1min Snip The speaker talks about his fascination with language construction and how he created his own language and writing system. He keeps a word from that language as a unique identifier. The speaker mentions Tolkien’s deep study and creation of languages like Elvish, highlighting the creativity that comes from developing new forms of expression.
Speaker 1
That. But you can go really deep into it. You can think about how do you build your own grammar? What kinds of what is the space of all possible grammar isn’t just like you can think about grammar is in a structured way in programming languages. You can think about grammars for human languages in a similar kind of way. And so what are possible grammars? How do you invent new kinds of grammars? And so I had this whole phase, right? I made my own language. I made my own writing system and kind of got into it. And the surface is a word that means a thing in that language that I will not reveal on this podcast. But it’s a remnant of that phase and it’s like unique enough and at this point second enough of an identifier for me that I kind of keep around.
Speaker 2
Keeping some mystery. That’s really interesting. I never really got into like the construction of languages myself, but I was interested in that idea for a time because I’ve always been really into Tolkien. And I think one of the things I later discovered about him was that he was also very deep into the study and the creation of languages like Elvish. I know that he fashioned with like an actual full system of grammar. And so it’s kind of fascinating just to see the creativity, the different ideas that can kind of blossom from just I want to develop a new motive of expression.
Speaker 1
Yeah, I think one of my takeawaysExploring Abstractions for Operating on Text: Tools vs Examples Building a vocabulary of text operations requires exploring different abstractions and approaches. One approach is to use tools that can transform text into bullet points or other structures. Another approach is to use examples and copy-paste style operations. By experimenting and testing, the best way to utilize the technology can be discovered.
Speaker 1
Hypothesis and I can try it maybe I give it to some people and then based on that I update my hypothesis and keep going and I guess in that sense it’s a little bit like research but all of this Is framed around quite specific questions and then hypothesis that go for those questions and so like a question that I’ve been thinking a lot about recently is what are the right talking About sort of vocabulary of ways of operating on text what are like the right abstractions for building this vocabulary is it like different kinds of operations or maybe it’s like example Based like maybe you have maybe one way to think about it could be I have this kind of tool that is a tool that I can apply to a piece of text and turn it into bullet points or turn it into some Other structure but a different way to think about it is maybe the abstractions shouldn’t be the tools maybe the abstractions should be like examples and sort of like copy style and Paste style and so maybe I should the operation should be I take some bullet points and then I say copy the structure of the text and now apply it to this pros and it transforms it into a bullet Points and these are two different ways of taking advantage of the same technology maybe but using it in different ways and which one feels better why and one would answer that just by Just building it and then seeing how it works.
Speaker 2
Yeah I mean a lot of that definitely does sound like researchA Pair of Personal Search Engines: Full Text and Semantic I created a pair of personal search engines last year. One is a full text search engine called Monical, which can search for names, locations, and keywords. The other is a semantic search engine called Reveries, which can find related information based on given articles or phrases. These search engines are useful and simple to use. While personal search engines may not work for everyone due to the lack of interesting data, many people find the idea appealing.
Speaker 1
So this these are a pair of projects that I made mid last year I want to say mid 2021 and they’re a pair of personal search engines they search over an index of I think somewhere around like 25,000 or 30,000 personal data points and so they include things like journals articles and bookmarks online contacts what are the things I’m blanking but there’s a there’s like seven Or eight data sources I guess tweets social media posts things like that and and their repair because monical is a full text search engine I can search for things like people’s names And locations and keywords and reveries a semantic search engine so I can give it like an article or a phrase and find things that are related to it and they’re both useful the semantic Search is not anything fancy it’s just a bag of words on word embeddings but it works well enough in my use cases that it’s it’s fast and simple and it was learning about word embeddings At the time and they’re probably and I think a lot of people like the idea of a personal search engine I don’t think it works well for everybody because I thought like humans just don’t Produce that much data unless people are really like like most people just don’t have enough interesting information they produce outside of their work life and usually at work you Have a pretty good search engine over what you produce whereThe Web Browser as a Tool of Thought The web browser is a valuable tool for storing and organizing interesting articles. The future of the web browser lies in its potential as a tool for thought and command line apps for thought. This idea suggests that graphical interfaces may not be as effective for organizing and processing information.
