Wednesday 11 December 2019. I’m on a train from Sacramento to Emeryville to meet up with friends for dinner and am kind of forcing myself to do it. I don’t want to do it because ugh effort but on the other hand, these are all people that I like and like to spend time with and I know I’ll have a good time. So! Dinner it is.
The train ride from Sacramento to the Bay Area/San Francisco always feels weird to me as a Brit:
Sacramento to Emeryville is 57 miles by train; the train I’m on will do this in 1 hour, 37 minutes. It’s 9.6 miles from Emeryville station to the restaurant which at this time of day would take about an hour by public transport or an hour by car. Total travel time? 2.5 hours by public transit, 2 hours by car.
London to Brighton on the other hand is 52 miles. It takes 59 minutes to get there by train, and both stations are actually, you know, in their respective cities. Or you could drive, in around 1.5 hours.
But to put this in context, Sacramento is California’s State Capitol. California itself is practically the same as the United Kingdom, so Sacramento is London. San Francisco is, well… it’s also practically London. London is a bit weird because the U.K. is so small, every single thing apart from industrial manufacturing is represented in London.
Birmingham is the second biggest city in the UK and 126 miles from London. You could drive there in 2h18 or a train will do it in 1h21.
Manchester is the third biggest city in the UK and 208 miles from London. You could drive there in 3h49 or a train will do it in 2h14.
It gets more ridiculous: Sacramento to LA is 386 miles, a 6h7 drive or a 6h16 train ride. London to Edinburgh is 411 miles, a 7h45 drive or 4h20 by train.
America is insane, and don’t even get me started on taking the train from Portland, OR to Seattle.
I am writing this listening to the cover of Blue Monday by HEALTH, from the Atomic Blonde soundtrack.
I’m going to try to do a short one today. Ha! Let’s see how that pans out.
So there’s a Google Assistant smart alarm clock made by Lenovo. It is called the Lenovo Smart Alarm Clock and well, let’s just say that we are going to consider in this section exactly how smart it is, and what that might mean. But first, there is some news about it!
You can tell it is a smart alarm clock because it has software and that software has been updated. The smart alarm clock can do a new thing now and there is a blog post telling us all about the new thing it can, like some sort of puppy that can roll over, but only when it’s connected to the internet.
Lenovo Smart Alarm Clock has a new ringtone called “impromptu”, which:
gives you a ringtone that fits your situation, based on things like the time of day or weather. It’s powered by machine learning technology from our Magenta open source project. For example, if your alarm goes off early in the morning and the weather is less than 50 degrees, you may hear this ringtone. [blog post]
Here’s what I think:
I like the idea of non-explicit environmental cues that communicate information or state about something. I like the idea that I do not have to read something or look at something and comprehend it to know that, for example, it is going to rain today. (Why is it always the example of the weather with smart alarm clocks, or traffic, or… the other things).
I like the idea of ambient information that can help you with your day or help you make decisions without it being explicitly transactional, e.g. “OK Google, what’s the weather like?”
This is different from the iOS 13 feature of the welcome message text in the morning that tells you what the weather’s going to be like and, I don’t know, how many calendar appointments you have that day or when your first thing is.
But here’s what I don’t like, and here’s where my cynicism or I would like to say realism creeps in: how long is this going to last? If I’m going to learn what different alarm tones mean and how they correlate to temperature or likelihood of rain or traffic patterns, then how long do I have to learn that correlation? And once I’ve learned it, how long until the feature or product is end-of-lifed?
What I’m getting at here is the two tensions of rapid product development and experimentation in the marketplace - hey, we’ve got an idea, let’s deploy it (I mean, we can) and see if it’s any good and hey, if the product fails, too bad - on the one hand, and on the other, the fact that it takes time for us to learn things, and my assumption [citation needed etc] that if we’re talking about grokking and internalizing cues like “a subtly different ringtone means a different thing” then… that’s a lot!
