Introduction

My name is Stefano and I am head of partnership community at MindStone.

How many of you have heard about MindStone?

All right. Catherine, yes, there is a reason because we met in London 10 years ago and she's now in Atlanta and she came just for this meetup to see her old friend struggling in this world of AI.

So what is MindStone? Let's have a look.

What is MindStone?

Pretty much, it's pretty much written there. We want to be the only one platform to learn about practical AI.

And what do we mean by practical AI? So we do not want to focus on theory. We do not want to focus on the wider implications of AI, but we really want to help professionals and people in their daily lives make the best possible use of these tools that are available so that they can earn back some time to do other things.

And I'm really happy that we are here in Atlanta today. I get the sense that it's a really creative city, the music, the marketing. And so I'm really curious to see what comes out over time from these meetups, what comes out from the speakers that we're going to hear over the next few months and perhaps years.

The Purpose of MindStone

So practical AI, how do we do this? We do it in three ways.

The first is through these community meetups. And so we do this in 12 cities now and we want to be in 30 by the end of the year.

And so it's just a moment where people come together once a month and they just see how AI can make things better, can allow them to become more productive and get some inspiration.

And we've done this for two years. We started in London with pretty much the same number of people.

And now in London, we have 200 people showing up pretty much every month. Same in New York, same in Toronto. These have been the founding communities. And I hope Atlanta, under Marco's leadership, will be there very soon.

And so the meetups, but then you see there is a videographer here who is recording. And so we take the content and we put it on our platform. So you can access all the recordings from all the videos for free on our platform.

Meetups and Community Engagement

But then we realized that the content, this content is just not enough. It's good to get inspiration, but ultimately people need some core AI skills.

We will all need some core AI skills to be able to use the tools, and that's why we offer training. So we have a range of training courses. We would do B2B courses for more than 100 people, but we also have a community court running every month where you can just register and follow our flagship program.

This is just to give you a sense of where the communities are today. There's going to be an Atlanta dot very soon.

Thanks to Marcus, who has just been fantastic, by the way. He just jumped straight on this with no hesitation, pretty much. And we put this up together in three weeks, four weeks. It was just incredible, like the energy and the commitment that he has shown so far.

Just a bit more about the training program, just for you to get a sense. You know, it's a weird model, I have to admit. It's quite unique because what we found worked is to do some live sessions with world-leading experts on the platform and then combine that with some learning which is tailored to each and every one.

So you basically have the live sessions where you get to see how someone works with these tools in the best possible way. But then there is a customized learning experience that gives you homework and practical advice, which is very relevant to your job.

The Training Program

This is just a list to get a sense of what the company does.

Practical AI in Action

Now, speakers are going to be, we're going to have two speakers today, me and Marcus, and I will focus more on the practical side of things. So I will show you a little demo of how we can use AI tools for writing, for producing content, for writing LinkedIn posts and

and blog posts and what i will try to show you is that you can do quite a bit to personalize the writing style and really make it sound like you would sound to get away from this stereotypical gpt tone so it's pretty much related to this idea that ai can empower us can empower human creativity and it doesn't need to lead to this standardization or across the meaning that many people are afraid of.

And then Marcus will focus more on the conceptual side of things. He will tell us what this means for the future of creativity.

Are there any questions at this point? Any curiosity that you have around MindStone? Or... What they just said.

Did everyone who came out, did you have a chance to register? You either got the Google form or you got the MindStone registration. Just so I know, how many people were able to register inside MindStone?

Oh, sweet. Just after you come see me, I'm going to get you set up on the actual platform. Yeah, so just come see me at the actual platform. Of course. I can show you. It's a very good question.

So you can access it on the phone. It obviously works better on the screen. So you see the events. You see the content. You would have the videos, you would have some articles that we upload. And then you've got the programs, if you are enrolled to the programs. This is more or less how it looks like.

And then if you register, you also get access to this AI coach, which would give you feedback on your prompting skills.

