Welcome, welcome, welcome. Nate, you're in the building, aren't you? I think, I think
you're on mute. We're also, by the way, for the audience, we're recording this on video
too, so there'll be some content that comes out after this. So I can see Nate talking
away. I should be available now. Can you hear me now? Yeah, I can hear you loud and clear,
loud and clear. So look, Rob then intros, I intros myself last week, we just dive straight
into the topic at hand here, which is AI in crypto media. And we're joined by Nate
Whitehill. I'll give a bit of a background around who he is. And then please dive into
that after Nate and kind of give a bit more detail around what I've said. But as a without
further ado, I guess Nate Whitehill is a tech entrepreneur with nearly two decades of experience
developing web apps and then plunged into digital assets and crypto after discovery
of Bitcoin in 2015. This was a pivotal moment for Nate as it pretty much led the trajectory
to where he is now and a large focal point around what we'd be discussing today, which
is the creation of crypto slate in 2017, which is a crypto media publication that now boasts
millions of monthly hits to their site. Crypto slate does more than just cover the news though.
Nate, I won't steal your limelight there at all. So I let you delve into some of the detail
there and the particulars. But before we hand over the mic to you, I would just encourage
everyone even now or after this to check out crypto slate if you haven't before. It's an
incredible site. And I've said this in private when we've had conversations, the UI and UX
is amazing. And like I say, there's more than just news there. But without further
ado, I guess Nate, over to you. Is there anything else you'd like to add to that? And great
Yeah, cheers. Thanks so much for having me. That was a great introduction. Yeah, as you
mentioned, I first got involved in digital assets in 2015. Over the next few years, I
had the idea to create a website that combined data like coin market cap and news like coin
desk. And that was kind of the genesis for crypto slate was to have daily articles, as
well as very contextualized understanding of all the different digital assets for I think
at the time, there was like 900 assets on coin market cap. And we painstakingly went
through every single page on coin market, every single listing at the time to add data
to crypto slate. So that was really kind of how it all came about. Right now we publish
around 10 to 15 articles daily, we do insights as well. So those are more data oriented analysis
about what's happening. It could be on chain, a lot Bitcoin, macro ETF related type of content
as well, as well as we have the the directory of companies, products, people. So we really
like to define crypto slate as a contextualized source for understanding the digital asset
Perfect. Thanks a lot. And I ask every guest when we bring them on a bit more about your
journey. You've mentioned is 2015. You mentioned since Bitcoin, please tell us some of the
trials and tribulations around that and then how crypto slate came to be, I guess, even
from an organizational perspective, is it just yourself? Do you have other partners involved?
And how do you get it up and running?
Yeah, so my business partner of the past few decades, Matthew Blankhart, he's he's someone
that I've known trusted since we were really like 11 years old. We've been friends. But
yeah, he's he's the CTO of crypto slate. So him and I together, we built the entire website
ourselves. We both have technical backgrounds. And we were part of an incubator in 2010. Tech
stars, I'm sure a lot of people are familiar. So we have a lot of tech experience. And by
the time our our other good friend and investor encouraged us to buy Bitcoin in 2015, that was
really kind of like, you know, going down the rabbit hole. And then about a year later, we
discovered a theory and realized, wow, this space has so much more potential for for
builders than just just Bitcoin. So that was kind of where we were at at the time. But yeah,
over the past, gosh, almost seven years now, crypto slate is developed into quite the
Perfect. Did you huddle Bitcoin the whole way? Or were you trying to trade?
Yeah, unfortunately, I tried to trade a bit too much. So I didn't huddle nearly as much. But I
still do. Bitcoin is my biggest position in my portfolio. And that's the asset that I believe
For sure, for sure. I'm becoming more and more of a Bitcoin maxi. But we will naturally
given who you are and what you do touch on crypto and digital assets and that space. But
really, we're looking to frame the conversation, I guess, more so around AI. So how you've
implemented AI into your business, how some of your colleagues and members of staff at crypto
slate are using it and just generally how you see this impact in, I guess, crypto for one, but
also the media space for two, which obviously play a part in both. So tell us about, I guess,
your journey into AI and how you got involved in and started using these tools in crypto
Sure. So it would have been the first week of 2023, where we discovered GPT 3.5, started playing
around with it instantly became convinced that this is a major technological shift that no one
can, you know, no one can deny. It's gonna have massive ramifications and implications for all
businesses globally, including for media. At the time, it wasn't quite in a position where we
could leverage it, but we were using it on a personal, you know, in a personal way and starting
to really just experiment, see, see what the potential was. At the time, back last year, at the
very beginning, AI or GPT 3.5 specifically tended to hallucinate a lot, meaning just make up
random facts. So we knew it wasn't, you know, where, where integrity and editorial clarity are so
important for publishing news articles on a daily basis, we knew that it wasn't something that we
could implement into our daily processes. So it really was probably like three to five months after
that. So probably between like March and May, that we started to actually think about what are the
ways that we can add AI to our workflow added to, to really like enhance and help us editorially,
help us be faster, be more creative. And just to be clear, we don't use AI to necessarily to just
create articles or write articles from scratch, it really is more of like a check and balance for what
we're doing. So we use it when it comes to being more creative, we've built AI tools internally,
that for example, came to the text or the content that an author has put together and suggest
headlines. And that's kind of how we see it. It's really it's great at making suggestions, it's great
at helping people think outside the box, be more creative, things like, you know, suggesting sub
heading or headlines, like I said, and it can be kind of an editorial check and balance as well. So
for example, we have our editorial guidelines, we have all the things that we're checking for when we
are reviewing a piece of content before it's going to go out. And one of our internal tools can
analyze all of all of the content for a given article, and then determine whether it's meeting
the editorial criteria. But I should be extra clear that AI, at least in its current state is not a
replacement for people, we do need people to write the content, we need people to edit the content,
AI just provides an additional check and balance.