For example, how long does it take to get used to something in a new environment? To birds singing in a different way outside your window? To having birds at all singing? To a different pattern of light? Would you even be able to correlate a different ringtone with a traffic slowdown? These things feel like maybe they benefit from being slow.
Other approaches (that I’m sure many people, and also some readers) include ambient lighting, and none of this is impossible, right? You could set up a if-this-then-that bridge that mungs weather data into your Phillips Hue and know by the color of your accent lighting whether it’s going to rain later and you need your umbrella (again, always these examples!)
But one thought I have is that those settings are idiosyncratic, right? The Lenovo Smart Clock could give you options for how you want to be notified about environmental conditions, but that would only work for you. Say a guest comes and stays — how do they know what the alarm clock means when it… speaks in this way? When it emanates/radiates information in this way?
I‘m thinking that natural cues resulting from the, uh, natural world, produce a shared vocabulary. The sun shines through curtains or blinds more; it is summer. “Everyone” can experience that. There’s nothing here about, say, a universal grammar or language for information-about-the-world. Could there be? The thought of there being a land-grab for such by Google or whomever is frankly a bit sickening, even though the existence of a universal understandable grammar would be quite nice, you know? Who the hell would even come up with a mapping for such concepts?
I also have some snark about the Lenovo Smart Alarm Clock, and I do that by linking to its FAQ page on its support site.
Now, there are many frequently asked questions on that page, or at least questions that are presented as “frequently asked”. We have no way of telling if those questions are actually frequently asked, or if they have been anticipated by the product team as being potentially frequently asked, and they’re putting together a page just in case anyone has questions like:
how to set the alarm?
how to set the alarm volume?
is the screen a touch screen?
Now, I may come across as a cynical bastard here, but I would have thought that of all alarm clocks, a smart alarm clock might be the kind of alarm clock that is least relevant to a support article about setting an alarm. Granted, the answer to that frequently asked question is “Talk to the Lenovo Smart Clock. Say something like “Hey Google, set an alarm for 8.a.m.” and yes I have seen horrendous Sony alarm clock manuals. Or horrendous Sony manuals in the first place. But this is supposed to be a smart alarm clock. Why call it smart if, I don’t know, it can’t answer questions itself about how to set an alarm to an extent that the product team are so worried about how you might set an alarm there’s content about it?
I may also be being too harsh. I can imagine that someone else who does not own a smart alarm clock, say, me, has chosen to inflict it upon someone who is not me, as a gift. The entire episode has backfired on me because the receiver of the gift is now asking me for technical support, i.e. “how do I set an alarm on this smart alarm clock you gave me” and I, a non-smart-alarm-clock-owner, must resort to googling and relaying to the gift receiver, “well, you say, uh, Hey Google, set an alarm for 8 a.m.” in the hopes that this will help my enemy set an alarm.
Anyway. “Consistent long-term grammars for ambient information radiators”, go read my paper on arxiv or whatever.
Oh right, and there were other apps that did this. The reason why I got worried about these things just not lasting long in the marketplace is because Uniqlo came up with one of these in 2012 and it was discontinued in 2015. So, you know. Sucks to have learned that.
Agent-distributed computing to save the day
In response to Jack Twitter proclaiming that he’d like to call forth some sort of flood and have a do-over for that particular social network, I saw Alex Stamos recommend this:
What I understand Stamos to be proposing here is Twitter running personal daemons - a bit like the encapsulated code+data General Magic/Magic Cap agents, a bit like AWS Lambda-style server-side computing, where you (or someone) gets to define a set of content filtering behavior that would run in an isolated process. I mean, almost as if something like a mail provider was able to let you run your own custom spam-sieve or trainable mail filing system. How novel!
But I do really like the idea of user-defined constrained programs running on server/service infrastructure doing operations on that service data/content, because you can also imagine a marketplace or directory for those services. There’s a lot there!
