AI Assistant

WHAT DO YOU RECOMMEND

Maybe an AI assistant that can e.g. summarize large texts like Notion recently implemented it might be a good idea.

HOW COULD IT BE DONE

An AI assistant could simply be implemented by using the functionalities of one of the already existing, very extensive open-source AI frameworks.

REAL WORLD USE CASES

    - summarizing large texts (like books or long abstracts)
    - automatic creation of flash cards for learning

RECOMMENDED ALTERNATIVES

As mentioned, Notion implemented such thing recently. Also, there are independent applications like ChatGPT that can do this.

ADDITIONAL CONTEXT

At the time I wrote this: none

4 Likes

This could be an idea for when the API becomes available. Due to AnyTypes offline first design, I don’t think any AI integration is feasible for the larger public. I’ve moved your feature request to the relevant subcategory.

PS: could you remove the pencil icons? Those are placeholders/indicators of which parts of the template to edit.

3 Likes

Be very careful of this one. In some of the groups I belong to people are leaving Notion for the way they have implemented this.

4 Likes

To be honest we do not see any real value in the current implementation of AI assistants in context of our app. The hype is big but it will take tike to separate the wheat from the chaff. We have couple of ideas on Anytype way of on device AI usage but for now it is to early to speak out them.

14 Likes

hello, are you able to elaborate on this one a bit more please? What are their reasons for leaving? Is it because it’s a separate pricing? Or just privacy concerns (on top of the already existing ones being cloud first)? Or the way it was coded/integrated?

1 Like

I think AI assistants are necessary, if you can combine search and AI assistants in Anytype, just like Chatgpt combined with Alfred, it must be great~

2 Likes

Thoughts on how Tana implemented this?

Hello, i am wondering if you will add ai to anytype some day. I think it will be a good step to make anytype the best.

3 Likes

Lets get a private chat gpt function please or something similar like GPT4ALL

3 Likes

Combine this with self hosting an Anytype node and you’re all set!

3 Likes

I’m somewhat against this, in general…

Too many people have given too little thought to the risks involved in posting PII / sensitive data to a 3rd party endpoint to “automate stuff for you”

The risk of compromise, in many directions is real… and there’s too little awareness of how the sausage is made, in general for this to be a ubiquitously good thing ™

taking the security concerns off the table for a second… One of the OTHER BIGGEST challenges with “ai” as it sits these days, is that it’s a system that generates coherent-sounding gibberish…
IMO, far too many people seem to have the misconception that the API is thinking for you, when, from my understanding, a more apt analogy is more like:

:robot:In 98% of the examples I was shown, these two words fit very well next to each other in contexts similar to the one I’ve been presented with…

So one gets results that feel… coherent… sane… but the devil’s in the details… and it’s real easy for that to go all sorts of sideways…

jus my $.03

8 Likes

Refreshing to see people in this thread bringing up the privacy concerns regarding AI and options to do so in a contained way.

3 Likes

My biggest issue with AI is the amount of time all these apps wasted implementing a near useless (to me) feature.

Every cent of dev budget that went into building notion AI and not making their mobile app liveable contributed to me looking for alternatives (like Anytype).

6 Likes

I found this related video about proper integration: using the notes as part of the database (starts talking about it about circa minute 30).
This is game changing because it’s not just about summarizing or extending an idea, it’s about using the notes as a database to interact with in natural language.

Personally, this if ever implemented would be something I would always want to isolate from my own content or turn off entirely. I think it is an interesting and useful technology when used in appropriate situations but one in most situations for notes I am taking it is something I would want to avoid.

4 Likes

Just want to voice my support to NOT implement AI into Anytype, at the very least not in native form. Besides it being ridiculously overhyped, there are real issues with data privacy and copyright that go against what Anytype stands for. It’s also not as intelligent as people make it out to be so the exaggerated utility would compromise the foundation of the entire data input system. This app interested me because of the, “Enjoy True Privacy” motto. I would move to something else if AI is integrated into Anytype because privacy and no-nonsense reliability are more important to me than a mirage of new tech “convenience”.

