If you have been alive at any time during 2023, you have likely heard about generative AI and how every business is experimenting with the technology to identify and implement efficiencies.
It is pretty much been the only topic of conversation during Amazon Web Services’ re:Invent 2023 conference in Las Vegas this week, as the hyperscaler has been building tools aimed at delivering generative Ai to more of its customers.
One of the tools it is making more widely available to the public is PartyRock (no affiliation to LMFAO), which is a platform designed to make the creation of applications wildly easy for newcomers, and ultimately spark a desire to experiment, as well as prompt further interest in the field.
AWS debuted PartyRock earlier this month, describing it as an, “intuitive hands-on, generative AI app-building playground”.
“In just a few steps, you can create a variety of apps to experiment with generative AI. For example, you could build an app to generate dad jokes on a chosen topic, create the perfect personalized playlist, recommend what to serve based on ingredients in your pantry, analyze and optimize your party budget, or create an AI storyteller to guide your next fantasy role-playing campaign,” the hyperscaler posits.
We got some hands-on time of our own with the platform, and like anyone who has an internet connection and a Google, Apple, or Amazon account, we tested out whether it is as easy as AWS is touting it to be.
The results are immensely impressive, with PartyRock only needing a simple description of the app you’d like created. In our case, we wanted to make an AI chatbot that would be able to provide assistance to those who had tech support queries for consumer electronics. The prompt was simply “Help people with tech support for consumer electronics,” and in less than one minute our simple chatbot was created.
From there things get a little more interesting, and you can start customising the application to refine, or remix as AWS puts it. The application we had PartyRock generate featured two different input methods, with a live chatbot input akin to ChatGPT, as well as user input window that featured summary and detailed answers based on large language models (LLMs).
On this front, PartyRock allows you to choose which model you want to make use of. At the moment, Anthropic’s Claude is set as the default, with AWS and Anthropic collaborating quite closely these days.
It’s unclear whether Claude 2.1, which was unveiled last week, will be added to the mix down the line, but the choice of different models all depends on how you want the interface and response feedback of the app to be for users. In the case of Claude, answers are a little more “human-like”, with the 2.1 version of the model said to be even more so.
Tinkering around with the AI app, we can edit the welcome message it greets new users with, as well as how it handles prompts that present difficulty or simply cannot be answered. The likelihood of the latter is all dependent on time, as PartyRock does not have the latest information at its fingertips. Within the context of our Hypertech Support AI app, it cannot provide feedback on content specific to 2023.
To that end, we can ask it how to reset a router, or the difference between 2.4GHz and 5.0GHz WiFi, but we cannot find out South African pricing for the iPhone 15 Pro. The iPhone 14 Pro, however, is not an issue at all.
Other options available in the advanced settings are the temperature, which dictates the randomness of the outputs generated, and the top P, which is how tokens are prioritised for a more focused or diverse level of responses. These settings are not for novices though, but are at least interesting to have at your disposal in seeing just how effective an AI app would be in the real world.
Even though there are some limits to what PartyRock can do, in terms of ease of use, along with serving as a great introduction to what generative AI can be used for, it is a handy tool to leverage if you simply want to experiment.
If you want to take your experiments in generative AI outside of models like ChatGPT to see if you can create an app for yourself, PartyRock is a great starting point.