As generative AI enjoys its time in the spotlight, exciting innovations are being developed for experts and also those with limited technical knowledge, and PartyRock from Amazon Web Services (AWS) is becoming a popular playground for both professional and novice developers.
Mike Miller is the Director of Project Management at AWS, and he joins us today to share about the inspirational AI-powered products and services that are making waves at Amazon, particularly those with generative prompt engineering capabilities. We discuss how Mike and his team choose which products to bring to market, the ins and outs of PartyRock including the challenges of developing it, AWS’s strategy for generative AI, and how the company aims to serve everyone, even those with very little technical knowledge. Mike also explains how customers are using his products and what he’s learned from their behaviors, and we discuss what may lie ahead in the future of generative prompt engineering.
Key Points From This Episode:
Quotes:
“We were working on AI and ML [at Amazon] and discovered that developers learned best when they found relevant, interesting, [and] hands-on projects that they could work on. So, we built DeepLens as a way to provide a fun opportunity to get hands-on with some of these new technologies.” — Mike Miller [0:02:20]
“When we look at AIML and generative AI, these things are transformative technologies that really require almost a new set of intuition for developers who want to build on these things.” — Mike Miller [0:05:19]
“In the long run, innovations are going to come from everywhere; from all walks of life, from all skill levels, [and] from different backgrounds. The more of those people that we can provide the tools and the intuition and the power to create innovations, the better off we all are.” — Mike Miller [0:13:58]
“Given a paintbrush and a blank canvas, most people don’t wind up with The Sistine Chapel. [But] I think it’s important to give people an idea of what is possible.” — Mike Miller [0:25:34]
Links Mentioned in Today’s Episode:
Mike Miller 0:00
So as this thing gained steam internally, there was just this light bulb moment of oh, my gosh, like all of our customers and even customers who don't know anything about building on AWS can use this thing to really gain some intuition and understanding for how generative AI works.
Rob Stevenson 0:18
Welcome to how AI happens. A podcast where experts explain their work at the cutting edge of artificial intelligence. You'll hear from AI researchers, data scientists, and machine learning engineers, as they get technical about the most exciting developments in their field and the challenges they're facing along the way. I'm your host, Rob Stevenson. And we're about to learn how AI happens. All right, welcome back, everyone to your favorite AI podcast. It's me your host, Rob Stevenson, and I have another amazing guest for you this week really excited about this conversation. He is the Director of Product Management at AWS, Mike Miller. Mike, welcome to the show. How are you today?
Mike Miller 1:02
Hey, Rob, thanks for having me. I am doing great.
Rob Stevenson 1:05
I'm really pleased to have you because it feels like you touch a lot of different products in your role. Like product management is typically like a product if you're, if you're speaking to someone from a startup, but Amazon AWS obviously puts out loads, you get to play in a lot of different arenas. We'll get into that. But first, I would just love to know a little bit about you, Mike, would you mind sharing about your background, how you came to be working in the AI space, and particularly in this role at AWS?
