the concept space into parts that are true vs other parts. For example, saying "even number" partitions the concept of all numbers into those divisible by two. The statement "divisible by two" also needs clarification, as divisible needs to have meaning, as well as the number "two".
Everything written down expresses a concept. Like every sentence partitions the concept space where that statement is true vs other. A story is essentially the sections of the concept space that are true for all sentences in the story. (ignoring tricky statements, such as "this statement is false").
First thing is first: there are more concepts than there are statements to express them. The argument for this is identical to the one that is used to prove computability: there are more subsets out there than there are programs (or sentences) to express them.
But of the concepts that are expressed as sentences, what can we say about them? For one, the concept should not be contradictory: statements should create a concept sub-space---if statements are contradictory, the space will be empty.
- Alex; 20190818
Opinions on continuous entropy, Limiting density of discrete points.
I'll just go right out and say it: there's no such thing. A continuous distribution has infinite (or undefined) entropy---any estimation from finite samples will produce a quantized (discrete) entropy, which is not the same thing.
The key here is that the quantization boundaries (quantization buckets) matter---entropy will reflect this ``arbitrary'' choice, and not some underlying distribution property. For example, if you pick 1000 buckets, then after enough data has been collected, the entropy will be related to one-in-1000 choice, or about 10bits.
The adjustments (such as counting number of zeros, and adding log of that to entropy) also do not produce a correct result.
How then to measure entropy (or information content) of a continuous distribution? Perhaps there's no objective way to do that?
- Alex; 20190817
Long speculation ahead:
``The second law of thermodynamics states that the total entropy of an isolated system can never decrease over time.''
So what exactly is entropy and time?
If we take a shuffled deck of cards, and shuffle it again, did we increase entropy? (perhaps, but because we can't tell the difference between before and after, perhaps not?)
What if we start with an ordered deck of cards? Shuffling obviously increases `disorder'.
How would we measure this process? We'd need to capture some notion of order before and after shuffling, and then compare them. If we cannot tell the difference (both orders look `random') then entropy hasn't increased---if we can (ordered deck becomes random-shuffled), then entropy increased?
If the original deck of cards isn't ``random'', but was actually generated by a pseudo-random shuffling, would shuffling it leave entropy the same? It seems like it would increase entropy if we can somehow attribute more `order' (even if it's pseudo-random) to the before state than afterwards.
But now it gets arbitrary: shuffling will increase entropy according to our knowledge of the initial state (we're biased towards what we consider ``order''). But by another measure the deck may become more ordered. Yes, the number of orderings is astronomical, but the key is that our memory of the before state is just an arbitrary arrangement, and any shuffling from there will take us farther away from that original observation.
Which is why it seems the farther in the past we look, the more `ordered' things were. Perhaps it's just our perspective playing tricks on us.
And what if we remove memory: we cannot remember (nor compare) the past state to the after-shuffling state. Does entropy increase? No. So memory is key to entropy.
How about time? If entropy doesn't increase without memory, then the concept of time shouldn't make sense without memory.
So the second law of thermodynamics should read: recording the order of things consumes memory, and when that order changes, more memory will be consumed to record those changes. There's no time except for such changes.
Where is all this being `recorded'? (where is that memory?). In the positions of things---if we can observe the universe now, and play it back to determine what it was like in the past, then we effectively have a `memory' of what it was like. Where does that leave the universe before the big bang?
If there's no time without memory, then what does that do to the speed-of-light limit?
- Alex; 20190815
On way to Lake Placid decided to check the weather, and it would rain all night long: no star gazing, and wet-tent camping---so decided to turn around and head home instead.
- Alex; 20190812
We're bored of sitting home again: went to camp by Seneca Lake (one of the finger lakes in upstate NYS).
Perseids meteor shower is going on, and we picked a remote place with clear skies to observe it. Set alarm for 3am (once the moon disappears). Got up, wondered out of the tent into the dark forest, got to a clearing, and lied down on the ground. The milky way was amazingly clear. Saw about 5 meteors (3 very clear ones).
- Alex; 20190811
...and back in NYC :-)
Swapping tires on car at BJs (had appointment since 3 weeks ago).
- Alex; 20190729
...on the way to Delhi airport.
Dinner at Amrik Sukhdev.
- Alex; 20190728
Another day trip to Chandigarh; dinner at Barbeque Nation.
- Alex; 20190726
Day trip to Chandigarh.
- Alex; 20190725
...and back to Ambala.
- Alex; 20190724
We're all doing a trip to Kasauli (it's a short drive from Ambala).
So... compared to Shimla... Kasauli is... well... sucks. Definitely not a place to visit twice.
