Nikita Roy [00:00:02] 
This is Newsroom Robots, the podcast where we explore the intersection of artificial intelligence and the news industry. I'm Nikita Roy, data scientist, media entrepreneur, and one of the many founders currently building their ventures at the Harvard Innovation Labs. On the newsroom robots, I'm excited to bring you insightful conversations with industry experts about how AI is impacting the way we do journalism. It seems like every week there's a new AI tool or update, and figuring out how to actually use it can be tough. That's why I'm excited to announce Newsroom Robots Academy is launching a live, cohort based masterclass called generative AI for media professionals. I'll be co teaching this with Jeremy Kaplan, who's also joining me on today's episode. Jeremy is the director of teaching and learning at the City University of New York's Craig Newmark Graduate School of Journalism and writes the Wonder Tools newsletter. We'll help you gain practical AI skills. Registration is open now and the link is in the show notes in today's episode, Jeremy and I are going to be exploring the latest AI tools and share how we've been using AI in our workflows. Hi Jeremy. Welcome back to news from Robots. It's so great to have you back on the show again.

Jeremy Caplan [00:01:33] 
I'm delighted to be here with you, Nikita.

Nikita Roy [00:01:35] 
Jeremy, it's been a year since you've been on the podcast. Now it feels like forever ago in terms of the world, in the AI world, since everything is, there's been so many new changes and developments that have happened, and I'm excited to have you here back today to walk us through all of those development and exciting AI tools maybe that are changing the way in which we are working, the latest that's happening and helping us with all the way from like video production to editing to newsroom workflows and walk through all of the cutting edge developments that are happening at the moment. And I'm eager to hear about how all of these tools are impacting personal productivity and just how how we can become better workers in this very AI driven world. So I want to jump right into all of that. You write a lot about the tools on the Wonder Tools newsletter, and you've got your finger on the pulse of so many innovative tools. So what do you think are the key AI tools that are at the moment making the biggest impact on newsrooms?

Jeremy Caplan [00:02:37] 
Newsrooms are using AI in a whole bunch of different ways from the very beginning of the process of gathering information and gathering ideas and collecting story ideas to delivering and presenting the stories at the end of the process. In terms of the beginning of the process, there are tools that really help with background research and understanding what's happening, understanding complex issues. They include AI search tools like perplexity, where you can quickly educate yourself on a complicated topic, get up to speed with an AI powered search that actually has citations so it points back to the original sources. You can even focus the search on a particular subset of sources or subset of data. You can frame fairly complex sequence of searches to really understand something. And for journalists dealing with complex issues on deadline, that can be a really helpful starting point. A couple others that I think are helpful in newsrooms these days are tools for research on scientific topics or other complicated research topics. These are tools like consensus and Elicit, which allow you to search through basically millions of professional journals in a whole range of different domains and get a summary of what research says on the topics that you're exploring, along with citations to the original studies. So you can look back at what they said. You can even get a summary of how many people were involved in those studies and how legitimate they are from a scientific perspective. And it really speeds up the process of doing that initial research. There's still a human element when you go back and read the original study or think about it or talk to the scientists more thoroughly, but it really helps advance that initial process for journalists and newsrooms of understanding the issues, reading the original background, getting to the contextual understanding. And then there's new tools like Upward AI, which helps you once you're gathering that research, to organize it and put it into a format that you can then basically query and build into your own material. And this brings us to AI for notes because so many as journalists, whatever part of the newsroom you're working in, you're gathering notes, writing drafts of things, you're putting together outlines, you're thinking through a topic, you're doing interviews and pulling tidbits from research that you're reading. And once we have those notes, in many cases we lose details, we lose ideas in the muddle of a huge volume of material that we accumulate over time. So one of the things I've been really excited about recently is new tools to apply AI to our own notes. So rather than AI that's just trained on giant corpuses of data out there, we can actually have AI that's specifically trained on our own notes and that we provide custom instructions for and we give a specific amount of content to, because then that AI can be really tailored to what we need it to be used for and there are a few new tools that are really exciting in this arena. One is notebook LM, which is a Google tool that anyone can use for free that allows you to upload hundreds of documents. And these can be notes, Google Docs, variety of different kinds of materials that you've gathered, and you create a notebook that you're then applying AI to. And it's helping you by drawing connections between different aspects of your notes and citing the original place in your notes where those ideas appear, that information appears, and it allows you to then construct new notes based on prior insights that you have, or prior research that you've done, or prior thinking or writing that you've done. And that really brings a new power to reporters to make sense of their own notes and materials.

