First create a product that deluges you with a crap ton of useless work spam, make it impossible to search or organize information (was it a 1:1 chat? Group chat? Room? Thread?), add tons of useless distractions like emoji reactions, THEN offer an AI to search through it.
The emoji reactions limit the noise a lot in practice. Imagine the spam of: yes, this, correct, exactly, no, let's do this, ... which now lives folded into a less-than-one-message space right under the original message. I remember the pre-reaction crap and regardless my view on the emoji itself, it fixed so much in chat communication, I love the idea.
This. I didn’t think I’d like it either, but turns out it’s far easier and cleaner to convey simple thoughts with a tiny picture. Not sure how accessibility tools like screen readers handle those, though.
As a sighted person, I still find it really useful that I can hover over any emoji in slack and get its textual representation, e.g. :thumbsup:. I believe screen readers just process those.
The only gripe I have about this is a super implementation detail (perhaps on our instances only): the “thank you” hands emoji is :pray:. There is also a :thank: that is the same emoji, but it’s weirdly lower res.
No mention about whether they use their own in-house LLM or whether they use a 3rd party like OpenAI, does anyone know more about this?
Edit: from the linked Privacy Statement:
> Contracted service providers may also deliver artificial intelligence and generative artificial intelligence capabilities to analyze data, determine trends, make predictions and create AI-generated responses
So it looks like they may use a 3rd parties to generate AI summaries and responses, sounds like a privacy nightmare for companies.
I hate the trend of people's name getting associated with companies they have no connection with. Tesla is bad enough but are you saying the ml department of salesforce is called Einstein? Hilarious.
At the last few companies I worked the Product Owner gets to name the Product. The general path has been "look up Latin, Greek, German, Gaelic, cartoon, fiction and find something that is attractive to the PO".
I don't even know what a Y combinator is but this is Hacker News which is somehow from Yahoo! iirc.
I am pretty sure that if some of those companies using Slack is operating in Germany (or in some places that has GDPR) it might be a compliance issue due to the fact that this data can be transferred/processed over sees.
Slack, and messaging apps in general are the absolute last place I want AI functionality. I don’t want to think I’m chatting with someone when I’m actually with their AI PA.
Making a fake Chatbot that can masquerade as a person is unlikely to be a feature of slack. In the article, it focuses on summarizing, which I personally feel would be immensely valuable. It would be great too if the AI can learn my personal relevance somehow so it can guess which messages I’d care about and which I don’t.
I'd be very careful about trusting an AI to summarize human communication particularly for work stuff. Things get forgotten and incorrect info gets posted (private Google docs link instead of shared link).
People also send screenshots and non text info like Slack calls. Is Slack going to listen in on calls as well?
Just curious, what tools other than AI would accomplish this? Wouldn't pulling information from chats require contextual understanding (NLP)? I guess maybe it could auto-tag messages rather than summarize content?
You see, OP doesn't want AI, but a neural network seq2seq transformer ML model, trained on a vast amount of text data, say, now that would be something!
For example a tool that lets me take a group of messages and turn that into a ticket or a wiki page with a couple of clicks and keep references synchronized back and forth.
It feels very similar to a few years ago, where everybody and their mom thought they needed a blockchain. However, AI seems a lot more promising to me than the crypto industry ever did.
Masterpiece