Speaker 2
Write yeah exactly I haven’t used it as much for the social element but just as a place to like keep highlights for like very long articles and things I found interesting I found it immensely Valuable but those tools are definitely really interesting takes on this so kind of driving with this a little bit after building monocle you wrote this article the web browser as a
Speaker 1
Tool of thought and I’d love to hear you maybe re-articulate some of the ideas you had there yeah let me actually this for a while let me let me google my own blog post to remind myself what
Speaker 2
The ideas are um I can I can prompt engineer here a little bit here so extracting a line I have opened you at the time we’re becoming increasingly convinced that the future of the web browser Is the best tool named medium for thought and that at the time with zetal costin and journaling we’re busy making more effective command line apps for thought rather than dreaming up
Speaker 1
Graphical interfaces yeah I think there’s two interesting ideas here at least that I can recall at the moment one is this idea that no single like your your life and your ideasNotation’s Role in Bridging Abstract Ideas and Real World Perception Good notation should bridge the gap between abstract ideas and our senses. It allows us to represent previously unthinkable thoughts. Whether it’s in dance, mathematics, or poetry, notation is meant to help us interact with the real world. For example, math notation hacks our visual system to make complex concepts easier to understand.
Speaker 2
Yeah I think we were kind of getting into this earlier in that there’s like a back and forth between the notations we use those modes of expression and then the things that we can think And you have some really interesting ideas here just in terms of the perhaps maybe not goal but like aspirations for what notation could be in terms of allowing us I think you have a quotation Here that I pulled out that says notation should be judged by its ability to contribute to and represent previously unthinkable unexpressible thoughts. Can you tell me just a little bit more about like your your aspirations for what notation could and should be in that regard.
Speaker 1
Yeah so good notation has a few properties um one is and this is sort of notation generally whether you’re talking with dance or mathematics or poetry or whatever uh one property is that It should so when you really think about what notation is um its job is to kind of bridge between the symbolic abstract world of ideas of whatever domain you’re working in and the like The senses and the sensory systems um that humans have developed over at the hundreds of thousands of years of evolution to like be really good at interacting with real world objects And so as an example like math notation is about sort of hacking our visual system to-Mechanically moving symbols and hacking our physical metaphors Our brains are wired to reason about the physical world, but using abstract notations allows us to offload some of that thinking onto symbols. Generative models could be a way to create new notations. Effective notations allow us to arrive at new conclusions and work with new abstractions easily. Notations are a great way to expand our thinking abilities.
Speaker 1
Is another example where you can sort of mechanically move around the symbols and like mathematicians or cringes that you like treat dy as a number but you can move around the symbols And kind of achieve the outcome that you wanted and I think that that’s the way all of that is like hacking our the physical metaphors that are brains that are learned to reason about the World and using them to reason about abstract objects um the notations that are sort of really good in our society right now are really domain specific and they’ve been invented manually Over hundreds and hundreds and thousands of years um but it seems like maybe uh sort of generative models that we have today are maybe interesting ways of imagining new notations um One way to think you can as as you alluded to uh that you can measure whether notations are effective or not is that they are kind of you’re offloading some amount of thinking with abstractions Onto paper onto symbols um so that the brain does what it’s good at which is reason about things in the way that you can reason about physical objects and then you’re kind of hijacking That ability to reason about abstract things and so one way to measure whether notations are good or not is can you arrive at new conclusions can you um work with new abstractions easily Using notations that you couldn’t before which is where this sort of like think previously and think thoughts uh I think comes from um and uh one thing that’s great about notation which Is aUsing Language Models to Improve Language Itself Using language models to improve language itself by creating a physical representation of ideas stacked like lego bricks on a table, with arguments that fit together and disagreements that don’t.
