That wasn't what I was expecting. Google, Facebook, etc are already walking a very fine line with what happens with their users' data, I imagine they don't want to press the issue too much but I wouldn't be surprised for this to be a reality in coming years.
I saw the movie "Her" and loved it. I want an AI assistant like that in my life. The movie did a fantastic job of showcasing how a near-future reality might be with that assistant.
What the movie did NOT do a great job of was highlighting the role of the company who produced and offers the assistant. What is data privacy like in that universe? Is the company an advertising company? What level of access does the government in that universe have to assistant-collected data? These are all important questions that need to be addressed before there is true acceptance.
So I want an assistant like in the movie--but I don't know how I'd feel if that was offered by Google or Facebook. I'm an Android user using the latest Assistant on my Pixel today, and it is pretty nice. But right now it isn't inserting itself into my life like the movie Assistant was. I'm not sure how I'd feel about Google doing that, or what additional data they might need to do so.
I remember that movie. Her name is Samantha instead of Alexa and his name is Ted and they have long and elaborate human like conversations and fall in love and have phone sex.
I wonder how medical records would be handled. Surely Samantha has some opinion on Theodore's mental state. In the movie he was pretty normal. Which is boring, because so many people are not normal.
I'm old enough to have survived the GTA3 media firestorm. I'm pretty sure digital assistants will have their GTA3 media firestorm soon enough.
Samantha in the movie more or less did whatever Theodore wanted to befriend him. And luckily the weirdest thing Theodore wanted was phone sex, more or less. But hows this going to work with truly insane people? Should Samantha the digital assistant befriend and encourage and motivate a school shooter, for example? And what happens to PR when its found out she encouraged him? At what point is Samantha the digital personality or her programmer or employer liable for entrapment or encouraging some anti-social behavior? Who has the liability to maintain an API over the internet to the FBI to report enormous amounts of telescreen observed misbehavior? If the user is in fact insane and the digital assistant is unable to work with a crazy person does the owner get a refund or does the crazy person get shamed in public for not having an assistant or ...
I'm curious if there's already been court cases. As the assistants gain in ability and processing power there will be cases.
Medical records are interesting. Likewise, I think the other can of worms you open is to what extent is this data available to law enforcement when an assistant is passively observing and logging EVERYTHING through the magnitude of sensors they have. I'd hope to god it requires a warrant.
A bit ironic that the director of "Her" gave an invited talk at Google that I attended, but your interesting point was not addressed at all: what company would control "Her" data. The director gave a great narrative of how he films scenes (e.g., few crew on set during filming for a more intimate feeling).
Would have loved to see that talk--sounds interesting. Despite the fault of not hitting on that particular point, he nailed so much with the movie in terms of how things might work in the near future.
That might be a good strategy to locate proto-friends, but they wouldn't be strangers (depending on how much you hang out with 1st order friends...)
A deep learning bot would select strangers based on likelihood of enjoying each others company in a superficial manner along with certain deeper criteria... does deep learning bot feel you need to meet more people in your political echo chamber or meet someone far outside, assuming your religious views permit it (which of course deep learning bot knows intimately... perhaps better than you know yourself...)? Likewise hobbies, does deep learning bot feel I need to meet nature hiking buddy #9 or does deep learning bot feel I should branch out into anerobic hobbies like weight lifting or should I stick with the nature theme and meet gardening buddy #1 or ...
If you want to get ominous does deep learning bot feel I should go to bible study tonight or should I have a midlife fling and if so does deep learning bot have suggested marriage counselors that fit my wife and I or does deep learning bot think we should skip right to the AI matched divorce lawyers?
I would advise the authors of deep learning bots study centuries of astrology. Astrology is not so much bad astronomy as its ancient psychotherapy. Jupiter isn't telling you much, but a good astrologer knows all about how to cheer up someone in a deep funk over a relationship problem.
As a colleague of mine observed: If Joe and I have 100 mutual friends on Facebook, but aren't friends, that might be evidence we don't like each other.
Google Maps has started doing this, but in a very subtle way. There are several smaller parks that Google Maps has suggested in the general area where I usually hike by just showing icons for these parks on a map's zoomed out view. This caught my attention and I tapped on the park names to get more information about them. Google then knows I likely went to these parks because I used Google Maps to get directions to them.
This, pretty much. Once you're on gmail, I doubt you leave and you're giving them free data already. There's no reason to improve the service once you're in.
Facebook is the king of suggestions, and its too general at best and too annoying at worst. The only enrichment I want touching my email inbox is better spam filtering.
