The article is pretty light on details, but it appears Beam is attempting to solve the bookmark problem. But that brings up a really fascinating question: what are bookmarks for, and is there really a problem? If not, what is the real problem?
Before Google, bookmarks were necessary if you ever wanted to find your way back to a website. Now, all you need is some dim memory of a website to find it with Google. Currently, I only use bookmarks like I used to use fast-dial — to quickly navigate to websites without having to Google for them (or for internal work URLs that are not crawlable). The bookmarks bar has become a built-in concise portal for my various go-to sites.
So if the original purpose of bookmarks has been achieved, what is the real problem here? I remember the many hours I used to spend organizing my bookmarks into tidy folders, like a personal Wikipedia of topical links. Many of these sites were blogs that were deprecated by the migration to social media and aggregators. My topical folders are now my Twitter feed and subreddits. So the idea of bookmarks as the “organization of specific topical data” is now outsourced to others.
And yet still, I have a nagging feeling that there is something missing from our experience of the Internet, and that some aspect of the concept of bookmarks reveals that problem.
Beam suggests that it is marking some piece of information as important — not marking a URL, but marking information itself. As information has grown exponentially, I do often find myself feeling like I’m drowning in a sea of important information, because my mind can only remember so much. There’s a nagging feeling that I’m losing important information, and I know that although I can (and do) bookmark URLs, I haven’t used a non-toolbar bookmark in years.
I find myself utilizing various hacks when I find a piece of information I think it important and don’t want to lose: I send myself a link via Slack, or sometimes by text message. How often do I actually revisit those links? Maybe once or twice in years
So there’s the problem identified in behavior: bookmarking or otherwise saving URLs, but never actually visiting those links. Am I the only one who does this? Probably not. When I add a bookmark, it is always accompanied by a sense of futility, but I do it anyways! When I leave a tab open to reference later, I know I’ll probably just close it later without a glance. But I still do it.
The true problem underlying this futile behavior is obvious and agreed upon: information overload. But we can be more precise: information overload that makes all information ephemeral, when some information should be permanent.
When I read an article about Beam, what I want is not only the article itself, but a kind of rap-genius of meta-information I’ve previously marked as important running alongside it. I’d see information about other browsers. I’d see quick summaries of the dead social-bookmarking sites and why they failed. I’d see pithy HN comments about why some people seem to get VC money thrown at them.
It would be like this HN comments section, but the only person commenting would be myself with full recall of everything I’d ever read, happily chirping away at every paragraph, line, and word. It would be a genius version of myself with a photographic memory.
So there it is: the real problem — we are not geniuses with perfect memory. But it would be pretty cool if we were.
If there were a GPT-X continually fine-tuned on everything I had ever read (with meta-data about how long I had read and whether I found it important), then I think it would be possible to at least solve part of this problem — that of not having a perfect memory. But ML has a long way to go until it is capable of producing genius, let alone intelligence.
This is to say that Beam has taken on a problem much harder than “bookmarks”, and it is going to take more than a few million to solve.
> But we can be more precise: information overload that makes all information ephemeral, when some information should be permanent.
On this particular point I disagree, though I love the rest of your writeup. The information was always there, somewhere; you just couldn't really access it. Google and other aggregators have made it possible for us to operate "one level higher". I don't have to remember exactly what Bayes theorem is and how it can be used, I can just remember the type of problem Bayes is excellent at solving, and rely on the internet to bring me quickly up to speed. It _has_ made us all geniuses, but we take it for granted because we've gotten so used to it that it seems obvious.
I think what Beam is attempting here is a similar leap in technology, and it should be interesting to see if they succeed and change our relationship with information. It definitely feels like something is missing, because the internet is too unreliable; if a video is deleted or a link is broken, you're out of luck.
Before Google, bookmarks were necessary if you ever wanted to find your way back to a website. Now, all you need is some dim memory of a website to find it with Google. Currently, I only use bookmarks like I used to use fast-dial — to quickly navigate to websites without having to Google for them (or for internal work URLs that are not crawlable). The bookmarks bar has become a built-in concise portal for my various go-to sites.
So if the original purpose of bookmarks has been achieved, what is the real problem here? I remember the many hours I used to spend organizing my bookmarks into tidy folders, like a personal Wikipedia of topical links. Many of these sites were blogs that were deprecated by the migration to social media and aggregators. My topical folders are now my Twitter feed and subreddits. So the idea of bookmarks as the “organization of specific topical data” is now outsourced to others.
And yet still, I have a nagging feeling that there is something missing from our experience of the Internet, and that some aspect of the concept of bookmarks reveals that problem.
Beam suggests that it is marking some piece of information as important — not marking a URL, but marking information itself. As information has grown exponentially, I do often find myself feeling like I’m drowning in a sea of important information, because my mind can only remember so much. There’s a nagging feeling that I’m losing important information, and I know that although I can (and do) bookmark URLs, I haven’t used a non-toolbar bookmark in years.
I find myself utilizing various hacks when I find a piece of information I think it important and don’t want to lose: I send myself a link via Slack, or sometimes by text message. How often do I actually revisit those links? Maybe once or twice in years
So there’s the problem identified in behavior: bookmarking or otherwise saving URLs, but never actually visiting those links. Am I the only one who does this? Probably not. When I add a bookmark, it is always accompanied by a sense of futility, but I do it anyways! When I leave a tab open to reference later, I know I’ll probably just close it later without a glance. But I still do it.
The true problem underlying this futile behavior is obvious and agreed upon: information overload. But we can be more precise: information overload that makes all information ephemeral, when some information should be permanent.
When I read an article about Beam, what I want is not only the article itself, but a kind of rap-genius of meta-information I’ve previously marked as important running alongside it. I’d see information about other browsers. I’d see quick summaries of the dead social-bookmarking sites and why they failed. I’d see pithy HN comments about why some people seem to get VC money thrown at them.
It would be like this HN comments section, but the only person commenting would be myself with full recall of everything I’d ever read, happily chirping away at every paragraph, line, and word. It would be a genius version of myself with a photographic memory.
So there it is: the real problem — we are not geniuses with perfect memory. But it would be pretty cool if we were.
If there were a GPT-X continually fine-tuned on everything I had ever read (with meta-data about how long I had read and whether I found it important), then I think it would be possible to at least solve part of this problem — that of not having a perfect memory. But ML has a long way to go until it is capable of producing genius, let alone intelligence.
This is to say that Beam has taken on a problem much harder than “bookmarks”, and it is going to take more than a few million to solve.