Nikita Roy [00:06:08] 
I like how you put that. Organized it as sort of AI for research and AI for notes, both of which are tasks that you need to do as a journalist getting a quick overview of a particular topic. Perplexity is so great for doing that, instead of chat GPT, which I keep telling folks about, so that you can get those sources and verify and fact check that information, and then you have AI for notes as well, but kind of also just so much of evolution. Been there for a lot of the different tools to help you with, like images and videos. What about the different workflows that exist in the newsrooms where you're taking content and making it into different, different formats?

Jeremy Caplan [00:06:46] 
Yeah, there's a tremendous amount that the AI tools can do in terms of helping us convert from one format to another. Just to wrap up that point about AI for notes, a couple of other tools along those lines that are really useful include Claude's new projects tool, which basically allows you to upload up to 500 pages worth of material, 200k in terms of the kind of AI technical side. And you are able to put in all sorts of different kinds of documents, your own notes, your own drafts, your style, guide, your past examples. And you can then basically provide a custom instruction which allows the AI to then really help you understand what's in your own notes and materials and reframe that, reorganize that, pull out bits you might note, have noticed, and summarize that, create outlines from that, or even create a draft of new materials from that that you can then build on. An additional way to do that is through a custom GPT that you can create with OpenAI's custom GPT. So there are now multiple ways to create custom solutions where we can apply AI to our own materials, which can be really, really powerful and useful. And I think we'll see more and more of those AI tools that we can apply to our own notes. One last example I'll mention is Mem AI, which got a $20 million investment a couple of years back from OpenAI, which was already identifying notes as an area where AI can be quite useful. And they are working on and releasing soon a new version, a really big update to that platform, which will make it much easier for us to apply AI to our own notes in a kind of tool, like an Evernote type tool or a notion type tool. That's great for anyone who's interested in personal knowledge management, so called PkAM in terms of converting material from one format to another, or journalists who are kind of wanting to convert material from different formats. One set of tools that I remain really passionate about are tools that work with our voice, and our voice is like a natural way we express our ideas and think through things out loud and help us avoid kind of starting with the blank page, the cold start problem. And one new tool I'm really excited about in this, in this domain is called talktastic. It's from the same team that created Oasis, which is a mobile app that I really like for doing what I call Bionic transcription. So it takes your audio from your voice, and instead of just recording it, it also transcribes it. And instead of just transcribing it, it transforms it into whatever format you want. You can give it custom prompts, so it can turn it into an outline, or a summary, or a draft of a presentation, or a draft of an email, or a draft of a newsletter post, or a blog post, or a yemenite video script, or whatever else you want to do. And now that kind of tool, which was available for mobile, is now in the form of talktastic, available for your computer, and you can use it in combination with any app you're using. So if you use Microsoft Word, if you use any email tool, if you use any, basically any tool or app on your computer, you can talk to your computer and then basically apply AI to whatever you're talking about and paste that in directly to whatever app you're using. So it's helpful in the workflow that journalists have for just getting information into whatever app or tool they're using. And there are some other great new apps along these lines that do this kind of transcription, like letterly is one I really like, and audio pen. So there are multiple different tools for mobile phones, as well as talktastic and MacWhisper, for example, for desktop. And those are simple ways to transform your voice and voice recordings into audio in terms of converting text into video and audio into video. There's also some really interesting tools in this realm. Hyper Natural is a new one this year that I think is really exciting because it can take virtually any input you have, whether it's a link to an article that you've published, or a blog post or a newsletter post, or it can take an audio file if you recorded a podcast. Or it can take just some text that you have, and it can convert it into a video, a shareable video that you can use for social media, for example. And you can customize the dimensions, whether it's vertical, landscape, square, you can customize the length, you can customize the illustration style. So the AI is essentially generating images and even now video to accompany the meaning of the material you've created and giving you essentially a social video in a couple of minutes. That might have taken hours for an editor to create in the past, and you can customize it. So what's really important, I think, in this realm is that we still have the human in the loop, right? We don't want just the AI to spit out something and kind of take it or leave it situation. Instead, the AI kind of suggests a draft, and you can say, you know what, I don't like that particular image in the second section of this, or I want to replace it with my own image of this, or I want to have something different in this section. And so you still have authority to continue making edits. You can shorten it, lengthen it, change it, change the dimensions, publish it in multiple different formats for your different platforms. So you really have a lot of flexibility. It's fast, it's easy to use, and it's relatively inexpensive. And that from hypernatural is just one example. There's so many. Auggie is another great social video kind of tool. And there are so many of these platforms now that are emerging in the video realm. Specifically, descript has made major advances, for example, in helping you create social video and other social content out of your audio and video material. And VEED IO is making advances. Vimeo has some cool new video AI features. So all of these platforms are increasingly baking in new AI tools. The next generation of Adobe premiere, which many people use, final cut pro, they're both quickly adding AI features. So this is a tremendously exciting period for people who are creating multimedia in newsrooms to be able to much more efficiently, much more quickly generate social video in particular. And we haven't even gotten into the realm of generative video, right? So now we have also runwayml and soon sora from OpenAI. And so now we're going to have a whole new realm where we can generate video beyond what we've so far imagined. But before we get to that, we still have this powerful moment of being able to convert existing materials and ideas into social video, which is really exciting.