Speaker 1
Ideally the benefit should go both sides and one way to look at a lot of what i’m thinking about is how do you use language models to improve language itself um as a byproduct of training A language model you learn a kind of continuous mathematical representation of language and maybe that’s not just a byproduct maybe that’s like a uh a way to give uh language more embodiment And so i talked a lot earlier about um the fact that physical objects have physical affordances that let us sort of meet equok what they’re for and what they’re used by um physical objects Of hints that tell us what they’re made of their attributes and then what we can do with them and ideally i think um one great way to work with ideas would be that you can kind of see them as Objects laid out on a table and arguments look one particular way claims look one particular way maybe um one if uh an object an idea is talking about a particular topic that that’s like A particular shade of blue or something and then you can kind of assemble an argument or an essay by like stacking these things on top of each other like lego bricks and then arguments That fit together sort of like have affordances that fit together like lego bricks and then things that don’t tend to like disagree by literally not fitting together and um this is a Kind of now like very concrete notation um or like a physical embodiment of these ideas but you can itThe Concept of Color Pickers and Creative Benefits Imagine a world without color pickers, only palettes of colors. But then someone invents a color picker that you can describe colors to and it gives you a shade based on your description. Eventually, you realize that the color picker has its own sense and experience of the world. It’s like it wants you to use certain colors. But really, all you need is a color picker with a gradient that you can point to. With a quantitative model and a vocabulary for navigating the color space, you can have a more precise and controllable color picker interface.
Speaker 1
Of a sense of like its own intent so uh an example that i use a lot is like imagine that color pickers are never invented and all existed in the world all that exists in the world were like Palettes of colors and one day someone invented a way to choose and discover new colors and the way you use this color picker is that there’s this black box this agent and you talk to it And describe the color that you want and you describe like i want the color that describes the warmth of a sunset and it gives you a shade of orange and you’re like oh this is really exactly The word i was thinking but i guess it’s close enough and eventually what you might discover is that you start thinking oh maybe this color picker has its own sense and its own experience About the world is like and it wants to you it’s wants me to you want to need to use this shade of orange and um and that’s if like given what we have today it’s kind of silly right like maybe There’s benefits to sort of creative benefits to having this conversation interface to color picking but what you really want is just like a color picker with a gradient and you’re Just like point at the color that you want and you can you can build this um more precise more controllable color picker interface because you have a quantitative model of the range of Colors can be represented and you can um develop a kind of vocabulary for how to move around in the space with things like color and hue and saturation and using all of these both the mathematics That let you describe the space and the concepts like color and saturation that you move around in the space in a structured way you can develop a really precise wayThe Depth and Mastery of Musical Expression and Interface Design Watching someone play the violin is like plucking a sound from the heavens. It’s amazing to see their mastery and range of expressivity. Good interfaces should have a learning curve that extends forever, like mastering the violin. It’s not about learning in 5 minutes, but about sacrificing a little for value.
Speaker 2
Way that she’s able to it’s it’s like kind of plucking a sound out of the heavens and like you know producing it with your your violin and there’s something amazing just to like watch somebody Be able to do that but then also taking that a step further and as you said thinking about like what are what are the decisions they’re making why is that the case I think there’s a lot of
Speaker 1
Depth to it yeah I think this like trait or this like sign of mastery or virtuosity where you have this like incredible range of expressivity that you can just keep going into the learning Curve extends forever into infinity I think is it to bring it back to interface and stuff I think is a really great sign of good interfaces there’s a great talk by rich hickey the creator Of closure or he talks about he he compared he talks about this sort of like learning curve fetish that people have in the programming language community and broadly in tech where people Are like this is so easy to learn you can learn it in five minutes and and that’s kind of like selling like a tar here like a tar here is good but like it’s not you can’t like master it and the Way that you can master the violin and if you are building creative tools I think it’s worth thinking about obviously the learning curve is important it shouldn’t be needlessly difficult To learn but there is value in sacrificing a little bit of like that like learning in five minutesExploring the Implications of Fiction in Building the Future Exploring the limits of fiction and the implications of technological advances is crucial. For instance, envisioning a future where systems make all decisions for us has dystopian consequences. It is important to think through the long-term effects of our creations. Thinking about a far-distant future can inspire fascinating thought exercises, like obsessing over language and notations.
