Slack's core flaws require topic modeling to fix

...aren't these all problems that emails have too, in more-or-less equal measure?

Of course, Slack needs to be better than email, or there's no point—but I'd posit that it already is, despite being unable to address these problems. The overall interface of a "chatroom" just works better for quick conversations compared to long email chains. Maybe not much better, but better.

As I see it, the problems pointed out in this article are the eternal problems of all communication within a large group—including face-to-face. It's why small companies can be more nimble despite less resources, and why managing the size of a team is so important.

And while the author's suggestions are great in theory, I can't imagine how you'd even go about implementing them. At least, not until AI gets much, much better than it is today.

Slack just need to fix threads. The UI right now is crazy if you are in more than one thread, I don’t understand why you have to click “x more replies” to see new messages - I don’t think they fixed this in the new update either. Right now it makes me (and others) avoid threads which clutters up channels for everyone.

I don't want an algorithm helpfully randomly deciding that someone trying to ping me is "noise". AI is not robust enough for this yet, and never will be.

Even Facebook has enough restraint to not include AI in their messaging app.

I much prefer Zulip to Slack because it forces you to split your convos into topics, making it more like a forum than a realtime chat system. It also intentionally eschews push notifications in favor of keeping everything asynchronous (unless you're specifically @mentioned).

I haven't seen any chat system other than Zulip quite nail this yet, but I'm optimistic that others will eventually catch on. I'm not convinced that human-provided topic labels could ever be replaced by an AI deciding which things should be clustered, but maybe I'm wrong...

Since "topic model" is not explained in the post: https://en.wikipedia.org/wiki/Topic_model

Everyone I work with is part of at least a dozen Slack workspaces with disparate notification settings. I get loads of notifications at times when I can't reply immediately, but there is no system comparable to _not archiving an email_ so I remember to come back to it. I get the convenience of semi-synchronous conversations and being able to engage and catch people up better than forwarding an email chain, but Slack and its ilk are very, very far from replacing email.

> Having AI in your DNA isn’t just a buzzword anymore

I haven't read or seen anything that suggests this, including in this article. If this article is suggesting that all companies have to use AI/ML to survive then that's wrong IMO.

AI/ML is great for a certain subset of problems but I don't think textual inter-human communication is one of them and there are a lot others where AI/ML will not be a silver bullet.

I’m short on real time chat mattering as much in 5 years.

My company is sick of the nuisance, and is working towards real time communication via git & CI/CD, PD alerts, and self guided copy paste of boilerplate framework manifests, without the constant forced check ins to sooth attitudes seeking conformity and compliance in behavior of people who are generally non-hostile.

Traditional office life is falling down just as much because the people that built it are dying off as it is tech.

The masses commuting to offices on the far edges of our towns is a generational thing that the future isn’t obligated to give a shit about.

People like slack because using it feels like "working".

It's not. Everyone who has described to me how their "productivity" went up because of slack I am able to explain away. It's not work, it's keeping up with slow chats and a mess of interlaced conversations. That's not work.

Test: In a popular channel wait until someone asks a question. Immediately then post another question. This should be ok behavior (it is with email). See what happens. The previous one quickly goes away as other people type. What a mess.

Chat isn't how work gets done, it's how people trick themselves into thinking they are "working".

When all you have is a hammer…

I think that Slack has economic problem as well. They are just chat, no matter what they say. And Slack is pricey, too. Once the coronavirus pandemic is over, but the recession/depression hits hard, many IT heads will look for ways to cut costs. Many organizations will dump Slack in favor of other solutions.

slack requires some priorities (better platform support for drawing on screen, less lag on connections like 50mbit down/ 5 mbit up) and for that, they need more programmers who know what they are doing and less product managers who want to "renovate the UI".

The solution isn’t AI. The solution does however require some intelligence to skim over the content and re-label it.

Humans are the engine here. And Slack is already starting to show some promise with things like threads, reminders and actions.

However to be truly beneficial, this has to be baked into the core of Slack’s being; it must support labelling of historical activity, it must treat threads differently, and reminders and actions need to be customisable based on your intended level of interaction with the underlying conversations.

Until Slack gets on top of that, shifting the UI around won’t help.

I don’t mind real time chat honestly, but I wish it were done right, at least on par with email:

No erasing edit history

Open source

Self hostable

The iPhone means the last two won’t happen and people who set their clients to auto load remote images means the first one won’t.

My experience is that basically all types of topic modeling tools just kinda blow.

You might as well use traditional clustering techniques instead since topic models are basically a type of clustering

On Slack, I just mute all the channels, unless I'm active on a project related channel. But having to jump through each workspace to clear out messages is a pain.

One view to see all or every workspace unreads would be great... I'm jumping through hoops just to find the damn unread messages sometimes.

Slack, like Dropbox and almost any tool in this category are great for a few people, and horrible the more you become.

Dropbox for +100 people isn't fundamentally better than a windows network drive.