Nikita Roy [00:13:01] 
And when you talk about the social videos, part one interesting conversation I actually recently had with the video editor over at the Daily Maverick in a south african newsroom, and they were actually telling me about how they're very into descript. They also use Adobe Premiere Pro, and a lot of video editors that I'm talking to really are liking descript because of that nature of editing a video the way you edit a Google Doc, quite literally, and so they can get sent it over to journalists, and journalists can get just edit the video versus having to go back and forth with them and talk to them as well. But one very interesting point that she actually brought up, I asked her about these tools that you're taking, right? Converting a story into videos, and they felt that they were losing their ip and their brand id basically when those tools were being used. So it was very interesting also to hear and see how do you keep your brand id maybe when you are using these kinds of tools, what differentiates you from the rest of it and not being noise. So I was wondering if you have thoughts on that specifically when you're using all of these tools, what is the trade off mainly that you're looking at?

Jeremy Caplan [00:14:10] 
I think it's a great issue for people to wrestle with and consider. I think one factor is the degree to which you can customize and personalize the content you're creating. So with hyper natural, for example, you can upload material, including your logo, including your styles, to ensure that what it's generating is generated in a way that's distinctive and still reflects your brand. So you're not having just a generic production, it's actually reflecting your own brand. And in the case of other AI tools, we still retain the control of the storytelling and the narrative and the kind of content we're putting in and the way in which we're steering the AI to create something as an output. So we need to, for example, train the AI with very specific, custom, detailed prompts and even documents that are setting context. In the case of Claude projects, we're providing foundational documents to help show what our style is and what our approach is and what our tone is. So that helps ensure that it's still reflecting our style, and then we're going to humanize it further after that. So once we take an output from an AI, we're still further customizing it, in many cases adding context for it or refining it further. So I think that does help us maintain our personality and our brand and our style. However, I do think we're moving into an era when creating good content, or good enough content in many cases is going to be accessible to many more organizations and individuals and creators. And so that means that the power we had in previous periods to basically stand out just because we had good production value or because we had a brand name, I think that will be diminished. And that means that we'll have to fight harder to be distinctive. And part of that is having more distinctive voices and more distinctive storytelling. I think part of it is having distinctive stories to tell or conceptual scoops. And so it's not just about the presentation of it, it's about the kind of idea generation and the coverage that we're doing and spotlighting things that others aren't paying attention to, zigging when others are zagging in terms of the topical focus and the subjects that we're choosing to spotlight, and the way in which we are doing our reporting and information gathering. So I think there are other ways in which we can be distinctive as news organizations in this era. And this newly accessible, powerful kind of technology means that we can't necessarily rely on some of the old ways that we were distinctive, which in some cases had to do with our production value, or just that we were the, among a smaller group of publishers doing this kind of content.

Nikita Roy [00:16:41] 
Yeah, exactly. And I think you summed that up very well and said that really well in terms of how producing good quality content is going to become easier. Just the production value of that is going to be easier to do. You don't need to have a whole studio right now to record a podcast and do videos, because these tools help very well with editing, these AI editing softwares to take out all of the noise and to do that. And that's something that you can just do with descript at the moment, with the click of a button. I don't even need a producer if I were just using Descript's AI tools to do that. That's interesting to see what differentiates us from the rest of the content and noise that's going to be out there, and how do we make ourselves more unique coming into all of this? You spoke previously about getting into all of the AI voice interactions that you were talking about where you can use Talktastic and it just speaks to you and you're able to use that content. Another aspect I wanted to get into was this voice to voice interaction that's also there right now with tools like chat. GPT. I love it a lot. Like I just got access to their advanced voice feature, which is having a lot of tones, and it can be really angry and sad and excited. And when I just got it, I just kept on telling all of the emotions that I know and made it speak in that way. But it's also really quick in the way it's responding. And so that I want to get into more in terms of like, how do we then focus on using AI for our own personal productivity and helping become better workers and better leaders in the way we are doing our work? Because now we have somebody with whom we can brainstorm and just have a conversation with. How do you see AI helping us just doing better work and being better people at work?