So, is this gonna work? Okay, this is gonna be handy.

How can I become better at prompting? Not really. I'm using an app which is called Super Whisper. basically allows you to dictate and tends to work quite well so in spite of my Italian accent is able to get most of my words and so you see our coaches elaborating a score and generating some feedback and

Improving Prompting Skills

And of course, this was a deliberately bad prompt, because it's clear what I want, but it's really missing specificity and context. And you need to consider that the more context you give to these models, the better. the results that you get out of them is, just for the way in which they are constructed. So the more information you give, the more the model would access the information that is relevant to the question that you're asking. So to make this more relevant, we would need to say, imagine I'm working on this particular problem, then how can I prompt better?

Does this answer the question? Does it give you a good overview?

Let's get started. Okay, he's not super comfortable, but he will do the job.

So, obviously, just for background, I'm heavily jet lagged, so I will be slow. I might be making some mistakes, and this is also a live demo. Like, I'm, you know, things will go wrong, and it's fine.

AI and Content Creation

AI is changing our relationship with technology. We are used to technology being very predictable.

A car would not break down very often. Google would always deliver a search query and I mean these models are probably right now 80% there and might stay there. So it's fine.

We will all do plenty of mistakes. I will probably also do some mistakes today. It's gonna be fine.

All right, so Let's have a look at how we can write a LinkedIn post together. And for the sake of producing some content, I will look at one of our past events.

Let's take an event that we did in Athens just for the sake of having some content. And I take this link. copy and paste it here.

And this is a custom GPT. How many of you have heard of custom GPTs or MyGPTs on ChartGPT? Half of you. And so it's a pre-programmed version of ChartGPT that would execute very well one particular task.

So what I'm going to do now is, I've pre-programmed this so that it sounds like me when I write my LinkedIn posts or a lot of the social media content. So I fed it a lot of material that I had put together in the past.

So, write a short LinkedIn post about this upcoming event. And this is a very bad prompt. But now we will see, if I put it just in the standard version of charge-upt, it's going to give us something fairly generic.

But look at what I was just able to produce. It's something that I could have written.

And if we go Athens, the cradle of reason, prepares to welcome a new rendezvous of minds, there is some French in there. I'm based in Paris. There are some references to philosophy, which is something that I am passionate about. So it is something, it's not the stereotypical text that ChartGPT would write.

Do we agree with that? Now, let's take this prompt and put it just on the general chat interface. Look at how different is it. And this would sound more like something that the stereotypical charge GPT would write.

And let's look at the reason for this. So if we go on the instructions, So these are basically the guidelines that this custom GPT would follow while answering to the questions or the prompts that I write.

It's fairly detailed. It's clearly trying to deviate from the typical LinkedIn jargon. But I also fed it some of the posts that I wrote.

And I could have made this a lot more, I think right now it's striking a good balance, but I could have made this a lot more accurate and precise. For now, it's working well. and I keep it up to date.

I refresh it once a month by fedding it some more content. What this allows me to do is to have a tool that basically gives me content that I could have written and saves me quite a bit of time when I need to post things in social media.

Generating Blog Posts

Now, let's try to bring this a step further. and write a blog post. And for this, I will give it a bit more background.

And these models are generally good at brainstorming, so you could talk to the models to get some ideas of what the storyline could be. In this case, I think I have I want to have some reference, I know I want to have some reference to philosophy at the beginning and say that it's interesting to bring the meetup to Athens, which was the place where philosophy came to age.

And I will close with a reference to the Academy of Plato and making some sort of link with the AI meetups. So this place where you go and learn about something new, you learn how to think more critically and advance your thinking. Let's see if this works.

Imagine you are an experienced blog writer. I want to write a blog post about this event. Start by saying that it's impressive It's important to bring this AI meetups to Athens where philosophy came of age because generative AI will bring to a similar change in the way we access knowledge.