No, for sure. And I think, I mean, similar ways to how I'm using things as well, ideation,
efficiency, even when it comes to brain dumps that I, for example, type, tons and tons and tons of
notes up. And sometimes it's not as necessarily organized as I'd like, but it's still all of the
content and research that's come from me. So even just to get things in shape, ready to publish
might be a particular way to use it. I like the fact you mentioned GPT, and I encourage everyone
here listening to use that as one of the base models, as opposed to just diverting to kind of
wrapper tools and other tools which are using these LLMs in the background, just to really
understand the fundamentals around how it works. So that's, I guess, more so natural language to
produce text or other documentation. Do you use it at all when it comes to image creation or video
creation, I guess, for social media or banners on your articles or anything else?
Yeah, image creation, it's absolutely incredible. It can do things that we were only dreaming about
back in the day, and we were using sites like Unsplash in order to get images or, you know,
modifying images using Photoshop. So yeah, image creation is one of our favorite things that it can
do, especially with Dolly 3 and Mid-Journey 6 are absolutely unbelievable image creation tools.
So how have you found things when it comes to, I suppose, consistency around content? I think
that's something that people struggle with. So anyone can go into a Mid-Journey or Stable
Diffusion or Dolly and produce something incredible as a one-off that may take days for
an artist and graphic designer to create, but for a brand, consistency is key and to make sure it
fits in within your brand guidelines. So how have you found that as a challenge and how have you
overcome elements like that? Yeah, so we definitely have our visual guidelines for
how we like images to appear. One kind of default thing that Dolly 3 does too often is it makes
images too visually complex. And I feel like when images are too visually complex, it takes people
a little bit too long to understand the visual message behind what a headline is. So there's
an element of, I guess, minimalism that we like to have in the images as well. We like them to be
beautiful, but also not so complex that people are looking at it and kind of confused,
and they may not think that it goes with the headline. So that's where Mid-Journey is pretty
incredible. For example, some of the Bitcoin ETF related announcements that have come out,
being able to have a group of businessmen standing around a big black rock with a big
Bitcoin and blazed on it, right in front of Wall Street, that's a pretty compelling and
visually interesting way to present the image for an article. So we are very much aware of
what prompts are required in order to get what we want out of it that goes in line with kind of
the visual guidelines we have for how images should appear on the site. For sure, for sure.
I'm interested, it's not just about the end product of the article and the copy and the imagery, but
the actual operational side of the business, do you leverage any AI tools or even automation tools
like a Zapier or Make to improve your process efficiency or even quality of output?
Yeah, AI, I feel like it really is important for all aspects of business now. So we do use it to
come up with creative ideas to reach out for potential sales. So for example, we sell advertising
on crypto slate and figuring out the best way to communicate comprehensiveness of our packages and
what the best option could be for an individual. We do use AI for a lot of other just creative
things for operational stuff like figuring out what are the best times for everyone to be online.
AI is really whatever someone wants to make it. So if that means they wanted to do a specific task,
one just needs to be as clearly defined as possible in what the AI agent should be doing,
like what's the persona, what's the best outcome, how it should respond. I think so much goes into
the actual prompt and then the follow up. So that's kind of how we view it as like an operational
tool for our entire business. And interestingly, you mentioned something which is working with
the tool. I've seen a lot online of people just going in, giving very vague prompts and not
necessarily designing where they're going with the journey on what they want from the task. So
for example, I always like to imagine where I am at the end of the task and what I want the final
output to be and then reverse engineer things with a series of prompts to get the output and
fine tuning along the way, as opposed to sometimes what I see being this great mega prompt that could
be hundreds of pages long or thousands of words long. And I don't see that being the optimal way
to use it personally. Do you agree with that? And are there any other tips you would advise on?
Yeah, I've never been a big fan of like the copy paste prompts that people like to use. One thing
I love about GPT is the ability to input custom instructions. So custom instructions for anyone's
aware that allows you to give a lot more context for how you want the AI agent to respond. So it
could be background information about who you are, about what your business does, about what your
goals are. There's all sorts of very unique and helpful ways to input information into your custom
instructions in order to get an even better output. And one thing that's awesome about the GPT for
Teams is that you can basically set up separate AI agents and tag GPT. So it's a lot easier to have
custom instructions for all different types of purposes. But it's been amazing to just watch how
much chat GPT has improved over the past year. It's actually, I would say if you fast forwarded
me to now from a year ago, I would think that this is like several years of progress. And this
has all happened in the past year. Oh, it's insane. And I think we've been both used in AI tools for
around the same period of time. And what I find fascinating is just how quickly everything's
evolving. Like you said, we even mentioned the tagging function. I mean, in GPT, that's only
a recent function where you can tag the different GPTs that you'd like to use in one chat, as opposed
to pre-setting them before you go into the chat. And they are listening to feedback. They're
responding. Competition has obviously been super hot with the likes of Microsoft, Google,
Amazon, and a lot of the big tech players you see getting involved, right? So that's super
interesting. And agree, play around with the tool, check out the power of the different features,
building your own DNA effectively, like you've said, around knowledge bases of documents or
PDFs that you might have that you can embed into it, or playing around with custom instructions
and custom GPTs or the equivalent for other models as well, for sure. So that's interesting.