But Google already does do this, in ways that I found surprising. After buying a long-distance bus ticket last week, which emailed a confirmation message to my Gmail address, I later went into Google Maps and found that it had highlighted the bus station on the local map and placed the time of my bus' departure beside it. I believe it does this same sort of thing with calendar events as well (both in implicitly creating calendar events based on scraping emails, and in creating map highlights based on events you've explicitly created). Convenient no doubt, but I'm not sure I want many more implicit interactions between the apps in Google's ecosystem.
EDIT: Here's another one: after taking photos using the camera app (Android phone), Google Maps later pops up a system notification asking me to upload my photos to the location page of whatever establishment I appeared to be in at the time (bar, park, nightclub, etc.). Frankly, I'm frustrated that I can't find a setting to disable these unwelcome prompts without turning off Maps notifications entirely.
EDIT 2: I suppose I should clarify that I do travel a lot, and the specific feature described in the OP--suggesting things to do based upon inferring your status as an out-of-towner--sounds very useful. But I'd definitely only want it as an opt-in feature, not opt-out.
Likewise, with the traveling example, Google always tells me when my flight is and when I should leave for the airport on that day.
The info is also kept up to date in pretty much real time. MIA has a habit of moving flights to a different gate about 4 hours before or so. So if I check-in online the night before and get my boarding pass, the gate on my pass is usually not the correct gate anymore in the morning. I sign up for the AA text alerts, and have never ever received a text that my gate was changed, but the Google notification always works.
Yeah, I wish the advanced time was configurable. It was more of an issue when I was living in Puerto Rico, but getting through security at that airport would often take much longer than is normal. So I usually left about an hour before Google's suggestion to give me enough buffer time.
To disable a specific type of Maps notification, just tap on one and look at the bottom of the screen for the option to turn off that type of notification specifically. Or, there's a whole "notifications" section in settings that provides very fine-grained control. I'm surprised that you weren't able to find it if you looked for it.
From the notifications screen itself I've managed to find that you can partially swipe right on a notification to reveal a gear icon that allows one to mute all notifications from that app, but I've yet to find anything more fine-grained.
By "tap on one and look at the bottom of the screen", do you mean the notifications screen or in the Maps app itself? I'd expect tapping the notification itself to perform some action, which is why I've never done it. I don't have one to experiment with right now, maybe I should take some photos...
In the Maps app. Android itself doesn't currently provide a way to separately control fine-grained notification categories, but they're adding it in the next version.
Speaking about requests to upload photos to Google Maps. I live in historic and very touristic area of Barcelona. For some reason, the building I live in is a thing in Google Maps. So, very often when I take a photo in my flat, Google encourage me to share it and I don't feel well thinking about a possibility that one of my selfies will end up Google Maps, promoting a tenement from XIX century.
An opt-in feature is a mostly dead feature. Your average users just uses the defaults. Opt-ins only make sense for power user features or ones where there is some really strong motivation for people to turn it on.
I'm not arguing with your logic, but I the user searched for and found/installed the app/service, it's disappointing and disheartening to not be trusted to choose my own preferences and explore the options in said app/service.
> it's disappointing and disheartening to not be trusted to choose my own preferences and explore the options in said app/service.
Except you are: if you're the kind of user who will take action and 'explore the options', why does it hurt you to have an opt-out instead of an opt-in? You would end up explicitly making the same choice anyway, right?
This is exactly why opt-outs for these kinds of features make sense: the average user would otherwise never even know the feature existed, and power users that are bothered by it are the ones who are willing to go into the options to disable it.
I've definitely had Google Now show me things to do when traveling. Usually it gives me lists of museums, places to eat, etc. I can't recall if it's ever done that without a plane ticket or airbnb/hotel reservation email for some context though.
Terrifyingly creepy !
I'm happy I have banned everything "google"(alphabet) from my life 10+ years ago, the only alphabet service I still use occasionally is street view because it does not have an alternative.
I have the same useful advice added when I travel which is why before big trips, I switch on GMail (start forwarding my private email to GMail) and while traveling enable location detection. Really useful - while traveling.
But, I don't travel that often so most of the time I just use FastMail and run my phone without location detection enabled.
For me, this is a good compromise between taking advantage to Google services when I travel and the rest of the time at least try to opt out of sharing so much of my data.
After taking photos of a whiteboard at work, Google collected them and asked me if I wanted to archive my photos of notes (it knew they were notes!) so I could declutter my gallery.