Then summarize the three talks and conclude by saying that we would hold these meetups once a month and it will become a school of thought and practice similar to Plato's Academy. All right, since my accent is improving, I made that many mistakes. I need to add the event here.

need to get rid of this let's see what happens one thing we're missing is the length let's do typical blog post length let's see if this works Well, don't think it's bad. And it's not perfect, but it's giving me something I can work with.

And I think we go back to this idea of these models being I mean, there is this idea that technology should give you the perfect answer. I think the point here is more that in this case, it gives you a base that you can work with.

And the fact that, I mean, it would probably take me 20 minutes to fix this, to make it a bit more similar to my style and my voice. But it would have taken me an hour to write it from scratch. And perhaps there will be things that you want to write from scratch. Because they are particularly important to you and you really want to, the fact that you're taking the time to write them matters to you.

But I mean, for a blog, in this case, I did something similar, not for Athens, but for another event in Madrid, and the fact that I was able to save 40 minutes to come up with something that still made sense to me was very valuable.

Experimenting with Writing Style

Now, we have time, because we only have two speakers, so I thought we would try to do this, pretty much from scratch. We would select a random topic.

We need to select a topic. We need to select some writer or a newspaper that we like. And we would try to come up with a text that's very ambitious or a writing, a different writing style.

So, What do we want to write about? Any event or anything in particular? Yes?

Oh, God. Look, if I had to stage this, I would have loved that question. I'm a massive basketball fan.

You wouldn't say it from my height, but I've played for a bunch of years. I played all my life, pretty much.

Why is LeBron James better than Michael Jordan? Let's try to come up with a list of arguments.

I will show you a cool tool, by the way. Have you heard about napkin? No? It's a fun tool that gives you infographics.

Yeah, look at this. You can select, this is something, this is a previous brainstorm that I did. And so it would analyze your text. And then it comes up. with a range of potential infographics that you can copy-paste.

And this was pretty impressive. This one that I did in Paris, a previous demo, and it was about the size of the plant-based market. And one of the options is this with plants, which is just, I would have, like, there are leaves. I would have not thought about this, but it's just a very good way of summarizing something which is... which is related to plots.

Anyway, so it's not a new napkin. LeBron. So, why is LeBron better than Michael Jordan? He's not, for starters. We can start there.

Okay, why is he, why is he not? Well, he bought his championships. He has fewer championships to start with.

Okay. He's got a better competition. He has fewer championships, sorry.

I'm starting to think I should have taken something safer. I've already... It's like a lobber. It's like a lobber.

Okay, fewer championships, but more seasons. I think Michael Jordan has had more points in playoffs, no? Oh well. Let's get some unbalanced opinion here, all right?

Debating Basketball Greatness

Imagine you are an experienced researcher of basketball statistics and history. Can you give me three reasons why LeBron has been a better player than Michael Jordan and three reasons why Michael Jordan has been a better player than LeBron? By the way, for the sake of brevity here, I'm using a light version of perplexity, so we're not gonna get that much detail, but let's stick to it.

Because running a deeper research would take us seven to eight minutes. How many of you have heard of perplexity? A napkin? Fewer, right?

No, no, we don't want the deep research. No, no, no. We want to wait three minutes. Nah, we don't want to. Yes.

Which one, sorry? Claude. Claude. Yeah.

Claude was recommended by a trust, but I guess, I mean, what was the difference? Because I stopped using when I started using these two, and I don't use them super often. Yeah. So I think perplexity is really good at research. The referencing is really good. The accuracy of the information is really good.

Cloud is generally quite good at writing. And one thing you can do with cloud. Oh, okay. Cloud.

Cloud. One thing cloud is really good at is integrations. You can link cloud to your Gmail.

So, can you find my receipts in the past three weeks, please? So this is something that would be pace takingly difficult and take to each of us who do any business traveling a long time. And so what Claude would do is go through the emails, takes a little bit of time, and now I've just done it in a very rough format. I would go through my emails and find a list of receipts. And so Claude is really strong on this integration side of things.