Booked hotel via hotwire.com, and got a cheaper deal than the locally quoted price (though hotwire.com said we'll get Sheraton or Hilton, and we ended up getting ``Kasauli Exotica''; which is probably the best hotel in that area).
Kasauli has a lot of monkeys! Tried to visit Manki Point, but apparently foreigners aren't allowed---it's near or on a military base, and they check passports :-/
- Alex; 20190723
Feeding folks in a temple.
Big people (adults) get big portions, kids get smaller portions. I must've served a hundred folks that day.
Awesome food. Hot weather.
- Alex; 20190721
- Alex; 20190720
Whatever food poisoning kiddo has we all got :-/
- Alex; 20190717
Catching the morning Amritsar Shatabdi from Delhi to Ambala.
- Alex; 20190716
Wow, it sure rains heavily here.
Missed Shatabdi train to Ambala, ended up spending night near train station.
Kiddo not feeling well---probably food poisoning.
- Alex; 20190715
Spending a day in Gurgoan, at the Golden Tulip.
Kiddo loves the pool :-)
- Alex; 20190714
Flying out to India.
- Alex; 20190713
Installed QuakeII, and ended up playing for a few hours. That game is damn good. The rail-gun mechanics are awesome.
- Alex; 20190709
The tire is completely flat :-/ It's amazing it lasted all the way till home.
Pumped it up, found the hole with soap-water. It's actually 2 tiny holes right next to the plug I put in last week. My guess is the rubber plug was old (not sticky enough), so once the rubber cement dried off, it started leaking. Anyways, ordered a new set of tires---will install after the trip to India. (It was time to replace old tires anyways, they have over 80k miles on them).
Finished Zucked: Waking Up to the Facebook Catastrophe by Roger McNamee. To summarize this book: FUD. FUD. FUD. Also, stop using Facebook. That's pretty much it. Not saying the book doesn't have good points, or that any of them are wrong, or incorrect. But you shouldn't expect some things of corporations. I don't have a FB account (nor own any shares), but even I know they monetize every bit of data they have on everyone. That's just business. If they don't put interests of shareholders first, they won't be in business for long. Yes, as individuals, we can choose not to buy their shares, or use their services (giving them data AND eye-ball-hours).
- Alex; 20190707
Spent the day driving home. The tire appears to be holding just fine.
- Alex; 20190705
There is an interesting must-see place near by: Hopewell Rocks. We got there just in time, walked in the sand/mud, saw the rocks at low-tide, saw the tides (which apparently rise at 1-foot per minute!). Saw the park folks close off the beach due to the tide, etc. All in all, this is the best part of this visit to Canada.
Next "small" stop is Reversing Falls. This is nice, but not worth the stop (except maybe as few minutes of rest from driving).
Entered US a bit late, and there's 50-miles of nothing. Next "big" town is Bangor. Found a campsite on the way (Parks Pond Campground) and spent the night. They had July 4th fireworks over the pond at 9pm :-)
- Alex; 20190704
Realizing that Nova Scotia is essentially a tourist trap. Things to see include lighthouses, cities, restaurants, museums, etc. (this is from the information desk at the visitor center).
So we picked a lighthouse and went there: Peggy's Point Lighthouse. It's nice. The rocks around it are nice. The whole place is packed with tourists.
Low-tire-pressure-indicator-light. Urgh. The same tire I plugged last week. 20psi. Good thing we had the tire pump with us. Got it back to 32psi, and moved on.
The next stop is Halifax, to a deceptively named: Point Pleasant Park. It's pleasant. It's nice. I jogged the circumference of the whole park (my guess a mile or two). Nice jogging park.
That's that for Nova Scotia. If you know of any other must-see things there, let me know.
Ended up in a hotel back in Moncton, mostly because hotel prices are generally outrageous.
- Alex; 20190703
Driving to Canada, and then through Canada. It's one big place!
Staying in a hotel in Moncton, just outside of Nova Scotia.
- Alex; 20190702
After a week of being back, we got bored. So off we go on another road trip: Nova Scotia,
First stop: Acadia National Park, where we got an awesome campsite, overlooking the ocean.
- Alex; 20190701
Trying out Overload. Seems to be a pretty faithful recreation of Descent. Seems awesome. Level design sucks though---but mechanics are awesome.
- Alex; 20190630
Today is officially my last day at Goldman. It sure was an interesting year. I was head of Americas compliance data---so got exposed to everything. Learned a lot about front-office and back-office data and operations, the US market making desk, treasury operations, etc.
- Alex; 20190629
Fixed the flat tire via `the plug'. I've seen it done about 10 years ago, and had the kit in the car forever. So took off the tire, located the puncture (it was a big screw), used that tool to make the hole bigger, put rubber cement over the plug, then shoved the plug into the tire... used a bike pump to inflate the tire. So far so good. Drove 30 miles on that tire already, and it seems to be holding.