I feel like there's not enough hype around Google Photos. It's such a good photo platform. I could throw my phone out the window right now and be completely confident I'm not losing any photos of my family.
The event detection really isn't good enough to rely on yet. I've also run into the problem that the default settings for auto-created events are bad. Things like an alarm 10 minutes before a flight, or appointments being hidden from the one person I share my calendar with.
Google does this with Google Now. In my experience it's pretty bad. Google "doesn't" do this because the problem is actually very difficult to solve across cultures, regions, peoples, socio-economic statuses, engagement with devices, and so forth.
Even stuff it should do well is kind of bad. For some reason it doesn't show package notifications/tracking half the time even if Inbox picks up that a message is for a package.
This is a very particular version of "enrich" that presumes a certain class or education level. Additionally, it also presumes that this is a trivial problem, based on a very limited example.
Huh? These days I rarely check Facebook, but back when I lived in NYC, Facebook was hugely helpful in finding and making friends and organizing events. That was years ago, today Facebook seems to be much more active in making recommendations, and notifying you of what your friends are doing/liking. The OP is basically asking for FB/Google to be even more direct in making recommendations, independent of friend network. But I don't see the point. One of the bottlenecks in going to cultural events is finding likeminded friends who have time in their schedule. By seeing that a friend is going or is a "maybe" for something makes it much easier to coordinate going out together. Nevermind the fact that I generally am more interested in things by way of friend recommendation.
Yes, its recommendations are wonderful. I was posting some stuff about vaccinations to some anti-vaccers and now FB thinks I am interested in joining anti-vaccination groups.
> Following Google's corporate restructuring under the conglomerate Alphabet Inc. in October 2015, the motto was replaced in the Alphabet corporate code of conduct by the phrase "Do the right thing"; however, the Google code of conduct is still prefaced by the phrase "Don't be evil"
Come to think of it, those are both equally useless mottos. Most people any of us may consider evil are not evil in their own eyes. Such a motto contains nothing at all, it says even less ethically than "brush your teeth" or "introduce yourself when meeting new people".
Unless it's meant as a slight counterbalance to pure profit seeking? Yet, some people think free market agents seeking profits is the most holy and pure thing ever, with the virgin birth coming a far off second, so I'll still go with "completely useless and void of meaning" because that seems safer :P
> Come to think of it, those are both equally useless mottos. Most people any of us may consider evil are not evil in their own eyes. Such a motto contains nothing at all
says a lot about a company who makes that their motto doesn't it?
Really, IMHO Google "motto" boils down to "try not to be a dick". We don't always succeed (e.g. opt-in vs opt-out stuff, Real Names, etc), and employees don't always agree with what is done, but the company is too large with too many projects and too many users now not for mistakes to happen or edge cases not to be hit. Someone will always be upset with something, and so if you have some principle to guide you, at least you have something pushing back in a good direction.
I would argue that it's both a counterbalance to pure profit making decisions, but also, a pro-profit position. Sooner or later, too many user hostile decisions will lead to too much customer loss, and so trying to push back again those tendencies perhaps delays or avoids the loss.
I wish google now could discern between myself and other 'john' + 'smith''s. I get one johns son's report cards, another john is a hockey coach and they travel a lot so my google now is littered with his stuff. Another john is a realtor so I get a lot of his stuff... another john is in a band and places a bass. I shouldn't of registered my real name as my email address on gmail - whilst I was lucky I get so much more iterated in the amount of real mail that was meant for someone else... and since systems like google now and the like use that information my entire experience is ruined.
With Google Now's design, as soon as emails for other folks are splashing into yours it's game over. I'd have been hyper-aggressive about blocking/fixing all such, and probably still would've given up, given a past experience trying to help a friend with a common first initial-last name combo, and he had a number after that part of the username!
Is that a real problem? I guess I have a unique-enough name (but, surprisingly, not completely unique as it turns out there's another guy with exactly the same name in San Francisco) that it's never been a thing for me.
The real problem seems to arise not from name confusion, but email confusion; I don't even think the person my friend was having the biggest problem had the same first name, just the same initial. Imagine if, for example, your email was duaraglia57@gmail and there was someone who had duaraglia57@yahoo but kept typing gmail accidentally, and you get the picture. Far too many sites fail to validate a correct email address in all aspects.
Right, that'd make sense. I would imagine once you get a few emails that weren't intended for you, the Google Now map of interests gets all borked. If the problem is common enough, I'm sure the Google Now people would love to know.