Mike Miller 1:28
Yeah, I'd be happy to Rob and I agree, I am one of the luckiest product managers, I think in terms of the breadth of technologies and fun projects that I get to work on. So for so by way of background, I've been at Amazon for about 11 and a half years, the first five years or so I worked on the team that built and launched the Fire TV Line of devices. So had a ton of fun there, you know, getting content into people's homes. But then I needed a little change. And I moved over to the AWS organization where I joined the AI ml team. And at that point, this was 2017. You know, AI ml was really kind of picking up steam, and AWS was just starting to get into the fray. And they were looking for a product that could introduce it to developers in a little bit more fun and hands on way. And so I took a leadership role on the team that built and launched a product called Deep lense, which was an AI enabled webcam. And we use this product to give developers a hands on way to learn about some of these new techniques of AI and ML. And this really was out of a learning that we had at AWS and at Amazon where we were working on AI and ML and kind of discovering that, you know, developers learned best when they found like, relevant, interesting kind of hands on projects that they could work on. And so we built deep lense as a way to provide a fun opportunity to get hands on with some of these new technologies. It was a big hit developers loved it opened up a lot of conversations about AWS and our capabilities. And so Andy Jassy asked us, hey, what's the deep lens of next year. So we kind of put our heads heads together and came up with a bunch of ideas. And we built and launched a product called Deep racer, which is itself a little 118 scale remote control car that we kind of bolted a computer on to. And what you do is you train a machine learning model in the cloud, to drive a car around a track, pretty simplistic simulation. And then you can take this model, you download it into the physical car, and the physical car drives around a track. And so of course, what's the car driving down the track without a little competition, so we built a league on top of it. And so we launched the deep racer League, the deep racer vehicle, as a way again, to get hands on and sort of have fun learning about AI and ML, it's been going strong for the last five years, actually, we just celebrated the fifth anniversary at reinvent this past year. And then we went on, we built a couple other additional products like that one of them was called Deep composure, which was about using music to teach AI a specific model called the transformer model, which if you've learned about generative AI, you know, that forms the basis of a lot of our generative AI, kind of foundation models and large language models today. So we were a little bit ahead of our time with that product. But it was pretty neat and fun and drove EO hands on usage of AI and ML technologies. So we kind of have been doing those things for the last five years or so then this past year when generative AI, you know, became the rage. And we realized there was so much opportunity here, we kind of took the similar approach. And we said, hey, you know, there's a lot to learn here. And a lot of sort of nuance with generative AI that's different. And we think this is something that a lot of people can be introduced to in a fun way. And so we'll get to it. But that's that's sort of the genesis of Party Rock and some of these other things that we've been doing. So I've had the very fortunate opportunity to work on a bunch of different really fun learning programs at AWS.
Rob Stevenson 4:49
It seems like AWS is customer, for lack of a better word is the developer. And so I'm seeing kind of a throughput in these tools. These products that you're describing, but they also vary wildly, right? Like you have everything from like just developer kits all the way to, you know, self driving car competitions. What is the throughput? What is like the common true north? If you're trying to serve this developer? How do you decide what to work on and bring to market?
Mike Miller 5:17
Yeah, that's a great question. I think if you look, yes, at the whole of AWS, we're all about helping bring workloads to the cloud helping, you know, developers and companies be more innovative in no cost conscious way, you know, through our core capabilities of cloud computing, when we look at AI, ml, and generative AI, these things are transformative technologies that really require almost new set of intuition for developers who want to build on these things. So part of the reason for these learning and hands on projects was really to help our builders or software developers gain more intuition about how these technologies worked. I think when we looked at generative AI, and the kind of capabilities that it brings, you kind of see that generative AI allows users of all skill levels, even users who have no idea what software coding is, the ability to get some value out of it. And we noticed that there wasn't really a great way for you know, those people to get hands on and, and really get some intuition about how it works, and how it might be applied, you know, for themselves. And so that's really this sort of common theme underlying all of these is that, you know, hands on learning. And kind of fun, is a great motivator to learning versus, you know, just like opening up a textbook or getting a worksheet or sitting through a course, like if you can have fun and sort of see the results of your learning in a quick way. That's the best motivator for kind of learning new stuff.
Rob Stevenson 6:43
Yeah, and this idea of people being at play of having a ton of fun with a product and what sort of innovation can that breed? Maybe we hamstring it by calling it fun or play. But that's kind of the case, when you speak to anyone about a problem they care about deeply. It doesn't have to be something that's fun, like talk to a cancer researcher. And they it's kind of like the light up when they speak about it, right? There's something that just fascinates them and makes them want to work on it, you get a sense, they would maybe do it, whether they were being paid or not. And that's from where we get people who really care about problems and come up with exciting solutions. And so it being fun and enjoyable and breaking up in that innovation seems I can understand why that approach has been fruitful and feels like it was fruitful with your most recent release, which is Party Rock. I'm sure our listeners will be familiar with Party Rock. It is a generative app for making generative apps, right, or really all kinds of things. I was playing around with it, I had it come up with a launch plan for a brand new podcast. And it gave me everything I gave me. It gave me topic ideas it gave me like the episode descriptions, it gave me artwork, which was cool. So this is multimodal in a way that other user facing generic tools aren't. I'm getting ahead of myself here. I really wanted to speak to you about Party Rock, because while you will ever see a lot of different products that feels like you will have been uniquely involved in party Iraq itself. So I would love to just know where this tool came from and how it fits into the strategy.