Jeremy Caplan [00:18:29] 
I would say there are many different ways that we can apply AI to our workflows. And if we think about all the different kinds of things we do in a typical day, it starts with email. And with email, there's a huge number of ways in which AI can be helpful. With tools like shortwave, which I think is one of the best AI powered email tools, as well as other tools like Superhuman, we can now not only summarize email. So if we have a long thread that would take seven minutes or nine minutes to read through in detail and get a quick summary of whether that's relevant for us now, and if so, what are the key points we need from it. In addition to that, we can get translations of emails. If we're in a thread that has multiple languages involved, we can also get a quick suggested draft. In some cases, superhuman does this. If it's a quick response from a mass email that you've received that you just need to respond to for some reason, but you don't necessarily want to craft something for the next five minutes, you can get a quick draft that's suggested for you. Then you can also use AI searching, which is really powerful with something like shortwave, where you can query your own past emails. For many of us, if we have thousands of emails every week or every month over the course of years, it's just a huge volume of material to dig through. People spend on average 30 minutes to an hour any given day finding, locating and responding to more complicated messages. And so if you can save a good chunk of that time by more quickly getting to a message or a past message or a past version of a message that you've already sent, that you're going to send again. With AI search in our email, that's a tremendous efficiency gain that we have the opportunity for. And the next generation of this kind of service within email will be basically identifying people we haven't written to in a while that we could reach out to, identifying opportunities for us to be more concise in our correspondence with people, identifying things we haven't followed up on, that we've forgotten about. So there's all sorts of ways AI can unlock even further productivity for us with email. But even the existing ones I mentioned, I think are helpful when it comes to calendaring, AI can be really helpful too. So we have lots of meetings every week. We have multiple people requesting appointments or scheduling threads that we're involved in. And increasingly tools will allow us to the AI tools, and some of them are already building this in with so called magic AI, basically where we can have our meetings, in some cases adjusted. If there's multiple people and we have a shared calendar, the meetings can be adjusted if something new comes up, rather than having a whole new email thread, that can be quite helpful. There's a tool called reclaim AI, for example, that will help you organize your own calendar based on your own rules and ideas. So, for example, if you want to make sure you set aside some time for meditation, or set aside some time for deep work every day, it will find that time on your calendar according to what meetings you've already scheduled, and it will schedule out some time for you. And if something else comes up that's a more pressing or urgent or that you have to attend to, it will reschedule your own meditation time or your own deep work time accordingly. And it will help you understand after the week is over how you've allocated your time and where you might want to make adjustments based on your own principles. So it's not telling you how to spend your time. It's not AI dictating anything for you. It's just AI responding to your own preferences and plans and helping you think them through in a more efficient and creative way. So with calendaring and with email, these are tools that are really helpful. With individual workflows, with word documents, word processing, AI can also increasingly be helpful. And there's a variety of different AI tools now for working with words and text and writing that aren't necessarily just about writing for us. That's definitely not an area where I'm encouraging people to just replace themselves with AI to do their writing for them. But it can really help with editing in particular and in thinking through our own thoughts and organizing our own thoughts. So these are tools like Lex, Lex app and butter is another one, a writing tool that's kind of interesting and useful. And there are other tools like Blaze which help with the social writing. So if you're writing, taking a blog post or a newsletter post, or an article you've written, and you want to create social media posts based on that, or you want to create a script for a little video for TikTok or Instagram or YouTube based on that, or a LinkedIn carousel or some other kind of social version of that. Tools like Blaze and Lex will help with that because they will help you think through the words you're choosing, but they will also help you with editing. So you can highlight a passage in Lex, for example, and ask it for some editing suggestions or observations. And Lex does something else which is really helpful for writers, which is helps identify common writing problems, right? The use of cliches, the use of passive verbs, the redundant language we sometimes use. My old teacher used to joke about the department of redundancies department and other common weaknesses in our writing. We can basically ask Lex to identify and comment on and just highlight, and then we can choose ourselves whether we want to make that change or not. So it's still about our writing, and we can even give it a custom prompt so we can ask it to watch out for weak or hedge words, for example, like slightly or possibly, or in some circumstances, other weak t kinds of phrases that we sometimes fall back on. So that can really be helpful in improving our writing or strengthening our communication, not just speeding it up. There's a common myth out there, I think, among some who use AI, that it's just about speeding things up and doing more in less time, right? That AI is all about efficiency and saving time and going faster, right? And there are certainly cases where it saves considerable amounts of time for us in doing menial tasks, like removing filler words in audio, for example, or removing background noise in audio. That can certainly be a menial task time saver. But I don't think it ends there. I think it begins there. I think that's just the beginning. I think it also helps expand the range of possibilities we think about, for example, with the writing that I was just speaking about. It can also help us identify blind spots or unconscious bias. Maybe we're using a certain kind of term or phrase without regard to who that might affect, or how it might be received by certain groups of people. Or maybe we're just not even thinking about a particular impact of something in a piece of our writing. And so the AI can actually help us, from an editorial point of view, consider more possibilities of blind spots we might have, or unconscious bias, or things we might not have left might not have considered, perspectives we might not have considered, or implications we might not have considered, or questions we may not have answered, or areas where we haven't provided documentation or evidence of something we're making a claim without evidence or without backing it up. And it can even help us consider more headlines or more subject lines, or consider more inclusive language in a certain part where we're not sure how to refer to something. It can help us include alt text to make sure we're inclusive in terms of accessibility for the material we're creating. So all of that is baked into these kinds of next gen word processing tools. And that's another area where, for individual journalists or media professionals, those tools can make a difference in our workflow. And increasingly they're baked into a lot of different tools we're using. So craft has a really good AI implementation. Craft do, which is a writing tool I like a lot. Coda and notion have AI built in. Google Docs is building in AI. So all of these tools where you can write, in addition to blaze and lex and butter, those specialized AI writing tools, even the more mainstream ones, have AI built in. And now even canva has canva docs with AI built into it. So AI as part of word processing and writing is really becoming a super common use for AI.