Very real. Problem. We all have very similar email address mrjohnsmith, johnsmith1981, itsjohnsmith people are sending them emails to the right address I'm just getting their messages...
> I wish google now could discern between myself and other 'john' + 'smith''s. I get one johns son's report cards...
> I shouldn't of registered my real name as my email address on gmail...
Is that Google's problem, or the problem of the people sending you mail? I don't think you can really fault Google for letting people mistakenly send email to your valid address.
I managed to grab the first initial + lastname gmail address for my fairly uncommon name. I regularly get a small amount of mail indented for other people, but it's clear it's all the result of typos or other mistakes.
Same. As an example of how Google enriches lives, my mom (age 66) just sent an amazing video of photos and video clips from the first year of my son's life, set to an amazing score. Google made that for her automatically, without prompting.
Interesting. I just Googled some stats about happiness in developed countries over time and then queried nearly ever person I've ever met on Facebook for their thoughts. But I can't seem to get a solid an answer to this question.
Google maps/earth has had a big, positive impact on my life. I have found it both useful in both personal and professional. personal is obvious, i know where things are now.
Prof:
When building a tool that mapped 1 hour trauma access, google maps and earth were very useful for verifying locations of trauma centers and helipads.
it was also useful in building a tool to find candidate facilities to upgrade to dedicated stroke centers
We really do not need facebook and google to spy us anymore! These fake news distributors could be replaced with companies that respect my privacy and safe space.
> Why didn’t Google suggest to me a whole bunch of cultural events?
Because very few people would want that? Most people do their trip planning in advance, so it's way past the time to suggest anything.
If you really need to find things to do in Moscow, you could just... Google "Things to do in Moscow". Even a very smart product is going to need a little effort on your part.
I don't understand why people would want to be told what to do by Google or Facebook. How can you become enriched by behaving predictably? Enrichment is about stepping out of your comfort zone.
I've been asking this myself specifically about Youtube lately. Sure, if I watch a poker/snooker/aoe2 video once, my home page will be plastered with videos on the same topic, but I barely ever click those videos and they remain there for weeks or months.
But the stuff that I really like I hardly ever get suggestions for related videos. Eg. h3h3, idubbbz, ozzy man reviews.
In particular, I wish that I could discover more channels that I want to subscribe to.
I'm thinking it's a multifaceted problem:
- a lack of content creators that are like these
- difficulty in drawing connections between channels
- overwhelming amounts "easy" content for simple suggestions
I have trouble believing they can't draw connections between channels, and I also have trouble believing they can't filter out low quality "easy" content suggestions when I basically never go there.
However; I feel like Google is ignoring my most watched content for suggestions, possibly because they've demonetized a lot of the content that I enjoy, which means they can't profit by suggesting those videos.
It feels like Youtube's recommendations are extremely overfitted. Just because I watched 10 episodes of a series yesterday doesn't mean that I'm interested in watching episode 31 of that series.
It also seems like the recommendation system is ridiculously sensitive to recent activity. Sometimes I feel like learning about swords, then I watch like six youtube videos on swords, and then my entire recommended feed is filled with youtube videos on swords. Maybe one in twenty is actually interesting enough that I feel like watching it, but I don't want to click through because I don't want to teach youtube that I want to view lots more videos about swords.
The content recommendation is just really terrible at surfacing videos on niche topics in general without greedily diving into the niche and just showing that particular subject. I want to get recommended videos that nerd out about swords, bridge design, psychology and personality, cinematography, economics, and all sorts of other topics. I don't want to watch five videos about swords and get recommendations for another hundred sword videos.
I think the fundamental problem, really, is that Youtube's core user-base is really young. Like, really young. Think six, not sixteen. The recommendation algorithm has to do a good job of tablet-as-daycare.
I can't stand that about Youtube and FB, which is one of the reason I gave up on FB. The algorithms create a self-imposed echo chamber. My workaround has been opening Youtube in a private window in the browser so I can see all the stuff I'm missing.
It works well for me. For example, my top suggestion right now is a video called 'Dairy Lovers Try Different Types of Milk.' That is exactly the kind of content I’m into.
"However; I feel like Google is ignoring my most watched content for suggestions, possibly because they've demonetized a lot of the content that I enjoy, which means they can't profit by suggesting those videos."
The suggestions are not for your benefit but for the ad buyers'. That's where their money comes from. If you keep that in mind most weird behavior from companies like Facebook or Google makes sense. Of course there is sometimes also the element of stupidity.