Mike Miller 8:06
Let's start at the beginning because it is got a really fun kind of origin story. And so if you look back at early last year, generative AI, you know, has kind of gone viral, everybody is recognizing that this thing is really going to be a new technology shift. And everybody's kind of figuring out like, how do I get an understanding of what this thing is and how it works. And one of our development teams, a team over in Berlin kind of had had the same thoughts. And so they built a little prototype, what they called a playground and LLN playground to start playing with some of the, you know, the most available MLMs men see how they work, give them prompts and see the outputs, but the sort of light bulb went off for those guys. And they said, hey, what if we build it in a way that it's easy to sort of connect these things and build some basic apps and see how they get connected. And then they made it available for the other folks on their team. And the next thing you knew through like Slack chat messages, you know, the word had gotten out, Hey, you guys got to check out this thing. It's really cool. I recall the day I got a Slack message, somebody told me to go check it out. I went looked at it. And it was a really interesting sort of perspective on getting hands on with these generative AI apps because you could use prompts. But then they had this idea of like widgets that you could kind of connect together so that you could have a prompt in one sort of feed into a prompt of another. You could sort of cut and paste, you know, data in and then like chat about it in a chatbot format, or you could do regular prompts. And so it started to like the light bulb would go off, things would start to click about how this thing worked. And it just started to gain steam internally by word of mouth as like a fun and sort of easy to use way to see the power of these things. And what they did was they also built like a little Application Builder, which itself was powered by generative AI so you could give it like one sentence and it would actually build up a little tiny app for you. So you could say like, you know, create a Story generator that takes, you know, a location and the name of a main character and write me a kid's story, you know, using those topics. And, you know, the app generator would go ahead and build the scaffolding and sort of put the widgets in place for that and sort of come up with prompts to use for that. And then as a user, you didn't just like get like a black box output, you actually got this little app that showed you the widgets and the prompts and sort of how things fit together. And so as this thing gained steam, internally, there was just this lightbulb moment of oh, my gosh, like all of our customers, and even customers who don't know anything about building on AWS can use this thing to really gain some intuition and understanding for how generative AI works. And that was really the birth of Party Rock was how do we take this thing that's, you know, fun and easy to sort of build these applications and externalize it and make it available to everybody? Not just, you know, AWS customers.
Rob Stevenson 10:53
That story is like the the reason that hackathons exist, right? It's like, how do we make this happen, you know, not by accident. In this case, it doesn't tell you what happened at a hackathon. It sounds like it sort of did grow naturally through some people working over there. But products that went hackathons are often very fun and much beloved by employee bases, and more often than not, do not make their way into the product roadmap at all. Right? So at what point did you realize wait, this, we could maybe turn this, you know, outward facing and that that does has utility beyond our own four walls?
Mike Miller 11:29
Yeah, it's a good question. You're absolutely right. I mean, hackathons. I love hackathons, just because it gives our builders, you know, creative freedom to kind of do whatever they want and sort of put things together. I think the difference with this one is that it was a little bit about timing, it was like the right sort of set of things put together at the right time, as well, we were looking for, you know, we were launching bedrock, which is, you know, Amazon's no foundation models as a service platform that allows companies to access these large language models in the cloud not have to host them themselves. And there's a range of sort of supporting capabilities around it. And so it was a little bit of like the right thing at the right time of like, Hey, this is an interesting way to show off bedrock. And in fact, if you look at the name of the product, it's actually called Party Rock, and Amazon bedrock playground, because we realized, like, this is like a super easy way for like, everybody to have a look at like, what bedrock is about what kind of foundation models exist? What are their pros and cons? How do they work? What's the relative like cost of using them, but then put them in this framework that makes it like super easy to generate kind of new ideas, whether that's prototyping new applications, coming up with new creative approaches to problems, or even just like experimenting for yourself? And like figuring out like, oh, how can I kind of put these things together to build something new. So it was a little bit of like, the right kind of tech and presentation, kind of at the right time.