… and then only two tiny bullets:
For the nth time, “digital transformation” and “innovation” is nothing without change, change always comes with risk, if you want innovation and you don’t want to change anything, just stop wasting time, or be totally OK with wasting money. cf Erika Hall, Scott Jensen. Much executive consulting around this area is as much about getting leadership OK with the idea that changing things necessarily involves risk but also that if you actually want to improve something in any way, things will have to change. One way of articulating this is to get someone to admit, out loud, that they would rather things stay the way they are — and talk in detail about what that is and isn’t, and what isn’t nice about the way things are — because they’re worried about changing things and not getting improvements.
A self-driving truck delivered butter from California to Pennsylvania (I bet a train could do it too) in 3 days and this fits my worldview of “self-driving makes sense in America in all those godawful long boring straight freeway roads going through hundreds of miles of nothingness” and less for “driving in populated centers around frankly insane, inconsiderate human drivers who have a death-wish and have genuinely taken “signaling” to be just a mere suggestion as opposed to a requirement or even something they could do out of self-interest in health and preservation”. This, again, feels like the early sign of “slowly, and then all at once”.
2000 words! Not bad for 45 minutes!
And I’m still on a train! Because it’s America and trains take forever! Maybe American’s actually like trains and the long journey time is simply an expression of love. They just don’t ever want to get off them, so they go as slowly as possible.
Much love to all of you and, as ever, send notes!
Best,
Dan
","body_text":"0.0 Situation report\nWednesday 11 December 2019. I’m on a train from Sacramento to Emeryville to meet up with friends for dinner and am kind of forcing myself to do it. I don’t want to do it because ugh\neffort\nbut on the other hand, these are all people that I like and like to spend time with and I know I’ll have a good time. So! Dinner it is.\nThe train ride from Sacramento to the Bay Area/San Francisco always feels weird to me as a Brit:\nSacramento to Emeryville is 57 miles by train; the train I’m on will do this in 1 hour, 37 minutes. It’s 9.6 miles from Emeryville station to the restaurant which at this time of day would take about an hour by public transport or an hour by car. Total travel time? 2.5 hours by public transit, 2 hours by car.\nLondon to Brighton on the other hand is 52 miles. It takes 59 minutes to get there by train, and both stations are actually, you know, in their respective cities. Or you could drive, in around 1.5 hours.\nBut to put this in context, Sacramento is California’s State Capitol. California itself is practically the same as the United Kingdom, so Sacramento is London. San Francisco is, well… it’s also practically London. London is a bit weird because the U.K. is so small, every single thing apart from industrial manufacturing is represented in London.\nBirmingham is the second biggest city in the UK and 126 miles from London. You could drive there in 2h18 or a train will do it in 1h21.\nManchester is the third biggest city in the UK and 208 miles from London. You could drive there in 3h49 or a train will do it in 2h14.\nIt gets more ridiculous: Sacramento to LA is 386 miles, a 6h7 drive or a 6h16 train ride. London to Edinburgh is 411 miles, a 7h45 drive or 4h20 by train.\nAmerica is insane, and don’t even get me started on taking the train from Portland, OR to Seattle.\nI am writing this listening to the cover of Blue Monday by HEALTH, from the Atomic Blonde soundtrack.\n1.0 The things that caught, etc.\nI’m going to try to do a short one today. Ha! Let’s see how that pans out.\n1.1 Rise and shine, information radiators\nSo there’s a Google Assistant smart alarm clock made by Lenovo. It is called the\nLenovo Smart Alarm Clock\nand well, let’s just say that we are going to consider in this section exactly how smart it is, and what that might mean. But first, there is some news about it!\nYou can tell it is a smart alarm clock because it has software and that software has been updated. The smart alarm clock can do a new thing now and there is a\nblog post\ntelling us all about the new thing it can, like some sort of puppy that can roll over, but only when it’s connected to the internet.