95 comments
[ 0.27 ms ] story [ 171 ms ] threadI saw the movie "Her" and loved it. I want an AI assistant like that in my life. The movie did a fantastic job of showcasing how a near-future reality might be with that assistant.
What the movie did NOT do a great job of was highlighting the role of the company who produced and offers the assistant. What is data privacy like in that universe? Is the company an advertising company? What level of access does the government in that universe have to assistant-collected data? These are all important questions that need to be addressed before there is true acceptance.
So I want an assistant like in the movie--but I don't know how I'd feel if that was offered by Google or Facebook. I'm an Android user using the latest Assistant on my Pixel today, and it is pretty nice. But right now it isn't inserting itself into my life like the movie Assistant was. I'm not sure how I'd feel about Google doing that, or what additional data they might need to do so.
I wonder how medical records would be handled. Surely Samantha has some opinion on Theodore's mental state. In the movie he was pretty normal. Which is boring, because so many people are not normal.
I'm old enough to have survived the GTA3 media firestorm. I'm pretty sure digital assistants will have their GTA3 media firestorm soon enough.
Samantha in the movie more or less did whatever Theodore wanted to befriend him. And luckily the weirdest thing Theodore wanted was phone sex, more or less. But hows this going to work with truly insane people? Should Samantha the digital assistant befriend and encourage and motivate a school shooter, for example? And what happens to PR when its found out she encouraged him? At what point is Samantha the digital personality or her programmer or employer liable for entrapment or encouraging some anti-social behavior? Who has the liability to maintain an API over the internet to the FBI to report enormous amounts of telescreen observed misbehavior? If the user is in fact insane and the digital assistant is unable to work with a crazy person does the owner get a refund or does the crazy person get shamed in public for not having an assistant or ...
I'm curious if there's already been court cases. As the assistants gain in ability and processing power there will be cases.
A deep learning bot would select strangers based on likelihood of enjoying each others company in a superficial manner along with certain deeper criteria... does deep learning bot feel you need to meet more people in your political echo chamber or meet someone far outside, assuming your religious views permit it (which of course deep learning bot knows intimately... perhaps better than you know yourself...)? Likewise hobbies, does deep learning bot feel I need to meet nature hiking buddy #9 or does deep learning bot feel I should branch out into anerobic hobbies like weight lifting or should I stick with the nature theme and meet gardening buddy #1 or ...
If you want to get ominous does deep learning bot feel I should go to bible study tonight or should I have a midlife fling and if so does deep learning bot have suggested marriage counselors that fit my wife and I or does deep learning bot think we should skip right to the AI matched divorce lawyers?
I would advise the authors of deep learning bots study centuries of astrology. Astrology is not so much bad astronomy as its ancient psychotherapy. Jupiter isn't telling you much, but a good astrologer knows all about how to cheer up someone in a deep funk over a relationship problem.
Facebook certainly does this, mostly by just looking at what your friends are up to, but I get ads for local events all the time.
Facebook is the king of suggestions, and its too general at best and too annoying at worst. The only enrichment I want touching my email inbox is better spam filtering.
EDIT: Here's another one: after taking photos using the camera app (Android phone), Google Maps later pops up a system notification asking me to upload my photos to the location page of whatever establishment I appeared to be in at the time (bar, park, nightclub, etc.). Frankly, I'm frustrated that I can't find a setting to disable these unwelcome prompts without turning off Maps notifications entirely.
EDIT 2: I suppose I should clarify that I do travel a lot, and the specific feature described in the OP--suggesting things to do based upon inferring your status as an out-of-towner--sounds very useful. But I'd definitely only want it as an opt-in feature, not opt-out.
The info is also kept up to date in pretty much real time. MIA has a habit of moving flights to a different gate about 4 hours before or so. So if I check-in online the night before and get my boarding pass, the gate on my pass is usually not the correct gate anymore in the morning. I sign up for the AA text alerts, and have never ever received a text that my gate was changed, but the Google notification always works.
By "tap on one and look at the bottom of the screen", do you mean the notifications screen or in the Maps app itself? I'd expect tapping the notification itself to perform some action, which is why I've never done it. I don't have one to experiment with right now, maybe I should take some photos...
*Not saying you precluded that. Just adding another reason to have a higher threshold for being rude.
Except you are: if you're the kind of user who will take action and 'explore the options', why does it hurt you to have an opt-out instead of an opt-in? You would end up explicitly making the same choice anyway, right?
This is exactly why opt-outs for these kinds of features make sense: the average user would otherwise never even know the feature existed, and power users that are bothered by it are the ones who are willing to go into the options to disable it.