And that's super useful. And I think the one thing I want to double down on there is
the copy and paste prompt. I think they go somewhere, but don't rely on them. Understand
the fundamentals of actually how the tool works, how to use the tool and how to tailor it for
your own good. So thanks for that, Nate. I think that's interesting. And I'm curious,
you've touched on so many areas you would use the tool. Is there any area where you think,
okay, well, we're not mature enough yet, either as a business or education, or just generally
with the tech where we just can't use that for a particular use case? I think to create an article
from scratch, only using AI is not, we're not quite there. And I think that also creates some
ethical implications as well. If a writer is, is basically writing and only using GPT. The other
thing that I think it's important to note is that there are so many AI tools, like you mentioned.
So we always like to run the same prompts across a few different platforms. So Cloud, for example,
at least as of recently has kind of had better results for us in some ways than GPT has. And I
think a lot of people would probably agree that over the past few months, the quality of output
from GPT has gone down quite a bit. And I know OpenAI said they fixed that pretty recently.
But it hasn't been consistent in that the prompts that anyone is using, and then the results that
are coming back are of the quality that one would expect. So that's why it's so important to use
multiple different AI tools, not yet too attached to this one. Yeah. And like you said, test the
same prompt and see what the outputs are. Test different situations. Some, some tools I find
are a lot more creative and some are a hell of a lot more concise. And for example, I mean,
on the flip side from what you said around GPT, I found when Bard first came out, it felt as though
it was a rushed piece. Google, seeing that people were coming to eat their lunch and they thought
we need to get this out. And I just don't think it delivered how I would have hoped and expected
from such a big player. But over time, it's definitely been improving. And of course, they've
introduced other tools like Gemini, and they've got a three tiered approach to Gemini. And you're
even seeing the likes of Samsung implementing these types of tools into their devices. So
it's not just software, but it's also hardware and being able to use AI on a local device.
I don't want to steer too much of an attention there. I just think that kind of context is
important for the audience for sure. So I want to, I guess, move more towards
crypto slate and your coverage. When we, when we spoke around this episode, you mentioned
you're moving more into AI content, or at least having some coverage. I'm interested to know
why and what that looks like. Yeah, I think there's a natural similarity between the type
of user that would be interested in crypto, obviously someone that's very open to embracing
early tech. And the same with AI. So I think it's kind of like a natural persona of someone that
is both excited about crypto and excited about AI. But I don't necessarily think it applies the
other way. If someone's excited about AI, they may or may not be excited about crypto.
So we try to like understand that differentiation. But as far as the type of content, we actually
have begun to cover AI news, even if it's not directly related to crypto, we covered it on
crypto slate. And we try to really like monitor, okay, we're, you know, we're covering AI now. And
is it getting the same results? Or is it getting the results that we had hoped for, meaning views
and engagement? And, and overall, I would say there have been a lot of AI stories we've covered
that haven't had a direct crypto angle to them, but still do really well in traffic. And as we
know, there's been so much drama in the AI space as well with, you know, the, you know, the temporary
ouster of Sam Altman, open AI and all the other crazy events that have happened. So it's fun for
us to cover. And as long as we see it getting good engagement, good shares, good, good time on site,
then we're completely open to it. And I'll be sure it's interesting, you say that I've been in
the web three digital assets space for a number of years as well. And I thought, okay, let's delve
into AI, see what happens. I'm sure it won't be as exciting as CZ had left by an upset the time. And
I thought, okay, well, AI might be a bit more mild. And then like you touched on at the whole Sam
Altman departure and reinstatement within the matter of days came and kind of rumbled the
industry. It's interesting, there have been other quite cool stories, I guess, from a reader
perspective. So even for example, how Google advertised Gemini, they made it seem as though
it's this perfect tool that can respond to audio and voice prompts immediately within a click of a
finger. And it actually turned out it was just really glossy editing and marketing. So the
industry didn't necessarily respond that well to it. So for sure, there's definitely some
interesting stories and soundbites that have come from the space already. Are there any particular
favorites of yours or even the audience so far that you found? Yeah, I mean, I think it as you
mentioned, it has been interesting how Google has embraced AI, I feel like they're much more on the
defensive about how they're developing their tools, because they recognize that over time, AI will
kill search results in the way that we're using them. So AI, you know, it's, it's basically become
what went from, you know, ask Google or Google it to ask chat GPT now. Yeah, I hear that a lot
more than people saying, you know, Oh, Google it now it's just ask chat GPT. So yeah, I think that's
been one of the interesting things is like how Google and I actually, of all the AI tools, I've
been using Google's AI tools, the least, it's really been chat GPT. Yeah, yeah, it's been chat
GPT and claw that have been I feel like the highest quality output.
Yeah, GPT and clawed. And I mean, from an image creation point of view, stable diffusion for me
and mid journey, I think takes the crown for now, at least. What I like about open AI is the
multimodality of it and how much you can do within a single chat, you go from data analysis to some
kind of text prompt to producing imagery to pretty much the whole host of different things you can
all do within one space. And I think a year ago, it was a lot more difficult to navigate,
you needed multiple tools, you needed to stitch the load of things together to make sense of it,
and it's jumping between kind of one tool or system to another. And I guess with time and
2020 was probably the year for it with multimodality and an ability to do numerous things through
one tool. So it's interesting you say that I don't think we're alone in our views of Google.