\nLenovo Smart Alarm Clock has a new ringtone called “impromptu”, which:\ngives you a ringtone that fits your situation, based on things like the time of day or weather. It’s powered by machine learning technology from our\nMagenta\nopen source project. For example, if your alarm goes off early in the morning and the weather is less than 50 degrees, you may hear this\nringtone\n. [\nblog post\n]\nHere’s what I think:\nI\nlike\nthe idea of non-explicit environmental cues that communicate information or state about something. I\nlike\nthe idea that I do not have to read something or look at something and comprehend it to know that, for example, it is going to rain today. (Why is it always the example of the weather with smart alarm clocks, or traffic, or… the other things).\nI like the\nidea\nof ambient information that can help you with your day or help you make decisions without it being explicitly transactional, e.g. “OK Google, what’s the weather like?”\nThis is\ndifferent\nfrom the iOS 13 feature of the welcome message text in the morning that tells you what the weather’s going to be like and, I don’t know, how many calendar appointments you have that day or when your first thing is.\nBut here’s what I don’t like, and here’s where my cynicism or I would like to say realism creeps in: how long is this going to last? If I’m going to learn what different alarm tones\nmean\nand how they correlate to temperature or likelihood of rain or traffic patterns, then how long do I have to learn that correlation? And once I’ve learned it, how long until the feature or product is end-of-lifed?\nWhat I’m getting at here is the two tensions of rapid product development and experimentation in the marketplace - hey, we’ve got an idea, let’s deploy it (I mean, we can) and see if it’s any good and hey, if the product fails, too bad - on the one hand, and on the other, the fact that\nit takes time for us to learn things,\nand my assumption [citation needed etc] that if we’re talking about grokking and internalizing cues like “a subtly different ringtone means a different thing” then… that’s a lot!\nFor example, how long does it take to get used to something in a new environment? To birds singing in a different way outside your window? To having birds at all singing? To a different pattern of light? Would you even be able to correlate a different ringtone with a traffic slowdown? These things feel like maybe they benefit from being\nslow\n.\nOther approaches (that I’m sure many people, and also some readers) include ambient lighting, and none of this is impossible, right? You could set up a if-this-then-that bridge that mungs weather data into your Phillips Hue and know by the color of your accent lighting whether it’s going to rain later and you need your umbrella (again, always these examples!)\nBut one thought I have is that those settings are idiosyncratic, right? The Lenovo Smart Clock could give you options for how\nyou\nwant to be notified about environmental conditions, but that would only work for you. Say a guest comes and stays — how do they know what the alarm clock means when it… speaks in this way? When it emanates/radiates information in this way?\nI‘m thinking that natural cues resulting from the, uh, natural world, produce a shared vocabulary. The sun shines through curtains or blinds more; it is summer. “Everyone” can experience that. There’s nothing here about, say, a universal grammar or language for information-about-the-world. Could there be? The thought of there being a land-grab for such by Google or whomever is frankly a bit sickening, even though the existence of a universal understandable grammar would be quite nice, you know? Who the hell would even come up with a mapping for such concepts?\nI also have some snark about the Lenovo Smart Alarm Clock, and I do that by linking to its\nFAQ page\non its support site.\nNow, there are many frequently asked questions on that page, or at least questions that are presented as “frequently asked”. We have no way of telling if those questions are actually frequently asked, or if they have been anticipated by the product team as being potentially frequently asked, and they’re putting together a page just in case anyone has questions like:\nhow to set the alarm?\nhow to set the alarm volume?\nis the screen a touch screen?\nNow, I may come across as a cynical bastard here, but I would have thought that of all alarm clocks, a\nsmart alarm clock\nmight be the kind of alarm clock that is least relevant to a support article about\nsetting an alarm\n. Granted, the answer to that frequently asked question is “Talk to the Lenovo Smart Clock. Say something like “Hey Google, set an alarm for 8.a.m.” and yes I have seen horrendous Sony alarm clock manuals. Or horrendous Sony manuals in the first place. But this is supposed to be a\nsmart alarm clock\n. Why call it smart if, I don’t know, it can’t answer questions itself about how to set an alarm to an extent that the product team are so worried about how you might set an alarm there’s content about it?