While this is going, we're gonna take a look afterwards. Let's look at what Perplexity is saying. Okay, unmatched longevity and consistency, superior all around versatility, greater team success across multiple franchises.

Flawless championship performance and big dominance. Unparalleled scoring excellence is for Michael Jordan. Defensive excellence and complete two-way dominance.

What do we think? Hmm. All right, so in a sense, like you know, there are many ways of, I see this text and now it's where the fun begins and we would just, I'm gonna export this as a PDF.

Now the fun begins and I could use the models to brainstorm on what kind of metaphor I could use to bring this article to life. Like here we are having, but let's try to do it the human way. So let's try to think about a story that could work in this case.

So, LeBron is a versatile player. who has had good longevity and has played in different contexts.

Michael Jordan is just this very clutch performer. This is that sort of story that we are trying to tell.

What kind of metaphor could we use? Do we have any ideas? No, okay. I don't either, to be honest, so.

Imagine you are an experienced writer. Here is some research about the differences between Michael Jordan and LeBron James. Can you suggest five metaphors that I could use to describe their differences?

Let's refine the PDF. Hm, hm, hm.

Hallucination in AI Models

Can you talk a little bit about the propensity for these platforms to hallucinate?

Yes. I mean, right now, I'm deliberately asking them to hallucinate. That's the point, in a sense.

But you know, in a situation like this, we were stuck. So we are just getting some ideas. So I'm asking them to hallucinate.

And by the way, the more you ask for impossible tasks, or the more the task that you ask for is not consistent with the context you're giving to the model, the more you are, you increase the propensity for the model to hallucinate. So my best definition would be It's like the AI is making up information.

But there are reasons why the model does that. So these tools have been programmed to answer the question at all costs. So the moment you give

not enough context for the kind of thing that you're asking, so there is this misalignment between how much context you're giving it and what you are asking the model to do, then the model will have a higher tendency for loosening. It's really important, what we saw earlier with the prompting, the specificity, the relevance becomes really, the context becomes really important.

Is this clear?

Customizing AI Behavior

Swiss Army Knife. Yeah, you can, what you can do, again, there is no, are you familiar with this? With this settings on ChartGPT?

like 50 50 i guess so let's look at them so you can customize it and you see that what i do I put an old series of instructions that reduce the probability of hallucinations. They don't eliminate it, but it just means that the model will be less likely to hallucinate.

Yeah. We go back to these systems not being entirely... I mean, they're very human in strange ways. Ultimately, they've been trained on human text.

They associate taking a deep... Huh?

All right, so... Do we like any of this? I like this. What do we think about this? I think it's the best of this. These are the first, okay.

Continue to imagine that you are an experienced writer. Now using the information in the article and only that information, I want you to put together a short text of about 600 words on the differences between LeBron and Michael Jordan. You should start at the beginning by saying that Lebron is more of an orchestrator, whereas Jordan is a singular force in line with metaphor two that you have found. And you should reiterate that in the conclusion to make it circular.

So what's missing from this, just for the sake of time, we have not chosen a style. But we will stick to it. In the long-running debate over basketball greatness, few compilers have been as persistent or polarizing.

Difference is philosophical. LeBron is a symphony. Jordan is a solo.

OK, I think this can be a lot better. Again, the point here is not to have something perfect. The point is to have something you can work with that will save you time. And if we spend a little bit more time thinking about the voice we want the model to have, I think it could become a lot better than it is right now.

Conclusion and Reflections

So the point of this whole demo for me was to show to you that you can actually use these tools for writing in a quite creative way.

And you can really, for brainstorming, for the production of content, and that you can really they can really empower you to save time, but while still carrying your voice and your unique vision on the world.

Of course, in this case, unfortunately, we failed to have a unique vision. We didn't come up with a metaphor, but I think if we had had some time to take a walk, we could have come up with it.

I hope this was useful.

Now Marcus will tell us more about the future.