In conversation with Ivan Ilin: iki, AI and the future of learning

iki.ai is an intelligent knowledge interface – an app that helps you or your team organize knowledge with assistance of AI agents. It was founded and bootstrapped by Ivan Ilin and Max Orlov and their small team. Launched last week of May and already recorded amazing performance on Product Hunt. In this AI in Coaching newsletter edition we are happy to chat with Ivan.

Jelena: Hi, Ivan.

Ivan: Hi, Jelena. Pleasure meeting you.

Jelena: How are you doing??

Ivan: Doing great. Thank you very much in the beginning for accepting to be a guest here in our AI in coaching newsletter.

Jelena: And one of the reasons why you're a guest for this edition is basically the recent success of your product that was also featured in the product hunt and so on. So it caught our attention at coaching center previously, but now it's really like something that is a huge success. So it's a great privilege to have you in the newsletter and also hear your perspective.

So maybe in the beginning, just a little bit of background around how you came to an idea, what's the story behind the product, and if you can share something with us.

Ivan: Thank you so much, Jelena. Yeah, I found it was quite exciting.

So the idea was that we saw a potential to use a natural language understanding to organize information and help people to gain knowledge in a more optimized way.

Because a couple of years ago, we just had Google with its blue links. And when you're searching for information, you had to always start from scratch. And some people could possess some professional knowledge and some just don't.

And you always start from scratch and try to navigate yourself in this amount of information out there. And it can be pretty hard. So we wanted to build a system where you can easily search for, you can dump some knowledge, and then you can query this knowledge base.

And then you can subscribe to people and see what they find relevant to different topics. To some extent, our professional network systems like LinkedIn solve it. Because if you subscribe to the proper people, your feed will be comprised of really relevant posts on machine learning or coaching or whatever.

But still, you have to choose these people. And then when you see this feed, you don't always have the time to consume it. Because it can be a lengthy post.

And sometimes you just need to save things somewhere to then later relate on that, reflect on that, and collect your professional knowledge base. So this is more the narrow vision of iki. And the larger vision is that all the information you need is outside in the internet right now.

And you just need some navigator and some algorithm to help you go through that in the optimal way and not waste time on filtering everything through Googling by yourself.

Jelena: Thanks for that intro. Maybe a couple of questions, two questions, actually, that come to my mind. One is, what is your kind of a metaphor for what IKI, your product, is doing?

What kind of a metaphor comes to your mind in terms of what is it in a metaphorical sense? And the other one is more concrete. And that is, what are the kind of substitutes for iKI or even to reverse the thing, what does it substitute?

What kind of typical products IKI can substitute?

Ivan: Sure. So the metaphor would be just a smart library, an LLM-powered digital library for professionals and teams. And before we had these libraries with people where you can come and ask the library worker to help you with choosing some book, particular thing, finding that for you, he can recommend things. And then he can just perform search. And that's what we wanted to solve with the algorithms in iki. And the second part of the question was about the substitutes.Or right now, normally, you just, if you were speaking about this narrow vision with the library, we just dump things somewhere, either to notes or to your, like you just open a lot of tabs in your browser or people save things to Notion. And then, when you find some relevant materials like GitHub repositories or tutorials, videos, you save it somewhere, either in the system where you consume it, Medium, YouTube, and so on. There are bookmarks, but they're not synced. Or you just open them somewhere or copy paster to your notes to open later,? and you cannot extract knowledge from this link. So later on, it just becomes trash because you never review that, you don't have time, life goes on. And what we wanted to solve is extracting knowledge from all these sources and you being able to query this knowledge in natural language.

So you just ask particular questions, for example, how to get to product market fit. And you already dumped a number of materials, filtered that you've seen on LinkedIn or that people shared with you that you find somewhere, but you didn't have time to review them. And they're sitting there in your library and when you ask this particular question, we extract all the information regarding this particular question from all the sources, all the links, we grab full text of any source.

And then we just feed them to our, like first we perform search and then we feed them to a large language model. It's called RAG, aka Retrieval Augmented Generation. And this large language model just produces a very nice answer, pretty concise and relevant to your question.

So that's the news, basically large language models are the new interface to information because they are capable of transforming it based on a query that you fed to them. That's pretty new. And before we didn't have this option, we just had coaches and not only coaches, but in general professionals who were able to answer a question based on their knowledge.

Right now, any large language model, given a set of information fed to it can do that. That's the new technology.

Jelena: Yeah, it's really amazing and so exciting. And maybe just to add a little bit of our perspective from coaching center, when we saw first iki, our first impression was like, wow, what a powerful learning management system. So in my mind, it was like some sort of a substitute for traditional LMSs.

But I know that in previous conversations with you that that is just one kind of an option. But if we take that kind of a direction, like taking iki for learning and as a learning management system, what are some of the things that people can do with it that they couldn't maybe with traditional? In my view, it's like getting so much broader options, having so much more dynamic environment for learning, for adding your notes, adding bookmarks from wherever, having AI in the conversation and some sort of a partner for learning.

But then also curious to hear your perspective. If we take that narrow direction of an LMS, what kind of an LMS would that be?

Ivan: Very good question. So yeah, obvious thing is just this question answering mechanics, but it's on the surface and then you can go deeper and try to work with prompts and basically given what you ask, you can provide tasks to this language model and to the system. And given what you ask, it will produce different results.

So you can always try to build your educational process with the system. So let me think of that. You can, for example, you can ask the system to form some questions for you that you should be able to answer given your previous learnings.

So that's one of the things. And then it can validate whether you answered them in a good way or in a bad way. So it's basically like a tutor.

You can treat it as a tutor. And then, so yeah, that's just a thing that works out of the box. The magic is that it works out of the box.