Rob Stevenson 12:49
You know, a lot of attention is being paid to generative lately, because of user facing products, or consumer facing products, rather, that you don't need to be an engineer to use that are no code or low code sorts of tools. And that makes them exciting and accessible. For example, I was speaking with your coworker, Emma, and she was like, anyone can use Party Rock, my mom was using it to make a quick buck out of the things that are in her fridge. And that does feel like a Rubicon moment, right? If Deanna Kwang is using generative tools, you know, then it's not merely going to stay in the domain of this tech bubble and the people talking about it, right. However, at AWS, you have this history of servicing developers do low code and no code tools. disenfranchise developers, how should developers be thinking about this ability for people with zero code experience to be able to put together apps?
Mike Miller 13:49
Yeah, absolutely not. I mean, if anything, it sort of grows, the universe, right grows the pie of kind of cool innovations that can happen. As far as Party Rock is concerned, Party Rock is all about, you know, increasing the reach sort of democratizing AI and ML and making AI and ML and generative AI kind of concepts available to all users, right, because we think, you know, in the long run, innovations are going to come from everywhere, right, from all walks of life, from all skill levels, from different backgrounds, and the more of those people that we can provide the tools and the intuition and the power to sort of create innovations, the better off we all are, right. In fact, we have a few programs in place already to support some of those initiatives where we're, you know, training, you know, millions of users on, you know, AI and ML concepts. And we've taken some pretty big goals to reach some targets for cloud computing and AI and ML kind of technology education for users around the world. So we kind of think of it as the rising tide lifts all boats, right and the more we can democratize this technology, the better off we're all going to be
Rob Stevenson 14:59
Now someone who has had a background in AI and ML is more familiar with some of the pitfalls and trappings and, you know, bad things that can happen when you develop a product. You know, we've all seen the various different kinds of product launches from huge companies, right from like the Googles, and Microsoft's, that just went disastrously. And those were made by specialists and experts, how much oversight is necessary, how much guard railing is there to take folks who have no technical background and make sure that their tool doesn't spit out something nasty?
Mike Miller 15:32
Well, there's a few approaches to this, right. So first of all, at the foundation model level, the way that companies are building and training these foundation models, they're incorporating, you know, safety measures into them. And with Amazon bedrock, which is this fully managed service that makes, you know, foundation models available via an API. There's an acceptable use policy and a responsible AI policy that kind of governs the usage of you know, those foundation models, and these are common things you would anticipate, like, you know, acceptable use policy describes, you know, that you can't use the models for illegal activity, for instance, and the responsible AI policy is that, you know, you can't use things these things for, you know, hate or related topics like that, right. There are also capabilities that Amazon bedrock has built in. So there's a capability called bedrock guardrails, that actually is an additional layer that can be configured on top of the foundation models that can detect and block user inputs or outputs from the foundation models that fall into kind of restricted topics. So for instance, Party Rock does use that we use guardrails to provide an extra kind of layer of safety on top of the capabilities and applications that might be built. You know, we also make it very easy for users to report violations to us, you know, right to the website, you know, and that's been very low, you know, the guardrails have done a really great job about, you know, just keeping things safe and aligned with our acceptable use and responsible AI policies.
Rob Stevenson 16:59
I'm curious how this tool fits into Amazon's kind of wider strategy and how it's thinking about generative, I guess, what is the strategy around generative?
Mike Miller 17:10
Well, I mean, I think if you just step back, and you look at overall, like artificial intelligence and machine learning, you know, for over 25 years, Amazon has been investing and developing these technologies, everything from, you know, customer facing things like the technologies that personalize your shopping experience on amazon.com, to the things that improve our operational capabilities, like aI powered, you know, robots that optimize order fulfillment in our warehouses, right. So there's been a lot of time for us to be working in sort of developing these projects, and seeing, you know, how they're used, how we need to train people to use them, what are the kinds of impacts that they can have, and this is what has resulted in a lot of that strategy, which is focused on, you know, tools like Amazon bedrock, which provide access to the widest range of third party foundation models, and also the goals that we're taking around democratizing the technology, because we can see the transformative nature and the power of it. And we want to make sure that we're doing what we can do to make this technology, easy to access and provide training and help customers gain intuition about each of these technologies as they released.