One big player, I guess we haven't touched on is Apple, who tend to always be late to the party,
but do things well. Do you have a view on where you think they might go with their technology?
Yeah, there's lots of rumors that iOS 18, which should be announced later this year will be the
most fundamental change ever. People are saying that it's because of their integration of GPT
specifically into Siri. So that could change the way that people are doing complex tasks
on their phone, all facilitated by AI and by GPT, it sounds like. So I think Apple at some
point will come out with something. As we know with Apple, similar to how the Vision Pro has
come out, Apple doesn't like using the word virtual reality or mixed reality. They use spatial
computing. So Apple's going to do their own twist and their own take on what they feel like the
language that how AI should be described. So I think it'll be really interesting to see what
ways in which the new version of iOS does change people's lives or change people's workflows with
their tools. For sure, for sure. I agree. And I mean, they always deliver and hopefully won't
drop the ball. Google did when they first entered the market. So I'm looking forward to see what
they'll come out with for sure. You mentioned something around chat GPT instead of Google it.
This ties into another question I wanted to ask around, I guess, the way open AI is using data
and information. And one of I guess the landmark lawsuits in this industry right now is New York
Times versus open AI. And to give a bit of context here, New York Times has basically said, well,
you're using our content, our data, our information. And a short, sharp response from
open AI is, of course we are. I assume there's going to be some kind of public good or utility
arguments for this. But it's definitely one to watch if you are looking at AI tools or how you
can implement it. How does that as a media business, do you think impact you from either
a copyright perspective or even a case of losing potentially hits to your site? If people are doing
something through GPT, do they really need to go to the New York Times or crypto slate?
Yeah, this is definitely a tricky one. I think right now, what's best is if like,
so perplexity, for example, perplexity does do a great job of citing the sources that they're
using when they give you an answer back. So I think that's the most important thing that
should be done now, meaning full transparency into what sources GPT or any other agent is using in
order to generate the responses. So that's rule number one. I do think it is also important for
there to be some consent for what data AI is being trained on. I don't think it should be
the case that AI can train on all data without asking permissions. I think there should be some
user that's giving their okay or not. And I guess part of the question there, which I'd like to
delve into is the revenue piece. So you've mentioned around citation and
authorization to use content. But how do you see it impacting the media world from a point of view
where people won't now navigate my site? Are ads effective, if that's the case? Are people
dwelling on my site? Does that impact SEO? A lot of, I guess, considerations there.
Yeah, I do think it is ironic that given how much GPT and AI has improved over the past years
and the media landscape right now, it looks like Sports Illustrated is defunct
and Los Angeles Times is doing a bunch of layoffs. Media is not going to have an easy
rest of this decade. Media will definitely have to learn, to adapt, to basically figure out ways
to provide more value. And one of those ways that we try to provide more value right now is just
through the original analysis of data that is not available elsewhere. So for example,
when all the Bitcoin ETF amendments were coming through over the past few months,
our senior editor, Liam Wright, he did an amazing job of feeding huge datasets and huge documents
into GPT in order to find what are the interesting nuggets that then we can use as the leads for
different stories and different analysis. And that worked out so well that we had Bloomberg
senior ETF analyst Eric Balconis giving us shout outs and giving us props for the type
of original coverage that we're doing. So I think given like we talked about earlier,
there's a huge creative element, but there's also a huge originality that can be derived
from using the tools effectively that will offset how GPT could really be destroying legacy media
businesses. But I think for big organizations like Los Angeles Times or Bloomberg or other
organizations, it can be difficult for just so many people are basically just cogs in the system.
So they're not really going above and beyond when it comes to the original analysis that they're
doing. And therefore, their businesses get affected. But for kind of lean organizations
like crypto site, we really have been able to take a very fast moving approach to how we can
use these tools to offset what may be a decrease in page views. But thanks to our originality,
our page views and traffic has been way up over the past few months.
Oh, yeah, I was I was checking out before I even spoke to you. And I've seen it. It's shut up.
I guess the short message there is a duck or die.
Really. And it's interesting, actually, I should have touched on this earlier, because you do have
a very lean operation, which presumably is a flat structure allows you to make these decisions
a lot more quickly. Do you want to speak a bit more around your team and how much you do within
that? I think it's quite incredible. I was shocked when you told me.
Sure. Yeah. So I've been an entrepreneur for 20 years. I've always understood how important it
is to be scrappy as an entrepreneur to not not get too ahead of yourself with projected growth or
how big an organization can become. A lot of people look at grilling a business is just
adding on more people. And that's that's not the right way to do it. The right way to do it is how
can we do the most with with the least meaning the least number of people that that way decisions
can be made faster. Everyone can have a much better sense of autonomy, much more power
within the organization to affect change. So yeah, I'm very proud that crypto slate is
probably the leanest organization, leanest media organization in all of crypto, like
websites like Coindesk, The Block, Decrypt. They're all great. They're all fantastic media,
but they have dozens and dozens of people there, you know, 80 to 150 people at some of these
organizations. And that's great, you know, that works for them. But, you know, crypto slate,
we've always tried to operate, as you said, very, very lean. So we're about a dozen people. And of
the people that we that we that work with us, everyone has a great sense of pride and achievement
and autonomy. And it allows us to make decisions and implement new technology or new processes
much, much faster. Not for sure. And I was I was really shocked when you told me how many how many
people there are, especially considering everything you do. I mean, we bumped into each other at
numerous events as well. So even you've got it to such a stage where you're even out there,
I guess, touching grass for one of a better term and actually meeting people and contacts and
creating these partnerships as well. So kudos to you, because, like you said, it doesn't happen
in many industries. Actually, there's a lot of fat that can probably be trimmed. And you guys
are doing an incredible job of, like I said, at the top of the conversation, millions of of hits
to your site and fantastic partnerships with man, a dozen people. It is quite crazy. So kudos to you.