Speaker 2
Yeah I think the the work of fiction is just in general a really interesting medium to explore beyond the limits of where we can actually build or we can actually create just the implications What the world could be if we had some of the things we’re trying to build and I feel like it’s also just very important for us to do that to keep producing that kind of fiction because I think That we often don’t think through the full consequences of what we’re creating so for example I think in in the paper in the nozery ken lu who I’m actually speaking to in January has this One story where he explores the implications of you know systems that are recommending things to you and making decisions and then what happens when that’s seeps into every aspect Of your life so you’re literally no longer making any decisions for yourself and I’ve heard people talk about this type of thing in a way where they’re like it’s great you know you don’t Have to think about what it is that you’re going to do tomorrow or something and it makes all these recommendations and that’s just like really good that’s progress and when you when You kind of take that and extend it to its logical conclusion though I think what you get is very very dystopian yeah I think there’s definitely benefits on like extending into extending
Speaker 1
The present I also like to what one of my sort of favorite thought exercises to do is like work backwards from like a very far-distant future and so I like to think obsessively about language And notations one thing that I think about a lot isMotivation and Insecurities in the Tech Community A person reflects on their insecurities and the pressure to achieve in the tech industry. They believe it’s taboo to admit wanting success, but it’s important to acknowledge and embrace ambition. They pursued various goals and achieved success, feeling no shame about it.
Speaker 1
So when I first came into tech a lot of my insecurities were about there were all these people around me that are like no not much older than I am who are doing things like anyone how can use And starting companies and it’s kind of I think it’s it feels to me like it’s taboo to talk about these things like talk about wanting to do these things but it’s true and I think it’s just Kind of lying to ourselves if we if we shy away from this fact that like we motivate these things and lots of people want to get on how to do is or get on product hunt or start a company or publish A paper or like whatever it is that the metric around the people around you’re encouraging and to to say that you can just make yourself not care about them I think is a lie or or you’re some Kind of a saint or something and so I think and I think it’s it does a disservice to people who are new to this community to say oh you should care about those because they it’s hard to not Care about them like that’s just the way that people are like implicitly sort of people measure measure their value and so what this piece talked about was I came in and I had this like Pressure just like everybody else to do these things that a lot of other people are doing like starting companies or whatever and um over time I kind of like did a lot of them I chased the Goals and I chased the the like virtue signals and I like got on hack and use I published side projects they like got popular whatever got followers on Twitter and I’m not really ashamed About that I thinkThe Most Effective Way to Accomplish External Validations and Goals in Tech It’s important to pursue goals and seek external validation, but the key is to focus on building what you find interesting and attracting like-minded people. Networking and gaining followers are helpful, but prioritize your own passions and values.
Speaker 1
They did them and um they are helpful like getting on hack and use is helpful people will find your writing and it’s valuable to have people that are reading your stuff and giving you thoughts And if you have a side project that goes viral that means that more people with interesting ideas are going to find you and you’re going to have interesting conversations and having A network is helpful like all these things are good and useful things and and it’s fine to pursue them and it’s it’s really emotionally difficult to just like decouple yourself from That um but what I realized when I hit a lot of these things that I wanted was that like actually like all the reasons that I could meet these goals all the projects that I did to get on the Front pages of these things the the things that enable that were not the things that I consciously made to get on the front page of product hunter whatever there’s things that I made because That they’re fine and so I think that the like lesson we should be pushing to people who are new in tech is not to say like I don’t care about like how to use or whatever um I think that’s just Kind of pushes people to hate themselves I think the the right thing is to say it’s okay to care about them there are they are helpful um it’s helpful to network and it’s helpful to get Twitter Followers and things like that but that the right way to pursue them is to just build the thing that you want to build and attract the audience of people that think like you and have the Same values and like to work on the same things in tinker and that’s just the most effective way that I found to accomplish all these sort of like external validations and goals is like Work on things that I find interesting because that’s where you’re going to do your your