\nI may also be being too harsh. I can imagine that someone else who does not own a smart alarm clock, say, me, has chosen to inflict it upon someone who is not me, as a gift. The entire episode has backfired on me because the receiver of the gift is now asking me for technical support, i.e. “how do I set an alarm on this smart alarm clock you gave me” and I, a non-smart-alarm-clock-owner, must resort to googling and relaying to the gift receiver, “well, you say, uh, Hey Google, set an alarm for 8 a.m.” in the hopes that this will help my enemy set an alarm.\nAnyway. “Consistent long-term grammars for ambient information radiators”, go read my paper on arxiv or whatever.\nOh right, and there were other apps that did this. The reason why I got worried about these things just not lasting long in the marketplace is because\nUniqlo came up with one of these in 2012\nand it was discontinued in 2015. So, you know. Sucks to have learned that.\n1.2 Some smaller things\nAgent-distributed computing to save the day\nIn response to Jack Twitter proclaiming that he’d like to call forth some sort of flood and have a do-over for that particular social network, I saw Alex Stamos recommend this:\nAlex Stamos\n@alexstamos\nHere is one option I proposed, allowing users to choose intelligent agents that run server-side in jails. Would be much more computationally efficient than doing this all over API.\n\nhttps://t.co/oOFOmj5wEX\nAlex Stamos\n@alexstamos\nTwitter should allow people to upload limited service workers to filter/block on their behalf. This would create a marketplace of anti-abuse solutions.\n\nCloudflare was able to build a secure option with Javascript, could also consider LUA or Haskell:\nhttps://t.co/o3foFDD4z6 https://t.co/hxkspowL8A\nDecember 11th 2019\n9\nRetweets\n50\nLikes\nWhat I understand Stamos to be proposing here is Twitter running personal daemons - a bit like the encapsulated code+data General Magic/Magic Cap agents, a bit like AWS Lambda-style server-side computing, where you (or someone) gets to define a set of content filtering behavior that would run in an isolated process. I mean, almost as if something like a mail provider was able to let you run your own custom spam-sieve or trainable mail filing system. How novel!\nBut I do really like the idea of user-defined constrained programs running on server/service infrastructure doing operations on that service data/content, because you can also imagine a marketplace or directory for those services. There’s a lot there!\n… and then only two tiny bullets:\nFor the nth time, “digital transformation” and “innovation” is nothing without change, change always comes with risk, if you want innovation and you don’t want to change anything, just stop wasting time, or be totally OK with wasting money. cf\nErika Hall\n,\nScott Jensen\n. Much executive consulting around this area is as much about getting leadership OK with the idea that changing things necessarily involves risk but also that if you actually want to improve something in any way,\nthings will have to change\n. One way of articulating this is to get someone to admit, out loud, that they would rather things stay the way they are — and talk in detail about what that is and isn’t, and what isn’t nice about the way things are — because they’re worried about changing things and not getting improvements.\nA self-driving truck\ndelivered butter from California to Pennsylvania\n(I bet a train could do it too) in 3 days and this fits my worldview of “self-driving makes sense in America in all those godawful long boring straight freeway roads going through hundreds of miles of nothingness” and less for “driving in populated centers around frankly insane, inconsiderate human drivers who have a death-wish and have genuinely taken “signaling” to be just a mere suggestion as opposed to a requirement or even something they could do out of self-interest in health and preservation”. This, again, feels like the early sign of “slowly, and then all at once”.\n2000 words! Not bad for 45 minutes!\nAnd I’m still on a train! Because it’s America and trains take forever! Maybe American’s actually\nlike\ntrains and the long journey time is simply an expression of love. They just don’t ever want to get off them, so they go as slowly as possible.\nMuch love to all of you and, as ever, send notes!