Rob Stevenson 18:22
democratizing access does exploding who can contribute? And who can make fun tools with this? This is, I've seen as across a few different companies is like, can we put these tools in people's hands and see what amazing things they create, and they don't need tons of resources to do so. That is very exciting to me, particularly as like a content creator. I want to know, like, what people are using it for, like, what kind of things you've seen come out of this already that you're excited about?
Mike Miller 18:46
Yeah, I mean, it's amazing. This is one of those really fun projects and products, where as a product manager, you kind of think one thing, and then you put it out there and you see how people are using it. And you're constantly amazed, I have to tell you, I've been amazed by the range of applications that people are building with this. And in fact, we're hosting the party rock genitive AI hackathon, which actually did we just closed submissions, and we're going to announce the winners in early April. And I think it'd be interesting, we can kind of take a look at like, the hackathon. And like the specific kind of categories that the hackathon highlighted gives you some indication of this. So one category was interactive learning experiences. So can you build apps that actually teach people things right? And it's really fun. One of the folks internally at AWS used Party Rock to build a party rock tutorial that taught you about like how to use Party Rock, and how do you like write prompts? And how do you sort of connect these widgets? So it was this really fun sort of meta experience, but, you know, we can imagine like virtual science labs and language lessons and all kinds of really innovative things that are focused on interactive learning. We've also got a category for creative assistants so we know that you We think about the AI as kind of your companion and sort of this tool that you can use to like, kind of bounce ideas off of and not just like, assume it's going to do a great job and give it 100% of the responsibility. Like, what kind of assistance Could you come up with to sort of use AI as sort of a creative assistant, you know, and then of course, there's entertainment. So we see a lot of people building like, you know, role playing games, or like chat RPGs, or interactive experiences, you know, there's folks using it for actually kind of like writing help, you know, parsing or optimizing paragraphs or tailoring for different audiences, or technical depth levels, you know, summarize in content, some of them are classic sort of generative AI use cases, but maybe tweaked or sort of formed away or packaged in a way that make them kind of more interesting.
Rob Stevenson 20:47
It sounds amazing. And it's very exciting to see what people are making. But even just like the product development stage of IT Standard, very, very fun and light. However, if it was, you may not have a job, I assume that it wasn't just like a straight line, a glide path to success and ease with developing this thing. So I was hoping we could kind of, you know, flip the rock over and look at some of the bugs and some of the challenges with developing this tool. I would just love to know, like, what kind of unexpected challenges you face along the way?
Mike Miller 21:15
Yeah, so I think, you know, we had the concept, right, as we talked about was sort of germinated inside the company. And we, you know, experimented with different approaches, right, the team did about these widgets and connecting them together and seeing what worked and what didn't work. And so we had a little bit of a leg up here, we almost kind of prototyped, and almost did like a little mini like alpha or beta of the product inside the company. So we could kind of see how people used it and sort of what type of things resonated. And just one of the things that we realized was the idea of sharing. And the idea of using an app as inspiration, were two things that we kind of heard about, but we really didn't nail in this sort of internal product. But when we built Party Rock, we kind of said, Look, we need a way for users to super easily share their apps. And that easy way needs to not just be like the app, but also the content, because what we found was like, we'd run apps and you'd like write a funny app, right? And it would come up with some great response and a funny image or funny sort of story. And you'd be like, Oh, this is hilarious. Like, I want to send that to my, my group of co workers and show them what this thing came up with. But there wasn't an easy way to do that. And so we sort of said, hey, when we're building this thing for external users, we want to make that sort of a streamlined and integrated part of what it means to be an app, like what is an app, it's shareable nature. So what you'll see in the Party Rock now that's available today, there's sort of two elements where that sort of surfaced like number one, you've got this ability to publish your app. And publishing your app just gives you kind of what we call a permalink. URL. And so hey, you can paste that URL wherever you want. You can throw it in your social, you can email it to somebody, you can send it in an instant message, whatever, you can post it on your LinkedIn, and other people with one click, they can boom, use your app. The other thing that we added when like use, let's say, you come up with something funny, and you want to share that miss this idea of a snapshot. And the snapshot had sort of like kind of a couple different uses, like number one, like yeah, you can take a picture of like the output of the app, and you can share it, so somebody can see what it came up with. But the other thing that that snapshot does, and the sharing ability does is it allows users kind of a way to get started. And so we named this remix, right? So you can take an app, you can click remix. And in the engineering world, it's like cloning a fork. Yeah, you basically fork that app, you make it your own. But it's an edit mode, I can change whatever I want. So we kind of design this thing such that sharing and getting inspired and seeing what other people make are like core capabilities of this thing. Because when we built this thing internally, we didn't do a very good job with those. And we said, oh, you know, I want to do this, I want to share this funny output, or I want to start with somebody else's app. Eventually, we sort of did those things in prototype. And then they became kind of core elements of the party rock that you see today.