I want to go back into, I guess, the cross section between AI and and crypto.
What are you finding is happening in the industry when it comes to projects using AI or businesses
leveraging the technology and maybe even forward looking how you see that AI and blockchain can
work symbiotically, I guess, to achieve even more than what they do in their own run.
Yeah, I don't have a ton to add to that. We have a list of so one of the things we have on crypto
slate is sector data. We have 50 or 60 different sectors. And an AI crypto is actually our top
viewed sector. I think if you Google AI coins or AI cryptos, you'll see we're in the top two or
three results. This is partly because we developed these pages prior to a coin market cap and coin
gecko. We rank really highly in Google. But as you mentioned, there is a ton of projects trying
to integrate AI, whether it's just really like an attention grab, or they actually have a real use
case. I'm not sure I'm not here to speak on that. But one of the ways I have liked that other
platforms or projects have used AI is Dune Analytics is a good example of being able to
just analyze huge data sets. I think that's one of the things that AI does really fantastically. So
they don't natively have a token. But that's a great example of crypto and AI integrating
symbiotically and providing value for the user. There's lots of other other projects, like fetch
and singularity. But I can't speak to how valuable or how symbiotic their integration of AI and
crypto is. No, I think another interesting point might even be around the copyright issue or what
we can see as a potential ethical issue or societal problem potentially, things like deepfakes.
Maybe you can use blockchain technology to ensure a certain source of where data and information
has come from or to actually aggregate and work with the large data sets like you've touched on,
like you've mentioned. I think other parts which you've spoken about in this conversation is
streamlining operations or actually how these companies work. I guess I'll leave this next
part up to you where I lead it. But as a developer, do you see potentially coding or development of
smart contracts or anything of that nature being impacted by AI technology? And how do you see that
working? Yeah, well, I mean, as we know, there has been a massive implication for anyone who
creates software and programs using tools like Copilot. So I think there's a lot of people that
think that the quality of code that has been produced since GPT and since Copilot has been
a lower quality of code, I think it still requires a lot of human editorial ability to
analyze the code. So if you're using GPT to code, it's really important that you also know how to
code yourself and be able to spot where GPT is making errors in its code. A lot of times when
we use it for small little coding related tasks, it does require a few extra prompts. In general,
I think AI will take anything at best, like 80 to 90% of the way there. But there always is a
human element that's very much required to make sure that what you end up publishing, or you end
up using as code for the website is actually correct, because we're not quite there. Maybe
once we get AGI in the next few years, AGI will be thinking so much more efficiently than humans,
and it will be able to do all those checks for us. But right now, the quality of the responses
is very much dictated by the prompts, but also the human review of the response.
So for anyone in the audience, AGI is artificial general intelligence, right,
which is a lot more powerful than what we have now. It's basically set to really
supersede what we can do as humans if the technology goes where people are anticipating.
But I agree, and I think it's not just unique to your business or the use cases you've mentioned,
make sure there is a human element. Understand the technology so you can improve your outputs
and efficiency, but don't just be lazy and rely solely on AI to do everything for you because
you will be left behind. And ultimately, like you said, the quality and the output
isn't necessarily as good. So use it to your advantage, but don't be lazy enough to try and
use it exclusively with no other human touch. There are other parts, actually, which are
quite interesting. So we've touched on kind of ethical use or how you leverage it within your
business. Commercially, are there any key considerations as a business owner around
using this tech? So recently, as an example, Mid-Journey changed some of their T's and C's
to basically put the owners on the end user around commercial rights.
Is that something you're tuned in with? And how do you, I guess, navigate that as a business owner
and share that with your team? I guess some element there of what you've said could be,
right? Well, if there's some kind of human touch, is it really as big an issue? And
like you've touched on other copyright considerations as well. So yeah, please be
interesting to hear your thoughts around that. Yeah, right now, I think it is very important
to have like disclaimers when AI is being used. So for example, if we have an image that was
generated by AI, you'll find right underneath that in the caption that we mentioned that
parts of this image may have used AI for the generation. And same with content too. We were
probably the first crypto media site to have like an AI disclaimers page on our website that showed
perfect examples of all the ways we're using AI and making sure that the reader has the full
trust and the transparency about how we're using it or how we're not using it. So yeah, I think
it's a very important editorial thing. And as you know, for for media, you know, we're
we're generating views and repeat engagement based off how much people trust us. So the more
forward we are in what we're doing and what our processes are like and how we're using tools,
the better as far as them having having the trust that we're not trying to mislead them in any way.