Nikita Roy [00:26:02] 
And what do you see as maybe use cases for voice to voice interaction that's there as well, not just AI embedded within text.

Jeremy Caplan [00:26:13] 
Yeah. So the chat GPT voices that you mentioned, I think is really powerful because we can start to have conversations with AI that are natural and that flow spontaneously. We can kind of improvise a conversation where we're learning about something. We can practice our language skill. For example, we can practice speaking in another language, or practice speaking in preparation for an interview or a presentation we're giving, and ask for feedback on that interview, or ask it the AI to play a role of the interviewer, or of the negotiator on the other side of a negotiation, or of an editor who we're going to have a difficult conversation with a colleague that can play the role of that colleague and help us prepare. And it can really be helpful to do that in a simulation way where you're using your voice, for example, it can help us to feel like we're experiencing that and feel what the nerves might be like in a different way than if we're just typing it out. And that has some potential power. Increasingly, one area of AI that I think is interesting, and I'm eager to see more of, is the AI that's not just about efficiency or even expanding the ideas we consider, but about helping us think through human kinds of issues. So PI and this AI hardware tool friend, and some other AI bots that you'll see on services like Poe.com, they allow you to talk with an AI to consider how you're feeling about something or how you are communicating with someone in a relationship. And I'm not necessarily talking about having an AI boyfriend or girlfriend or that kind of thing. I'm just talking about having AI assist you in thinking through your behavior or your attitudes or your feelings, which are important things. And some people have the luxury of having a therapist or having a close friend they can confide in on everything. But some people don't always have that or don't have it for every circumstance. And so some AI tools, and there are privacy issues here, and we want to be careful, take careful notice of where our data is going and what's private and what's not. So I want to make sure that mention that caveat, but assuming we have an AI tool that's run respectfully and professionally and carefully and privately, some AI tools can actually be helpful in that regard as well. This might be a good place to mention the local AI. In cases where you really are dealing with sensitive topics or something private, it can be really helpful to run an AI that's only running locally and only running on your machine, not uploading your data or your prompts or your responses to any server of any other corporation or anything like that. Increasingly, these tools are really available for free. They're open source. They're available for you to run privately on your own laptop. And to be specific, I mean tools like Jan AI, j a n AI, anything, LLM, misty, msty. These are tools that are free to use. You could download to your own laptop, and I've even tested them offline. So I wanted to make sure like, hey, there's no way this could be uploaded to some, you know, somebody, some company server or whatever, right? And so I'll turn my Wi Fi off, right? They still work because they're a fully local model. It means the whole model, the whole program essentially is running on your machine and not connecting to the Internet. And you can ask queries, you can have a conversation, you can talk about something that you're a little concerned about being private about, and it won't go anywhere. There's documentation to back that up, and you can look deeper into it. But that's an example of a use case for local LM, which for news organizations in the future, I think will be really valuable to have that security of having a fully offline or local system for running a large language model that is fully secure, fully private, and remains only operated by the news organization itself. And with the open source models, the newest open source models, which are really, really powerful, they're almost at the same level as the top models from OpenAI and anthropic and Google in terms of the speed at which they operate and the breadth of the creativity and usefulness of their responses, which means that you can really use them for a breadth of different kinds of use cases.