You don't have to program anything extra for that. You just have to drive the conversation in a different way, not just asking a co-pilot for answers, but just doing these mechanics and it will work. That's brilliant.

I mean, it's not just our success. It's just the brilliancy and the genius of large language models, how they work right now. And then you can integrate it a bit, I think, with the content consumption system.

So it's more of making a tailored system, but you can basically replace, you can make it interactive, I would say. So you can replace a person's proactivity, a teacher's proactivity with the system's proactivity. So it can track you, it can recommend to you what you should read next and what you should actually do next.

It can review your learnings and understand whether you got the topic and then maybe suggest you to have some further or dive deep lesson as one.

Jelena: I know that in our community, we have a lot of L&D professionals who are organizational learning and development people here, especially in the Southeastern Europe region. What's your kind of a vision for iki as a product and maybe also similar product in terms of future of learning, future human combination, and maybe even coaching in a more narrow sense?

Ivan: So basically, what we talked, what we discussed before is about a bit of replacing human in terms of these interactional part. But still people are the ones producing knowledge and all the knowledge that our current system possess is due to human knowledge.

So it's not that inventive still, but it's very good at transformation things, transforming things to given your query or given some task and so on. So people will still be leading all the educational process in terms of navigating what's important and sparking ideas, and then just guiding the whole process, creating knowledge. And then involvement is also important, you cannot trust AI in the area of emotions in general.

You can trust that is producing relevant or correct answers, but you cannot trust it in terms of your are being completely guided by AI right now. It does not motivate you. It does not encourage you but people do.

So still this part will remain and we don't have to fear AI at all, but only interactional and information retrieval and recommender parts will be sold. I mean, they are already sold and the systems will be way and way more integrated with LLMs, I guess. So LMS would go along with LLMs, that's my prognosis.

Jelena: Yeah, that's really interesting. Being a psychologist, I believe in, let's say one school of thought, like socio-cultural theory, which says that tools that people use and create also throughout history from like memory can be also seen as a tool and then writing and I don't know, then computers and even the internet. And now of course, the LLMs and the generative AI.

Anyway, that tools mirror in a way human cognition and that tools shape the way we think and also the way we think about our thinking, which we psychologists call metacognition. So tools have a huge like role in human cognition and metacognition. Having in mind this kind of a product like iki, what do you envision, what kind of an impact it may have on human cognition slash metacognition?

Jelena: What comes to my mind is like a dilemma actually, whether it would make us more like efficient in terms of like having an optimized, as you said, digital library or whatever, or something even more kind of ambitious, having some impact on transforming or augmenting the way we think, becoming more creative, having more like powerful maps of what we do and how we do things. So maybe in terms of this dilemma, transformation versus efficiency, what's your view there? What would we benefit more from these kinds of tools?

Ivan: I hope they won't take away our critical thinking and reasoning because that's what they do very well. I think that the paradigm will not change a lot because still, before you had Google for 20 years and people got used to having all the information at their fingertips, they still are not that proactive in research. Some are, some are not.

So we have this amazing tool, we have all the internet available right from any computer, but we're still too lazy and too swamped with our daily things to really dig into, like it changed the research process a lot, but still the basics are, I would say the basis is the same as it was before Google. Same with the large language model. But they'll definitely solve the interactional part.

And I would say they're a very good interface to information that would transform the whole internet and it is transforming right now. And a lot of companies will be created, a lot of value and so on, but I wouldn't say it would change our minds and thinking paradigm. I would say AI will augment us and it does, but it doesn't transform our habits too much.

One thing I see for sure is that human touch will be even more valuable than it was before. So before we had a lot of content created by people, right now, I guess in the upcoming couple of years, the balance will change and there are already companies trying to replace human writers just by, you give a topic and then you are provided with a nice paper comprised of different, a lot of different articles and so on. So these gen AI content will be a swamping internet and we'll have to fight it.

That's not a good thing. So human touch will be more valuable and hopefully we'll find our way through all this content mess.

Jelena: Yeah. I recently heard one, the thing that I like very much about AI and that it's maybe most impactful role is to make humans more human. So what you just said resonates in a way with that.

Okay, and maybe just regarding some sort of your vision of the future, what are your plans? You didn't brag really a lot, but the product of the day is a little bit of a thing and so on. So can you brag for a while and tell us where you are moving next with the iki?

Ivan: Sure, thank you. I mean, we're humble because we were the top product of the week for the whole week and then the last like nine hours, we dropped a few places lower and that's why I'm not that satisfied with that. But still, yeah, I'm happy with our launch and that our vision resonated with a lot of people.

And the vision is that we can help people organize their knowledge and give them a place, a personal Wikipedia where you can drop your knowledge and be sure that it never goes away and it's always at your fingertips. So that's the current version. And then we are heavily working on agents right now in order to outbound people's research.

To make it more proactive, make the service more proactive. We can grab the topics that you're researching based on what you save and then help you augment this with the web crawled content because it basically does what you do. So if you're reading about how to deploy a large language model on a local computer, there's already a bunch of stuff sitting in the internet about that and you don't have to search it manually.

We can grab that for you and then provide you with nice articles and summaries on how you exactly do that. And then we can practically ask you questions. So either you're doing that on an M2 Chip or you're doing that on other architecture or probably on AWS with Nvidia GPUs, and then we can tailor this information based on what you save.

So it can be a really interactive system and that's where we're heading to. Just a copilot that proactively augments your professional tasks.

Jelena: Wow, that's really exciting and looking forward to see that identic movements in IKI. Okay, thank you so much for sharing your story. Really inspiring and exciting.

And I hope also our readers will find this interesting and inspiring. Thank you even so much.

Ivan: Thank you, Jelena. Thank you for having me.