Rob Stevenson 24:00
I want to know, though, like, how did it break? Like, what were the parts where you were like, Oh, this part of it's just not working? We need to fix this before we flip it around and put it public?
Mike Miller 24:08
Yeah, I mean, I think we went through a few different iterations on the UI design, like, how do you make it easy to sort of connect these things, you know, there's a little at sign that you can use to kind of reference other widgets. We improve that a bit because originally, like, it didn't know like, what the universe of widgets was, like, now you hit that sign, it just tells you, oh, which of the three widgets that are in your project Do you want to link to so it's, it's like some common stuff that actually IDs and software developers kind of take for granted that we're now starting to surface for regular users. You know, we're built on top of bedrock right? It's an Amazon bedrock playground. So as new models arrive and bedrock, we add those into Party Rock as well. So you can see like the Jurassic and the sort of meta models are in there now. And you know, we recently you know, announced the new cloud models coming and so when those get to bedrock will surface those things in Party Rock as well. So giving users those options, you know, to play with. And, again, those might be for a little bit more advanced users. But also, we'll be building tutorials, right? Like, hey, how do you know which foundation model to choose? Like, what are sort of the things that you should take into account. And so those are the kinds of things that, you know, we want to grow the product into.
Rob Stevenson 25:13
I do like that it pushes you to use it for a certain use case, which is, I think, a great way to get started, right, like, given a paintbrush in a blank canvas, and most people don't wind up with the Sistine Chapel, you know. And so that is important, I think, to give people an idea of what is possible. I also noticed in the case that I use, which was like develop a launch plan for a podcast, it basically the only prompt it gave me was like, what is the podcast about? It just asked me that. And it didn't ask me to do like, oh, well, what are the first five episodes be? How many hosts? Should there be? What should the title but blah, blah, it was kind of doing all of that on its own. And that to me represented a pretty exciting leap and prompt engineering. There's a term for it, I'm forgetting where it's like, you take someone's prompt, and you just like, change it so that it makes sense to the technology? Do you know I'm saying it's like, you're basically automatically prompt engineering? Do you know I'm talking about?
Mike Miller 26:05
Yep, I forget the term too. But it allows, like you said, kind of you start with one thing, and it's sort of transforms it into what's an
Rob Stevenson 26:11
actual Yes, that's something that prompt engineer would come up with. So how much of that do you think is necessary? Moving on, I think of the way that we like, you, we search on Google, we don't talk to Google the way we talk to people, like we figured out over time, how to construct a Google search? Yeah. Is that going to be less important with generative because of the sort of automatic prompt engineering?
Mike Miller 26:32
That's a really interesting question. It is really interesting. And, you know, I think prompt engineering is almost training people to be more specific, right in their asks, right and what they talk about. But as these models become more advanced, and you have things like the automatic app experience, like in Party Rock, we're encouraging the models to sort of interpret, you know, what the user wants. And, you know, maybe there could be some back and forth to sort of refine, but I think that's super interesting. And one of the features that you'll see in Party Rock that sort of is related to this is that when we launched speaking of like, misses, not really missed, but just a learning, like when we launched internally in sort of the first version. And we surfaced image models, prompting, like using the right kind of words, and the right sort of formatting of the prompt to get the best image generated, took a little bit of skill, and you kind of had to know what you were doing. So what we wound up doing now, as we take the prompt that you give, and we ask the LM to write a good, you know, prompt, and then we actually use that. And if you hover in Party Rock, now we've got the little widget and it's got a little LLM button on the top or a little icon. If you hover over that, it'll actually show you what the prompt was transformed into, that actually went into generating that image. I'd say that's like a version one Dotto of that capability, we want to take that and make it a lot more useful. You could imagine us asking like, Hey, do you want us to, you know, refine your prompt for you, and you know, edit it so that it's a little bit better to deal with this particular model, and maybe show you these things and like an AB scenario, or like, you train you a little bit better on sort of how those things work. But that's kind of an interesting sort of application of what you're noticing, which is sometimes like prompt engineering does come into play, to get the optimal results out of a particular model. And I think with Party Rock, we want to surface that and we want to make it obvious, like, hey, in certain cases, we're going to translate things for you. Or we'll just let you, you know, prompt it yourself. So you can kind of see what actually happens, depending on what you write.