A lot of those issues are actually handled by our product:

https://insideropinion.com/

Which indeed, does use topic modeling :)

Yeah, Slack is a chat tool no work organization tool. We use Basecamp for that. Organizing work in an ordered and calm way.

I feel like a lot of the authors recommendations for Slack are talking points in Office 365 keynotes.

The thing I like most about Slack is the ability to add custom emoji – absolutely ingenious. We got a couple of different checkbox ones for status updates, with a question mark inside the box, an exclamation mark, crossed out, checked, which makes it really easy to convey different nuances of "done" and "not done". We have lots of product icons (AWS, GCP etc.) as emoji, so alerts can have the appropriate service logo in them, so people can scan and filter them much easier, plus consistent emoji for different urgency levels. And a lot of fun ones, too, which add some lightheartedness to overly serious conversations. There's a kind of meta-level communication and inside jokes going on in our team via Emoji reactions that otherwise would only work in person. Might not be for everyone, but it works tremendously well for us. Custom emoji give us a really neat way of making such a clean virtual space ours, of making it a bit more human; something that feels sorely needed in a world of cleanly UXed platforms and tools and walled gardens where people get data-driven into funnels.

Human-ness is also something that I feel state-of-the-art AI routinely takes away from things and spaces where it is a dominant feature. What AI we have is quite useful to blur the background of a video call; after all, that requires an understanding of what a human form looks like on video and not much more. After herculean efforts and gazillions of metric tons of training data, that works quite well nowadays. But getting human communication to a degree that would allow for any meaningful assistance in a chat app? Color me very sceptic. If my social ads are any measure for this, there's still a long way to go, and that should be orders of magnitude easier – I interact with a lot of content that isn't that hard to categorize, just show me photography equipment ads already! You don't even necessarily need to deep-mine my natural language content for that. And yet I get shown all those smart toilet ads, no idea why. That's not just unhelpful – it ensures the whole social media experience won't ever feel human and welcoming in the way random threads on low-tech PHP forums did back in the day, or the way a Slack convo can feel nowadays. Not while that stupid robot ant-brain is constantly breathing down my neck and interjecting smart toilet content; it's weird and uncanny, please stop.

I really don't understand why AI would be required to make people convert to premium in an app like Slack, or stay beyond the lockdown. Whatever happened to just making a great tool that's helpful and useful and a joy to use? If anything, it might be the opposite: Annoying, unhelpful, obnoxious AI assistance would drive people off quickly – think Clippy, but in a product that isn't remotely as entrenched and without alternative as MS Office back in the day. The way modern software tends to replace user choice with data-driven fits-for-most defaults, there might not even be a proper "off" setting for such a thing. Please don't, thank you very much.

>Slack actively hinders all non-communication forms of work by distracting the shit out of people and frying their brains worse than Broxholm.

>(Broxholm’s brain fryer was in fact beneficial, if parasitic. Slack is probably neither).

TIL the story and moral philosophy behind Broxholm’s brain fryer (which helps me understand Slack better).

https://en.wikipedia.org/wiki/My_Teacher_Is_an_Alien#My_Teac...

Here's a great review that questions its dubious assumptions about human nature and plot inconsistencies (click the "more" button at the end of "User reviews: LibraryThing member StormRaven"):

https://www.librarycat.org/lib/cctesttc1/item/139177749#

>[...] Eventually, the alien is revealed as is the alien's plot concerning Duncan and the machine used to make him smarter. This is more or less merely a vehicle for Coville to work into the book his argument that humanity is fundamentally inhumane. Duncan's previous behavior, bullying and crude, is contrasted with his nicer, more thoughtful behavior after he has been made smarter. Duncan is also alerted to the fact that the Interplanetary Council (an organization all the alien races of the galaxy belong to) is concerned by the violence and nastiness of humans and is considering what steps to take to neutralize the threat humans pose.

>Coville's thesis may be true, but I have some serious problems with some of the elements of the book. The most glaring is the idea that when Duncan becomes smarter, he also becomes nicer and more humane. One only has to think back on human history to realize that being more intelligent does not seem to correlate in any significant way with being nice. I also think that the way the alien treats Duncan - performing experiments on him without his knowledge, kidnapping and then imprisoning him to use his brain as a communications device - seems to pretty much destroy any claim the Interplanetary Council may have to the moral high ground. Coville's theme, that humans are bad and the aliens are more moral and kind, seems to depend on the idea that whatever bad things the aliens do is justified by circumstance (this is not the first time in the series that an ostensibly non-evil alien has kidnapped and imprisoned an innocent human to further their goals). This sort of moral inconsistency simply saps away some of the message that the books are trying to convey.

>In the end, some dubious assumptions about human nature and some plot inconsistencies regarding the moral nature of the aliens mar an otherwise fun little book about kids dealing with alien teachers. While My Teacher Fried My Brains has flaws that undermine the message of the story, it remains at the very least a decent book for younger readers.

Slack needs to buy hopin.to or a competitor. now.