No, for sure. For sure. Yeah, ethical use cases and best practice and being transparent,
I think is something that should always be the case anyway, but it's more acute now given the
types of tech and AI tools that are being used for sure. So I want to, I guess, take a couple of
steps back and that will allow me to take a couple of steps forward. We spoke about implementation
and how lean you are as a business. How was this decision made to use AI? Was was there
a series of meetings? Did it organically happen because yourself, Nate, were using these kind of
tools or someone in your team? It'd be good to to hear a bit more around how it came to be where
it is now for crypto slate. Yeah, I think it really started with my co founder, Matthew,
and our senior editor, Liam, we were very early adopters of GPT and trying to understand what
are the business benefits for us to use it. So it was definitely a series of early aha moments
and meetings that we had in order to determine what what's the best use case currently. And it's
really, as I mentioned earlier, it's only grown over the past year. It's it's become more and more
of a capable tool for all sorts of business tasks. But I think it is important if any business is
trying to figure out what their AI strategy is to get get the leaders of the company together,
and have a real, you know, an honest chat about what are the pros and cons for for using these
tools. So that's kind of like, how we looked at it, just to make sure that everyone is fully hands
on. And if anyone has objections that we can, you know, we can give good answers or justify what
those you know, those objections. Yeah, for sure. I there's some concern from people across
industries and roles around, okay, well, AI is here, everyone's saying how amazing it is,
everyone's saying it can be better than a human. This will take my job. What's your thoughts about
that? And I suppose as a business leader, overcoming particular doubts around how AI
might be used to oust employees and lower your cost base, we've seen it with major companies
coming out publicly and saying, Okay, well, we're replacing whole sales teams, because
AI can do the job. What are your thoughts around that? And how you
encouraged staff, as opposed to put them off of of embracing these kind of tools?
Yeah, once we decided as a business that we were going to be using AI tools, we really encouraged
everyone to use it, we actually kind of started to almost make it a requirement that they start
playing around with it. Because to use it to understand it to benefit from it, that's the best
way to offset, you know, AI taking someone's job is for someone to, let's say someone's a writer,
let's say someone's a writer. And you know, they previously produced all their articles just by
writing in the traditional human way, for them to use AI in a way to see how is how can AI outputs
be better or faster, and use that as a learning opportunity. So they, they become stronger writers
because they can, they can have AI as a resource, not a replacement, but but a resource for whatever
their task is. For sure. And you mentioned, you mentioned education or training or sharing
knowledge. Is that something officially you do? Or is it more so people just say, Nate, look at this,
I've just done this with AI tools, or what you have set meetings around this, you have a certain
agenda, which focuses specifically on this type of technology? Yeah, certainly in the beginning,
we did a lot more meetings to really encourage everyone to to use it. A lot of people in the
beginning were very hesitant to use it use AI in order to see like what what its capability or
potential was for for augmenting their workflow. But it became so important to us internally that
we basically just told everyone on the team, like, we need you to use it, we need you to use it to
understand what are the pros and cons for you or for your workflow and obviously, be fully
transparent about how you're using it with us. And if you do that, and we see the quality
improved, then then it's a win win for everyone. People are using AI and the quality of the work
is dropping, that they shouldn't be using AI that they should go back to, to doing what they're
doing and then figuring out how they can use it, you know, in a better way in order to improve the
quality of their output or to speed up their processes. But yeah, it's something that I think
every single business should should be orienting, at least the key players in the beginning to
really understanding what is the possibility for for AI to replace their roles in the next,
let's say five years. And then the more that they do that, they can take like a proactive defensive
approach by by being a very good, I don't like the term prompt engineer, but someone who knows
really how to prompt AI to get the best output. No, I don't. Front engineering, I think that
people confuse it with actually being a dev or an engineer creating AI as well. I'm not a fan of the
term, but it is a bit of a buzzword. I agree. Actually, I've managed teams bigger than crypto
slate. So they definitely needed some fat cut there. And as this might sound harsh, I'll get
to the reason why I said it shortly. But I said, your task is to make AI take your job. I literally
set a task to say make AI take your job. Of course, there was a caveat there, which is based on trust
and an understanding me and an understanding of it. This is just about driving quality and
efficiency as opposed to genuinely making you lose your job. But I agree, I think set that task.
I also think allow people to find their own way and I guess have a bottoms up approach as well as
top down to say, okay, well, I've been playing around with this tool within maybe some framework
or parameters. And here's what I've found, because it's such a novel technology. Still, despite how
long we've been using it, that people may have certain hacks or way of using it that maybe
senior management haven't necessarily identified or even have the time to identify or they don't
work as a content writer. I'm sure the writers within crypto slate are in it in the trenches,
basically doing this day in, day out. Whereas, of course, you've got a lot of other matters to
to pay attention to and address. So leaving it to the folks who you've employed as experts,
I think, to work out how it best works and then fostering an environment for feedback and
continuous improvement and knowledge sharing, I guess, across the business. So no, I agree
with everything you said. I wonder, Nate, if you have frameworks in place to do any of the things
I've touched on, or is it fairly organic in terms of how you share that information?
Yeah, we have created quite a few frameworks and internal tools that we've built out for specific
tasks. Unfortunately, none of that is publicly available. I'd almost consider one of our
competitive advantages for how much we've dived into ways in which AI can improve our processes
and allow us to do things faster. One example would be, anytime we record a podcast, we have
video of that podcast, obviously, but then we take that transcript and we run it into AI in
order to get what are the main talking points? What are some of the more interesting direct
quotes that were shared? And we then create an article based off the podcast recording. Article
be another source of driving awareness, driving traffic for a particular podcast episode. So
anytime it's your own content, meaning the transcription that you recorded,
I think that that's a great example of where one can use AI to summarize and to grab the most
interesting talking points and key quotes. And again, it really always does require the human to
jump in and to make sure everything was accurate. It's not hallucinating. It's not making something
up. So it doesn't replace what someone is doing, but it really can make it a lot more efficient
in order to get the desired output. Not for sure. And we're here now recording,
well, we're live on X spaces, but we're also recorded on video as well. And we'll be doing
the same to use certain tools to cut up this content into short or shorter pieces of content
where we can really delve into a particular topic or a gem that we've uncovered here today,
as well as then converting that into other content. Right. So it doesn't just have to be
video or audio because that's how we've recorded it. But summarizing key points, turning that into
a blog, creating newsletters off the back of some of the content you've created and then analyzing.
