IoT Using Generative AI: 3 Generative AIoT Applications Beyond Chatgpt

Generative AIoT applications beyond ChatGPT that you need to know right now in 2023

On November 30, 2022, OpenAI released ChatGPT. The chatbot, which is based on the large language model GPT3 went viral and set records for the fastest-growing consumer application in history, with 100 million active users in just two months.

Surprisingly, many people are unaware that the technology underlying ChatGPT is not new. The transformer architecture is based on a 2017 Google research paper titled "Attention Is All You Need." As of early March 2023, the paper had been cited over 67,000 times and can be considered a source of generative AI innovation.

In this article, we have explained the IoT using generative AI and 3 generative AIoT applications beyond ChatGPT.

So, why has OpenAI's ChatGPT become so popular even though it is neither new nor unique?

The answer is ChatGPT's ease of use and accessibility, as it was the first cutting-edge large language model made freely available to anyone with internet access.

Key Use Cases of Generative AI

Generative AI can have an impact on connected devices, typical IoT use cases, and IoT technology in general as a market research company focusing on IoT and IoT-related topics. Our findings were published in the Generative AI Trend Report. The report discusses the technology in general, examines the competitive landscape, and then delves into generative AI use cases related to IoT scenarios. Here are three of the nine use cases we identified at the crossroads of generative AI and IoT:

3 Generative AI Applications for IoT

Here are the top 3 generative applications for IoT

Application #1: Code generation for IoT

Large language models can be used to create, complete, or combine software code based on code snippets or natural language descriptions, and they can be applied to a wide range of domains, tasks, and programming languages. With such capabilities, these models can help both professional and inexperienced developers build innovative applications. Many IDEs already support generative AI. Although it is already used by software developers, many believe that generative AI will not replace developers anytime soon. Consider it another tool in the toolbox for code generation, similar to the no-code/low-code tools that have recently been added to the general software development toolbox.

Application #2: Robot control

Generative AI may have an impact on how autonomous (IoT) devices, such as robots, are controlled. Generative AI can generate control logic and commands for robots by capturing motion data from animals or humans. Instead of deterministically programming movements for each leg of a robot dog, for example, generative AI models can be used to generate individual part movements and make the robot dog walk complex, interconnected steps. Furthermore, generative AI models can assist robots in understanding their surroundings and connecting so-called horizon goals with more intermediate steps to achieve the goals (e.g., filling a glass with juice). Robots can then generate intermediate tasks without requiring human intervention to reach the higher horizon task.

Application #3: Social IoT devices

Today's IoT-connected devices enable users to access data via an API. Typically, these provide predefined sets of information, such as usage, battery, and asset health. However, generative AI has the potential to make device communication "more social" in three ways:

Allowing the device to respond to complex questions posed by the user.

Allowing the end user to communicate with the device to change settings.

Allowing the devices to generate answers using generative AI. The below example helps you understand how a large language model can be used to guide a stuck robot after it has been given unclear instructions.

Example

Amazon created the DialFRED framework to allow robots to ask questions when they are unsure. The questionnaire model is fine-tuned through reinforcement learning to ask the right types of questions at the right time to benefit task completion. The framework includes an "oracle" that generates answers to generated questions automatically using ground-truth metadata from the simulation environment. As a result, DialFRED offers an interactive Q&A framework for training embodied dialogue agents. ​