4 Likes

Even if any AI capabilities will be implemented in Anytype in near future (which is still controversial in our team) it will be enabled on demand and will not be a distracting flashy thing you can not turn off anyways.

4 Likes

If I could be of any encouragement here, I would be on the side of not seeing “generative” AI appear in Anytype. First of all, it’s so expensive because it requires vast amounts of cloud computing capabilities - it would essentially kill the idea of a fully offline functional app. Now you’re relying on the main internet again just to let AI work for you. It wouldn’t be sustainable for Anytype to adopt and bake AI into the app - it would officially ruin the app. So many platforms have become ruined with this “generative” AI hype to the point where I can’t stand it. It’s just a baggy extra feature that I don’t use. I have a keyboard and I can write - and I want to write.

It is replacing the most important human function - to communicate! To record history, do research, to write a letter to a friend. What in the world are we doing? Letting something that doesn’t have any sort of purpose, creativity, or mind be master of the future of learning and knowledge. Not to mention, generative AI is fundamentally worthless. It literally requires the masterpiece writings of CS Lewis or Hemingway and anyone who publishes, shares, or makes a literary work to spit out an essay or story and pretend to not be plagiarizing. It takes the famous art pieces by famous people such as DaVinci and Picasso to make pictures that don’t even look real. I read an op-ed (I think it was by someone at The Atlantic) a while ago that basically said this generative AI is wasting our time. My biggest problem with it is that it destroys human creativity and individuality. People who use it are blindly letting it infringe on their freedom to speak and express themselves and such people are missing out on being a part of and contributing to society.

Okay, yeah, I’m pretty opinionated on this matter, but I’m not saying all AI is bad. There are two different types of AI in my opinion - there is the type that helps someone do their job. For example, I am using Grammarly to check this right now - that’s okay, it’s making my writing better, but I still have control over what changes it requests me to make. Sometimes, I don’t correct something because it doesn’t make sense, but ultimately I have the final say in what I’m writing looks like, sounds like, and reads out like - it’s still authentic (just like someone who goes to press and an editor looks over their work - they still have freedom in how the book or literary work turns out and it still is authentically their work). Also, extensions that help you write code and correct syntax are okay, it’s still your code!

However, we get to generative AI, and it’s just scary. People are able to generate functional programs (but who knows what code these bots snatched to write it) - people who don’t even know how to program. You can prompt it to write you an essay - and it’s coherent (but personally, still unnatural all the same). In these cases, AI replaces the creative human. It’s upending the job market because companies now think they can hire AI instead of a human, but isn’t this just like the type of forced, unpaid, or underpaid labor that they utilize in communist societies such as China? Don’t you see, it’s the exact same thing that happened in the Industrial Revolution all over again with a brand new hype. Once again, it infringes on the ability of humans to be creative and passionate about what they do.

We were created to be creative, beautiful, unique individuals - each for a unique purpose. But this, once again, robs humans of their purpose in their vocations and totally upends our lives to let ourselves be ruled by AI, not as in some super-intelligent humanoid bots, but more subtly letting AI do what we were made to do. AI should be like all other technology, helping us in our jobs, making it easier, but above all, not doing it for us and not controlling us. You have no control over what AI spits out, but when you write or make something yourself, you are in full control.

We are letting AI control how we live, and it’s not good. We are entering an age where people are blind to history and where we came from and thus don’t understand where we are going. We can’t let AI infiltrate our lives. I guess I’m just saying, that we need to take a step back and not get hyped up about this. We need to be the individuals that we were made to be. We live in a world where people are constantly searching for purpose and meaning, but all searching in the wrong places. AI is not the right place and it never will be.

4 Likes

Please add paragraph breaks :weary:


IMO, Anytype shouldn’t have any AI built into it, even optionally. Not that AI won’t be needed by some people, but integrating AI means taking on a whole basket of security problems. It would be a better architecture if Anytype could simply make its data available via a simple format consumable by AIs. That really isn’t much more than an application-optimized export capability.

And Anytype wouldn’t even have to design that capability itself. AI is undergoing a period of exploration and experimentation. But in time that will shift to interoperability, and the AI community itself will build the solution.

What kind of security problems?