I think another step is analyzing. Well, how well did these pieces of content perform and
how can we maybe fine tune or tweak our prompts or frameworks to enhance that going forward? So
no, it's really interesting and agreed. It's what we're playing around with. And
Nate, I might have to pick your brains after this, maybe some of the ways you've done it.
Absolutely. Yeah. One other interesting thing that we'd like to do with it is,
for example, analyzing all of our headlines over a 30 day period and correlating that to
views and then feeding all of that data as a CSV into GPT in order to determine what are the
trends among the headlines and the topics and the views. So really, it's a great tool for just
giving insights about data or information that you already have and then figuring out
what is performing best. And then once you know that, then you can start tailoring your content,
tailoring your headlines more towards what is getting the highest engagement.
And you can even ask, we're on the topic of chat GPT, ask it, how would you improve
this piece of content or compare this piece of content to the frameworks or best practices
we've created and tell me where we've maybe missed something or we should include something
else that we've not got in the article or video or actually the framework itself.
So it's just an all encompassing tool. And I mean, needless to say now, we're massive advocates
and I just encourage everyone to, even if you don't use it to the level, I guess Nate and I
have spoken about now, use it for one task, use it for something super simple. Just use it once a
day and it will improve your career. It will improve your business really and improve how
you operate as a person. And even for me now, it's allowed myself and the team and Brain Blocks
really to do so much more with the few people we have than we would have ever been able to do
outside of using these tools. So yeah, agreed. And I just encourage people to go out there and
play around with the tools and see what happens. So we've touched on the power of AI and we've
somewhat scratched the surface of limitations, but specifically, what are some of the limitations
you found throughout embedding AI into crypto slate or just generally as a tool right now?
Yeah, so there is, as we've seen over the past few months, there is still an element where the model
quality does change. So for example, like we said, GPT, the quality of the responses did go down
significantly over the past few months. They say they've fixed it. I haven't used it enough in the
past few days to give my take on that. But that's something I think everyone should be aware of is
that if you use the same prompt for something, let's say a few months ago, you may not be getting
the same quality result now. And that's another reason why it's so important to use multiple tools
and to compare the results that you're getting across Claude, Bard, OpenAI, etc.
For sure, no, 100%. And as you say, it's ever evolving. Some of the bigger players haven't
even got involved yet. So keep in an open mind to try new tools. Again, it doesn't mean you have
to do it every day. You might have your core one or two tools that you go to and you know work for
you, but play around with something maybe once a week or a couple of times in the week and see
where it takes you. So I'm interested, we were speaking about operational impacts and ethical
considerations. Has this filtered into anything around contracts or legal speak with your partners,
not necessarily just a note or a claim that you use AI on your publication? More so on the back
end. Yeah, I mean, AI is great for analyzing legal contracts. I think for any legal contracts,
someone should actually be reading it word by word to make sure everything is accurate. But
as far as as far as AIs being kind of like a lawyer on your shoulder who can identify things
right off the at the very start about what may be missing or where problems are in a particular
contract, I think AI is a fantastic legal analysis tool. For sure, for sure. And I'm sure I've got
a lot of value from this, and I'm someone that looks at AI tools or uses them every single day.
So I'm sure the audience have as well, and I hope they do. Are there any, I guess,
golden nuggets or gems you can drop around specifically bringing AI into media
or tech ventures, even if that's just a one rule you would say to focus on or consideration you'd
focus on? Yeah, I think it's really important to embrace startups that are using AI. So what one
example would be spinach spinach will record your meetings and then give you an AI summary
of those meetings that can be a lot more effective than trying to take notes some of the time. And
we've really we've been an early spinach user really liked working with their their team in
order to improve their product. So yeah, just really kind of taking like a forward thinking
approach to to what are the startups integrating AI and then using those and then seeing if it
benefits you as well. But as far as golden nuggets, you know, I think the best thing anyone can do is
just to just to get their hands dirty to use the different tools and to to just, you know,
not give up so easily not be so quick to discount if let's say the result that they get is not not
ideal. I think as humans, we're too quick to judge most of the time. And by by by taking kind
of a slower more open minded mindset about about these tools, it can provide a lot of business
benefit and a lot of operational benefit a lot of personal benefit even for if someone is trying to
figure out like what the ideal meal plan is or how they should best structure time in their day. I
mean, AI can do basically like almost anything we wanted to do the quality maybe hit or miss.
But I think it's just so important to play around with it and to not not be quick to discount it
as this is just a fad. No, and interestingly, you've mentioned it, this has been very much
business or work focus, but use it in your day to day life, even if I know some companies are
putting restrictions on certain tools or a bit nervous about it, fine. See how you can,
like you said, create a meal plan, review what you're doing in terms of actual activity versus
diet and all of the rest if we're using that example, but it can pretty much be for anything.