- Alex; 20190627
Got a flat tire. In NYC! After a month of driving across the country and back, had to come back and get a flat right near home!
- Alex; 20190624
After a month on "the road", we're finally home.
9281 miles. $1263 on gas (33 fill-ups). $892 on hotels (16 nights; primarily via hotwire.com). Probably ~$200 on tent-camping (cash payments). $482 at Walmart. $126 at McDonalds. $23 at Subway. Probably ~$150 at other food places.
- Alex; 20190623
The FairBridge hotel in Wickliffe, Ohio wins the award for being the worst-hotel-of-this-trip. Yes, even worse than Motel6. Motel6 is cheap, styles itself as a cheap hotel, so any expectations are null. FairBridge modeled itself as a 3-star hotel---but it's actually worse than Motel6 in just about everything.
It's within driving distance of Niagara Falls, which is where we went.
Niagara is awesome when the weather is awesome. The place was packed. Very hard to find parking. the main parking lot was closed, so we got a spot in the overflow lot on the other end of the island. About half a mile walk from the main waterfall.
Did a loop walk around. Clicked pix by Tesla statue. Did BBQ right there. Day well spent.
At the end, decided that we should probably drive home instead of staying someplace on the way.
- Alex; 20190622
Heading to Cuyahoga Valley National Park. It's the last national park that's on the list.
It turns out it's not a very interesting park. There are a few places to see waterfalls (we visited a few). That's about it.
Kiddo loves running around. He ran down/up all those stairs in all the walkways around the whole park.
Ended up in a hotel in Wickliffe, Ohio. Very strange hotel (FairBridge). Their pool closes at 7pm, and opens at 12pm. Pretty much guaranteeing that anyone who checks-in in the evening won't be able to use the pool :-/
- Alex; 20190621
Due to rain, decided to skip Cloud Gate in Chicago.
Heading directly to Indiana Dunes National Park. It's the newest national park this year. We ended up in the "state park" by accident. After finding the visitor center, did the scenic drive along the coast.
This is my first time at the lakes. It's amazing. The beach looks identical to an ocean, but if you taste the water, it's fresh. Non-salty. There's no sea-water smell either. It's pretty neat.
On the drive to the hotel, windshield wiper fell apart, so we made an ``emergency'' stop at walmart to replace those. These were the original ones that came with the car---the 10-year-old wipers. New wipers are awesome, they actually lean the windshield, almost forgot how clean windshield looks like!
At the end of the day, ended up in a hotel in Elkhart, Indiana.
- Alex; 20190620
Austin, MN is apparently home of SPAM. They even have the Spam Museum. So had to drive by that on the way to Chicago.
Another long day of driving through fields and fields of stuff.
Ended up in a hotel in Itasca, Illinois.
- Alex; 20190619
...And so begin the most boring few days of the entire trip. Driving from South Dakota to Chicago (cloud gate).
It's one long drive. It just keeps going and going and going. There's nothing there. Scenery is just the same---forever. I90 sucks.
In a work-zone (1-lane reduced-speed no-overtaking zone), a truck overtook me using the shoulder(!!!). While he was doing that crazy thing, a rock flew up and cracked our windshield :-/
Right out of the work-zone (about 11 minutes after he overtook me), we saw that truck pulled over by the state troopers. I really hope it was for that overtaking (and not for speeding, which he was).
After a long day of driving, ended up in a hotel in Austin, MN.
- Alex; 20190618
Headed out to Devil's Tower in South Dakota.
After hours of driving, got to Devil's Tower in South Dakota, and realized that the correct Devil's Tower is in Wyoming. Yeah :-/ So much for "being on the way" and searching google maps by name.
Heading to Mount Rushmore, which is surprisingly close to the South Dakota's Devil's Tower.
It's raining, so clicked some pix, and ran back to the car.
Next stop: Badlands National Park. It finally stopped raining. Weather is awesome. We used the south entrance into the park, and drove the large chunk of the dirt rim road. Saw lots of bison, etc.,
Were planning to camp by Badlands Campsite, but that site was full, so ended up camping about 5 minutes out of the park on the edge of I90.
- Alex; 20190617
Decided to take a vacation from our vacation: staying in this Ramada hotel for a day and ``doing nothing'' the entire day.
Nothing includes: going to the pool, visiting the few kid friendly parks in the city, napping during the day, watching TV, and generally just "doing nothing".
Day well spent in Gillette!
- Alex; 20190616
||Or, you can directly go to a desired entry.|
|NOTICE: We DO NOT collect ANY personal information on this site.|
||© 1996-2019 by End of the World Production, LLC.