Nikita Roy [00:30:26] 
Yeah, and those local models that you're talking about, I think are so important and useful actually for journalists because who might think that you're not able to use any of these large language models because of all of the privacy concerns. You can harness that power locally on your computer, completely disconnect from the Internet. But there's also the advantage that a lot of these models that are there on your phone, like chat GPT perplexity. They have that voice function, right, where you can chat with it and have that conversation. And as you were saying, it gets into such a weird ethical black mirror episode, I would say, of when you start thinking about the ways in which people can use the voice context for AI. But I feel like there's also, people are better at communicating instructions and maybe giving prompts as such through voice. And so that's another way in which you're just talking to. I find myself, at least when I'm using the voice function, giving better prompts for chat GPT. And so I think that's that idea of using AI to help us in new ways, that we probably were not interacting with technology in that way. But apart from all of this, there's also, I think, this idea of how we are having AI embedded with our own personal apps to help us more of our personal consumption of AIh. What are the key apps and tools over there that you think are useful for people to have to just enjoy with AI and have those AI experiences?

Jeremy Caplan [00:31:51] 
I often am commuting to work on the New York City subway and listening to a podcast, and I find that I hear something really interesting, surprising on one of the podcasts I listen to and I can't remember it later on, or I'm thinking about something else later and I've forgotten where it was or when it was in the podcast. It's really hard to return to. So I love an app called Snipt S n I p d for allowing me to save a highlight in any podcast by triple tapping on my Airpods or double tapping on whatever your headphone of choice is, and it saves a highlight of that section. You can then share it. You can have it automatically added to your notion notes or wherever you keep notes online, and it also will give you a quick summary of different chapters of the podcasts that you subscribe to, so you can choose which ones you want to listen to. Sometimes it's hard to tell what's going to be in an episode just from its title, so it's really helpful to be able to see, oh yeah, these are the topics, and I can even scan through and get a quick summary of an entire hour long podcast and decide, oh, I only want to listen to the second half or whatever it might be. So Snipt is really, really useful. And the first times I've paid for a podcast app, frankly, to be honest, because the other ones are often free and they're all kind of similar, so that's really useful.

Nikita Roy [00:32:59] 
Also, you can see what other highlights others have put for that particular episode, which I like to go back and see for newsroom robots what people are snipping.

Jeremy Caplan [00:33:08] 
Yeah, that's super useful. It's super useful. And you get an email of summarizing the things you've highlighted for that week, which is also a nice way to kind of review what you've looked at. Another tool that I like along those lines, speaking of getting a summary, is read wise readwisereader, which basically allows me to ingest all kinds of RSS feeds, blogs, newsletters, YouTube channels that I want to follow, as well as any articles I find along the way that I don't have time to read at the moment. So the old Instapaper or pocket read it later services have been replaced for me by read wisereader. And in addition to just being a great inbox for all that material, it also allows you to apply AI to it. So summarize the longer things that you haven't had time to read through yet, and to query those things that you're reading or listening to, or get summaries of them, and that becomes a really efficient way to organize, you know, the material I want to read for the limited reading time I might have for a commute or travel. So I like read wisereader for that. And likewise is another app where it'll recommend books and movies and podcasts and tv shows and whatever else I might want to consume based on my own interests and past reading and my past Goodreads books that I've read. And I can also query it so I can say, for example, I love adventure books, where it's a nonfiction adventure and has X, Y, and Z characteristics. And it will basically recommend books based on the ratings and based on reviews and based on my prior appetites and tastes and preferences. And that's a really useful application of AI to personal reading consumption or movie consumption, which I really like in the managing information realm as well. When it comes to consumption, I find that the new generation of note taking apps, we spoke a little bit about note taking earlier, they're really making it easier to organize your own material, like even movies or books or podcasts or newsletters or whatever it is that you want to keep track of or keep a collection of. These new personal knowledge management apps allow you to put the information in and then query it with AI in new and helpful ways. So I'm thinking of apps particularly like capacities is a really great new notes app for those who have used notion in the past or Evernote in the past. Capacities is kind of the next generation note taking tool, which is really great, and it's particularly great for keeping collections of books and podcasts and movies and whatever else you're interested in collecting and organizing. Tana is another new one along these lines. Any type, they all work really well for this, and they allow you again to use AI and query with AI and summarize with AI and basically add a layer of AI where we want it or not. If you don't want to use the AI components, those are really some of the great new AI tools, and one I use now called lazy I really, really am liking because any note that I have in this lazy tool, it's called lazy because you can activate it anywhere you are on your computer with a little shortcut keyboard stroke. So you can add a note or add a clip from a website or from an email, and it will save that context so you can get back to it really quickly, which I really like because I don't necessarily want to switch contexts and open up a whole new app each time. So lazy is really quick with the keyboard shortcuts. But it also has this feature where any note that I make or anything that I've saved, I can apply the newest OpenAI model to. So I can take a complicated, rambling, messy note and then summarize it and clean it up and create an outline of it or summary of it or email version of it really, really easily. And it also has all the other note taking functions, so it's a really good alternative to just using apple notes or notion or some other note taking tool.