So yeah, I can not say this a much enough story, but I've doubled down on just using it. Get your
hands dirty for sure. The other part is spinach. I've not heard of spinach, but a lot of people
are using tools like Otter and Fireflies and a few other tools in that record meeting, summarize
notes, pick out key features around what has been said or talking points. And that could even be
turned into a follow up email. That could be turned into the how you improve your sales pitch,
your pipeline. It could be used for so many different things that I'd encourage. Yeah,
even using something as simple as a tool that can record your meetings and then
give you advice and summaries. Off the top of my head now, I forget the name, but there's a tool
that analyzes body language when you're on a sales call. So it was quite interesting to see,
actually, did someone drop off, part the way through the conversation? Okay, well, we can
tweak how we pitch. It's so powerful if you use it in the right way. So definitely recommend
playing around with some of those tools as well. So that's amazing. What's the name of that tool,
by the way? Off the top of my head, this is the one I forgot. So I'm going to have to get back in
the comments and I'll ping you directly on this one, but I'll get back in the comments. As soon
as this is done, I'll find it and drop it in there for sure. So I'll let you know it's knocking
around somewhere on my laptop. I've not used it in a lot of detail yet. I've generally used
fireflies and seen people using Otter AI. So I'll let you know and I'll let everyone listening in
know as well in the comments. Yeah, and for anyone that's interested, spinach is spinach.io.
Perfect. You heard it here, spinach.io. I'll check that out as well. I've not heard of it yet,
but it sounds incredible from what you've said. And I usually ask a favorite tool. It sounds like
you've got so many. So above and beyond spinach, are there any other recommended and the obvious
ones like chatGPC that we've discussed, are there any other recommendations around tools to use and
what they do? Yeah, I've been liking Perplexity lately. Perplexity is great at citing sources.
So if there's something that's happening in real time, Perplexity can do a fantastic job of
analyzing all the news at the moment, and then giving sources or additional contacts and
information for that. So Perplexity is awesome. Gosh, let me think. What else have I really liked
using? Yeah, I'm a big fan of GPT for Teams because if you use the enterprise version,
it's not training on your data. So that's one way to feel like you have some additional
kind of protection from maybe sharing details that you don't want future versions of GPT
to be trained on. I know that there's a lot of talk right now about GPT-5. The models are being
trained right now for that, and maybe we'll see that sometime this year. So yeah, for anyone
that's worried about their data and what they're putting into GPT, definitely sign up for enterprise
so all your data will be protected. And enterprise also allows you to share, like you've mentioned,
different chats and GPTs across your team as well, right?
Yep, yeah, it's far better than just a standalone version.
Perfect, perfect, perfect. So it's been amazing, super insightful, so appreciate your time.
I wonder if anyone from the Moby team just wants to jump on here and field any questions if we
do have them or allow any speakers up on stage. I'm also going to look at my phone here to see if
anyone's raised their hand.
Hey, I was trying to unmute my microphone.
I'm Tom from Blame Blocks. Amazing insights. I was just looking at the tool you just mentioned,
Spanish AO. Really, really good. I do have a question for you. As you said, you have like
a couple of, you have tons of writers on you. What you want is the, like you've been testing
different chat GPT versions of tools and responses with that. Which one would be like the one you
guys use the most that like delivers the best consistent results? Because like for me, for
example, to work with graphics, we have a lot of tools that you can play text to image,
but consistency is kind of hard to get. So on the writing side, how is consistency for you guys?
Yeah, so at the beginning of the year, GPT was producing pretty consistent results. But
towards the later half of 2023, we noticed that the quality went down quite a bit. So we kind of,
we tried using Claude a bit more. And we found that, for example, transcribing,
like a script from a video was Claude was doing a lot better than GPT. So yeah, in general,
I would definitely say like, the more that you can play around with on a regular basis,
the easier it'll be to determine what's giving you the most consistent results.
Nice, thanks. Thanks for letting me reply on that. I guess it's just about, as you said,
Yeah, that's the key takeaway. Thanks, Nate. We did have some hands up, but some hands have
also gone down. So we'll give it a few more seconds here. And then if not, then of course,
we can wrap up. But any lasting thoughts then from yourself, Nate, whilst we just
hear any more questions? Oh, gosh, I'm really excited for this decade in terms of what
capitalism is going to look like with all of the usage of AI. A lot of people talk about UBI being
inevitable. If that's the case, then the value of Bitcoin and crypto is going to skyrocket if
everyone's getting paid by the government for not working or for working minimally. So I think
we're in such an early inning right now for how this technology could fundamentally transform
society and the way that people work, the way that people live their lives. I also think that
once we combine robotics, like the Tesla bot, for example, with AGI, that it's going to completely
change the workforce, it's going to change the world. I don't think most people are ready for
how different society is going to look like 10 years from now. It's going to be like the biggest
change in how we are operating as humans in all of human history. No, no, agreed. And that was a
profound statement to end on. I'm not going to ruin it with anything else. But it's been a pleasure,
Nate. And I encourage everyone listening now to check out CryptoSlate.com, right?
Correct. That's it. So check out CryptoSlate.com. They've got some incredible news and updates that
aren't yet public. Nate told me, keep that out of the spaces. There will be things made public
soon enough. So check out CryptoSlate for your crypto digital assets and blockchain news, of
course, as well as AI. And obviously, check us out on Brain Blocks, where we'll be updating the
community on news, advice and practical uses of AI. Thanks a lot, Nate. It's been a pleasure,
and I'm sure we'll get you back here sometime. Thanks a lot. That would be amazing. Yeah,
thank you so much. It's been a pleasure. Perfect. Thank you. Take care.