Rob Stevenson 28:27
Yeah, it's like you're asking the tech how to use it. And historically, that looks like someone's snarkily responding to your, your Reddit comment RTFM, you know, instead, it's like, okay, well, hey, we realize this is new, and we're gonna teach you, this thing can tell you how to use it, or it can automatically take what you want into a language and understand,
Mike Miller 28:48
yeah, can it help you improve the way you're interacting with it? You know, along the way, I think we'd love to get there.
Rob Stevenson 28:55
So important, going back to like the Google example. It's like, we've all seen like our parents, or someone who is not like a digital native, their early Google search was like, I would like directions to the mall, please. Thank you, you know, and like Google knows not to index on the words, please. Thank you. And that's a, you know, a proto version of like, auto prompt engineering.
Mike Miller 29:16
Yeah, we've all figured out like, what does keyword search, like look like? How should it work? Right? And we're kind of getting that same intuition for generative AI.
Rob Stevenson 29:26
Exactly. Mike, well, hey, this has been really fun chatting with you. I loved playing around with Party Rock, I'm going to continue to do so I'm going to continue to recommend that moms do so. And it's exciting that like, Hey, this is a very technical podcast, and it's for a typical individual, but opening up access to people and seeing what they come up with can only result in more innovation. And it feels like that's the point of Party Rock. So I'm really pleased that it exists and that you came to talk about it today. 100%.
Mike Miller 29:48
And I think what we see with the range of users, you know, using Party Rock, all skill levels, even like seasoned developers, you know, are trying it out. They're throwing code in there sort of talking about design As they're asking technical questions, you know, I think everybody has their own little creative outlet as well. And that's where Party Rock really shines is in some of those more creative areas like coming up with new ideas or exploring or optimizing. And so I think everybody who listens, this podcast and their relatives should give it a shot. I mean, one of the neat things which we didn't talk about is how you access this thing, right? And so one of the things when we designed Party Rock was we said, look, we want there to be like a zero barrier to entry for this thing. So instead of requiring an AWS account, which you know, has a bunch of hoops to jump through, and you need a credit card, and all that kind of stuff, let's just make it login via your social account. So today, well, your Amazon account, yeah, your amazon.com your Google account, you just use those you log right in, it's in a web browser, you know, for limited time we have a free trial, which includes a bunch of credits that is quite generous it gives you quite a bit of usage of Party Rock without having to worry about you know costs you know, we've got a wide range of models from bedrock so it's a really powerful and just easy to get going way you know, to get your hands on so that's definitely a big plus.
Rob Stevenson 31:04
Yeah, and right now if you enter offer code how AI happens you get unlimited free credits to begin. No, I'm kidding. It's not not the case, but you won't need them because it's free already.
Mike Miller 31:14
So your listeners into the video, my eyes went really wide.
Rob Stevenson 31:18
If you look under your seats right now, everybody gets a pretty rocky count. Everybody you get one you get one. No, no, plenty of it. It's free out there to try. So I recommend folks go check it out. And hey, Mike, thanks for doing this man. I really enjoyed chatting. This has been a fun one. So thanks for being here today.
Mike Miller 31:33
Yeah, happy to be here. And thanks for having me on. Rob. I really enjoyed our chat. Thank you.
Rob Stevenson 31:38
How AI happens is brought to you by Sama. Sama provides accurate data for ambitious AI specializing in image video and sensor data annotation and validation for machine learning algorithms in industries such as transportation, retail, e commerce, media, med tech, robotics and agriculture. For more information, head to Sama.com