Nikita Roy [00:36:48] 
That's so many interesting tools over there. And I'm going to try out lazy because I would love for some shortcuts instead of having to constantly switch between windows. So that's my next thing to experiment with after this recording.

Jeremy Caplan [00:37:01] 
And it's lazy. So just in case anyone's looking for it, because it might be hard. If you search lazy, you might get a lot of pictures of people lounging on the beach or something. So it's lazy. So. And it's still in beta. So it's still an early tool. I just want to say that, kavya.

Nikita Roy [00:37:14] 
Okay. You know, actually, I'm just thinking about all you came last year and you were speaking about a lot of tools, and I think when you're coming back this year, it's a lot of new tools, and maybe 20% to 30% are some of the same tools that are having advanced features that you're now talking about. But the landscape is changing so much. So how are you thinking about just the investment of, like, you got this tool now. How do you know if this is helpful for you? Should you be using it? How do you experiment with that tool?

Jeremy Caplan [00:37:43] 
It's challenging and it can be overwhelming. Right. For any purpose you have, there's 13 different tools you could try and explore, and some new ones coming out for editing video or for writing every week. So it's definitely overwhelming. For those of us who love to dip in and try new things, it's like being at an infinite buffet and wanting to maintain a diet. It's hard to do for certain. No way around that. I like the Barry Schwartz model of paradox of choice, where you kind of start with an unlimited buffet and decide you're going to limit yourself to a few options that you closely consider. And then you set a set of criteria, a handful of criteria. So I want something that's free, or I want something that's within a $10 a month limit, or I want something that is careful with my privacy and data. You set a few criteria that are important to you. Maybe it's something that's available in a different language, or maybe it's something that is available on different devices, platforms like a phone as well as desktop. So each have kind of different criteria that might matter to us in different contexts. And then once you have those criteria, you apply them to a limited number of choices that you've narrowed it down to. Sometimes you can use external reviewers to help with that process. So someone you trust as a kind of curator of things can kind of help give you a decision set to consider from consideration set. And then you apply your criteria like the ones I mentioned. Then you basically pick stick and dig. So you pick one you're going to try, you stick to it for a while and say, okay, this is the kind of social video tool I'm going to try for this month, and you stick to it for that month. And you say, at the end of the month, I'm going to decide, is this something that's answered these three questions? Has it been materially useful? Have I actually used it? Has it saved me time or added quality to my workflow? And is it something that I think I'll continue to benefit from in the future? And if it answers those three questions in a positive way, then it's worth sticking to. And if not, then you move on and explore something different that addresses that need you have. The third part of this pick, stick and dig is you dig in. So it's important, I think, when you're evaluating tools, to get a little beyond the surface, because it might seem like, oh, it's just this, it's just the same as this other thing, or it's not really that helpful. Some people jump into chat GBT, for example, or Claude or Gemini or copilot, and they put in write a poem in the style of Michael Jackson about, I don't know, the Seine river or something, and they're like, oh, that's not that useful. And then they move on, but that's because they just touch the surface, right? So it's really about picking something, sticking to it, but then digging in beyond the surface and seeing different use cases where it could be useful and experimenting and seeing how other people are using it and finding really helpful ways in which it can move the needle on something important. So I really think it's important to use it for real projects that matter to you so that you really see that not just theoretically, but practically. It helps you save time and do something with higher quality so that you can free up time for your human creativity. That's ultimately the real benefit of all this. It's not just about doing more for the sake of doing more or doing things faster for the sake of writing more emails. It's doing things so that the things that we want to have time for, make time for the people and the activities that we want to make time for, we can. Because otherwise the lives that we live can be overwhelming. Right? We have so many tasks, we have so many meetings, we have a growing number of emails to address and tasks to fulfill and projects to manage that if we don't think carefully about how we're doing things in our workflow, it can overwhelm us and it can eat up all our time.

Nikita Roy [00:41:07] 
Pick, stick and dig. I'm going to be using that framework, Jeremy. I find that that's really an effective way to communicate how you should be playing around with these tools and seeing if it's helping you or not. Well, I feel like it's a whole laundry list of AI tools that we've had, and you've spoken about all of the different use cases that how you've been using these tools. But I want to know, has that been one particular tool, maybe apart from chat, GPT and all of those top tools, that has really made an impact on your own personal life and the way in which you have been interacting with this?