\nBest,\nDan","raw_format":"html","source":"import","summary":"This issue explores the disconnect between California's sprawling geography and the UK's compact infrastructure through a train journey, then pivots to examining how smart devices attempt to communicate ambient information through subtle environmental cues like context-aware alarm tones. The throughline connects physical distance and cognitive effort: just as traveling in California requires accepting vast spaces and travel times, learning to interpret a smart clock's weather-responsive ringtones requires time and repetition to internalize—yet both are increasingly designed for speed and disposability rather than the slower, more natural processes through which people actually learn and navigate their worlds.","reading_minutes":11,"links":[{"url":"https://www.blog.google/products/assistant/rise-and-shine-new-alarm-clock-features-assistant/","normalized_url":"https://blog.google/products/assistant/rise-and-shine-new-alarm-clock-features-assistant","domain":"blog.google","anchor_text":"Lenovo Smart Alarm Clock","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://www.blog.google/products/assistant/new-google-assistant-features-customize-your-alarm-clock/","normalized_url":"https://blog.google/products/assistant/new-google-assistant-features-customize-your-alarm-clock","domain":"blog.google","anchor_text":"blog post","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://ai.google/research/teams/brain/magenta/","normalized_url":"https://ai.google/research/teams/brain/magenta","domain":"ai.google","anchor_text":"Magenta","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://storage.googleapis.com/magentadata/lenovo_smart_clock/keyword_blog_example_alarm.mp3","normalized_url":"https://storage.googleapis.com/magentadata/lenovo_smart_clock/keyword_blog_example_alarm.mp3","domain":"googleapis.com","anchor_text":"ringtone","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://www.blog.google/products/assistant/new-google-assistant-features-customize-your-alarm-clock/","normalized_url":"https://blog.google/products/assistant/new-google-assistant-features-customize-your-alarm-clock","domain":"blog.google","anchor_text":"blog post","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://support.lenovo.com/us/en/solutions/ht508387","normalized_url":"https://support.lenovo.com/us/en/solutions/ht508387","domain":"lenovo.com","anchor_text":"FAQ page","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://www.wired.com/2012/05/uniqlo-wake-up-app/","normalized_url":"https://wired.com/2012/05/uniqlo-wake-up-app","domain":"wired.com","anchor_text":"Uniqlo came up with one of these in 2012","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://twitter.com/alexstamos/status/1204814997248823296?s=20","normalized_url":"https://twitter.com/alexstamos/status/1204814997248823296","domain":"twitter.com","anchor_text":"Alex Stamos @alexstamos Here is one option I proposed, allowing users to choose intelligent agents that run server-side in jails. Would be much more computationally efficient than doing this all over API. \n\n https://t.co/oOFOmj5wEX Alex Stamos @alexstamos Twitter should allow people to upload limite","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://twitter.com/mulegirl/status/1204475284809904128","normalized_url":"https://twitter.com/mulegirl/status/1204475284809904128","domain":"twitter.com","anchor_text":"Erika Hall","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://twitter.com/scottjenson/status/1204780933527691264","normalized_url":"https://twitter.com/scottjenson/status/1204780933527691264","domain":"twitter.com","anchor_text":"Scott Jensen","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"},{"url":"https://www.mercurynews.com/2019/12/10/a-self-driving-truck-delivered-butter-from-california-to-pennsylvania-in-three-days/","normalized_url":"https://mercurynews.com/2019/12/10/a-self-driving-truck-delivered-butter-from-california-to-pennsylvania-in-three-days","domain":"mercurynews.com","anchor_text":"delivered butter from California to Pennsylvania","is_archive":0,"archive_url":null,"archive_service":null,"category":"external"}],"sections":[{"id":1656,"ord":0,"number":"0.0","heading":"Situation report","level":1,"word_count":363},{"id":1657,"ord":1,"number":"1.0","heading":"The things that caught, etc.","level":1,"word_count":22},{"id":1658,"ord":2,"number":"1.1","heading":"Rise and shine, information radiators","level":2,"word_count":1195},{"id":1659,"ord":3,"number":"1.2","heading":"Some smaller things","level":2,"word_count":558}]}![\"Twitter](\"https://cdn.substack.com/image/twitter_name/w_36/alexstamos.jpg\")