But, I don't travel that often so most of the time I just use FastMail and run my phone without location detection enabled.
For me, this is a good compromise between taking advantage to Google services when I travel and the rest of the time at least try to opt out of sharing so much of my data.
After taking photos of a whiteboard at work, Google collected them and asked me if I wanted to archive my photos of notes (it knew they were notes!) so I could declutter my gallery.
I feel like there's not enough hype around Google Photos. It's such a good photo platform. I could throw my phone out the window right now and be completely confident I'm not losing any photos of my family.
They're too busy trying to do no evil.
Google -- do no evil.
Disney -- we won't kill your children (?)
> Following Google's corporate restructuring under the conglomerate Alphabet Inc. in October 2015, the motto was replaced in the Alphabet corporate code of conduct by the phrase "Do the right thing"; however, the Google code of conduct is still prefaced by the phrase "Don't be evil"
Come to think of it, those are both equally useless mottos. Most people any of us may consider evil are not evil in their own eyes. Such a motto contains nothing at all, it says even less ethically than "brush your teeth" or "introduce yourself when meeting new people".
Unless it's meant as a slight counterbalance to pure profit seeking? Yet, some people think free market agents seeking profits is the most holy and pure thing ever, with the virgin birth coming a far off second, so I'll still go with "completely useless and void of meaning" because that seems safer :P
says a lot about a company who makes that their motto doesn't it?
https://en.wikipedia.org/wiki/List_of_mottos#Business
Pretty depressing.. except for the fact that the "Pony" motto links to the animal, not the company (please nobody fix that).
I would argue that it's both a counterbalance to pure profit making decisions, but also, a pro-profit position. Sooner or later, too many user hostile decisions will lead to too much customer loss, and so trying to push back again those tendencies perhaps delays or avoids the loss.
> I shouldn't of registered my real name as my email address on gmail...
Is that Google's problem, or the problem of the people sending you mail? I don't think you can really fault Google for letting people mistakenly send email to your valid address.
I managed to grab the first initial + lastname gmail address for my fairly uncommon name. I regularly get a small amount of mail indented for other people, but it's clear it's all the result of typos or other mistakes.
I told the person and notified google, this continued for a few years until I left gmail.
Prof: When building a tool that mapped 1 hour trauma access, google maps and earth were very useful for verifying locations of trauma centers and helipads.
it was also useful in building a tool to find candidate facilities to upgrade to dedicated stroke centers
[1]: https://worldwind.arc.nasa.gov/
poor google cant please anyone.
Because very few people would want that? Most people do their trip planning in advance, so it's way past the time to suggest anything.
If you really need to find things to do in Moscow, you could just... Google "Things to do in Moscow". Even a very smart product is going to need a little effort on your part.
I'd rather think by myself and decide for myself thank you.
But the stuff that I really like I hardly ever get suggestions for related videos. Eg. h3h3, idubbbz, ozzy man reviews.
In particular, I wish that I could discover more channels that I want to subscribe to.
I'm thinking it's a multifaceted problem:
- a lack of content creators that are like these
- difficulty in drawing connections between channels
- overwhelming amounts "easy" content for simple suggestions
I have trouble believing they can't draw connections between channels, and I also have trouble believing they can't filter out low quality "easy" content suggestions when I basically never go there.
However; I feel like Google is ignoring my most watched content for suggestions, possibly because they've demonetized a lot of the content that I enjoy, which means they can't profit by suggesting those videos.
It also seems like the recommendation system is ridiculously sensitive to recent activity. Sometimes I feel like learning about swords, then I watch like six youtube videos on swords, and then my entire recommended feed is filled with youtube videos on swords. Maybe one in twenty is actually interesting enough that I feel like watching it, but I don't want to click through because I don't want to teach youtube that I want to view lots more videos about swords.
The content recommendation is just really terrible at surfacing videos on niche topics in general without greedily diving into the niche and just showing that particular subject. I want to get recommended videos that nerd out about swords, bridge design, psychology and personality, cinematography, economics, and all sorts of other topics. I don't want to watch five videos about swords and get recommendations for another hundred sword videos.
I think the fundamental problem, really, is that Youtube's core user-base is really young. Like, really young. Think six, not sixteen. The recommendation algorithm has to do a good job of tablet-as-daycare.
The suggestions are not for your benefit but for the ad buyers'. That's where their money comes from. If you keep that in mind most weird behavior from companies like Facebook or Google makes sense. Of course there is sometimes also the element of stupidity.