Jeremy Caplan [00:41:40] 
Well, lazy recently, Lazy Toe has been helpful in ensuring that the multiple projects that I'm juggling I still am able to really keep a close tab on in a way that is different from how I've managed projects in the past. So in the way I managed projects in the past, I had to kind of have a separate app and use something like Trello or other project managers where I'd open them up and spend separate, dedicated time on those separate places. And Lazy has really made it easier to just integrate and what I'm working on with my workflow. So I'm in an email, I'm looking at an email, I quickly save it to lazy with a keyboard shortcut and I attach it to a certain project without having to switch contexts, open up a different app, change my focus, create a new note, and do all those things that add friction to the workflow. So I would say that's increased the fluidity of my workflow and helped me avoid context switching as often and allowed me to retain a little bit more of a deep focus so I can really think about something I'm writing and just save a quick note with a keyboard shortcut without having to break focus on the piece of writing or work that I'm doing. So I would say that's been helpful and I want to mention that the team behind Lazy, and a lot of these teams behind the tools that I've mentioned, like the team behind hypernatural, the team behind talktastic and Oasis, the team behind capacities, which is just basically a couple of guys in Austria and one or two other teammates, these are all really micro teams. They're micro entrepreneurs, they're creators and individuals who are creating really new, exciting user interfaces. Maggie Appleton is envisioning new user interfaces, another great creator. These are people creating new interfaces, new ways of using AI, new applications for the large language models that are really useful in many cases, easy to use, in a lot of cases, free or very low cost. And they are to me, at the forefront of this generation and the new generation of AIH tools, which we hear a lot from OpenAI and Google and Microsoft and Anthropic and these big, big companies, perplexity. And they're doing great things and they're doing great work and they're innovating as well. But the micro entrepreneurs creating these little tools, these everyday tools that I've been talking about, I think are really underappreciated in the extent to which they're creating this new generation of AI tools and use cases. And I think we'll be hearing more and more about them in the years ahead as the true AI pioneers who are not necessarily just the ones building these giant models, but the ones refining the everyday little use cases, where the rubber meets the road, where you actually find value and efficiency and creativity and AI. So I just want to say a little bit of appreciation for those folks who are in the trenches doing that kind of work, sometimes outside of the spotlight.

Nikita Roy [00:44:27] 
And I think that also speaks to the capabilities of AI right now, where it is helping. We didn't talk about that, but it's helping a lot in the development side. As a coder, it is the best time ever to pick up programming and even learn anything. You just need to know the basics of how to build a software, and you have AI tools that are going to help build out that vision. That's what we're seeing. And I think that's one part that also gets me excited, because you're seeing that happening in the tech side, that innovation is also happening in the newsroom side, and there's a lot more. It's all about just keeping the user at the center of what you're building and seeing if they're excited by it, what are they liking, what are their pain points, and then building outside of that. And you don't need this giant team like OpenAI and Google and Microsoft, it's the small ones that are also at the top of your list right now. So that's what we're seeing.

Jeremy Caplan [00:45:18] 
Yeah, absolutely. And even people creating visuals of various kinds, like the team at flourish, which is a great Dataviz tool, napkin AI lets you create these amazing visuals. Graphics mind maps.org charts, other kinds of visuals that in the past would have been quite complicated to create. Unless you're a designer or a very good illustrator, you can now create them with some AI prompts. These are really allowing us to take AI in new directions, new creative directions, in addition to the coding that you mentioned, right, where coders are really empowered now in new ways, and people who thought coding was beyond them can now start to dabble.

Nikita Roy [00:45:52] 
Absolutely. Well, Jeremy, this is, as always, whenever I talk to you, it's always a wonder tool session. And I feel like everyone listening to this is going to be going away with a laundry list of AI tools that they need to then pick, stick, and dig to and try out. So it's been so great having you back on the podcast, and I'm excited to be now also teaching another course with you as part of Newsroom Robots Academy and having this AI course where hopefully we'll have people, a lot of people joining us as we walk through and actually show how these tools work, where AI can fit into your work and what we can do.

Jeremy Caplan [00:46:28] 
Absolutely. I love collaborating with you, Nikita, and I love learning from you and working with you, and I'm so excited. This is just such an amazing moment. If we step back from this arena of AI tools, it's just a phenomenally magic period in which to be creating things digitally, working on things digitally, collaborating with people. The tools we have are just really powerful and magical and fun and also challenging and overwhelming at times. But, yeah, I love chatting with you about this and exploring these new tools and this new terrain.

Nikita Roy [00:46:55] 
Thanks for joining me, Jeremy.

Jeremy Caplan [00:46:56] 
Thank you.

Nikita Roy [00:46:59] 
That was Jeremy Kaplan, the author of Wonder Tools and the director of teaching and learning at the City University of New York's Craig Newmark Graduate School of Journalism. Stay updated with the newsroom Robots podcast, and sign up for our newsletter@newsroomrobots.com. this podcast is made possible thanks to the Harvard Innovation Lab Spark grant. I'm Nikita Roy, and this is Newsroom robots.