How to Create Traceable Generative AI in an Era of Untraceability
ChatGPT’s mainstream introduction in 2022 kicked off an era of Generative AI that has brought both awe and apprehension. Its ability to generate human-like text, images and music has undoubtedly reshaped industries and opened new possibilities. However, with these advancements comes the pressing concerns of untraceability, where the origins of AI-generated content and training data remains obscured, raising ethical and accountability issues.
Though much of the mainstream discussion surrounding Generative AI has been focused on limiting its sentience, a more pressing undiscussed issue is the blurring boundaries between human and AI creations and the need for more traceability and transparency. One year later, the situation on the surface surrounding Generative AI remains largely the same. We don’t know if what we are viewing is human or AI generated.
In the midst of this technological excitement, Numbers Protocol has been working behind the scenes to serve the provenance needs of anything digital so when Generative AI came into the picture, we were more than ready to accommodate. Today we are going to showcase how Numbers Protocol works in tandem with AI platforms to create traceable Generative AI content.
Prerequisites
We will not be writing any code in this showcase as we will be using Instill AI’s Versatile Data Pipeline (VDP) to construct our traceable Generative AI. Instill AI provides no/low code tools to convert unstructured data into meaningful data representations. If you’re interested make sure to check them out here.
You will need to create an Instill AI account in order to create the data pipeline and a Capture Account to create on-chain records. Since we are interacting with the Numbers blockchain to create on-chain records we will need NUM token to facilitate these transactions. New Capture accounts are credited with 12 NUM Credits which equates to 120 registrations at no cost.
Here’s our tutorial video, if you prefer to watch a video.
Create a Capture Account
Create a Free Instill AI Account
Add New Pipeline
From the Instill Dashboard navigate to Pipeline & Click on “Create pipeline”
Set up Pipeline Properties
Instill AI will automatically assign a Pipeline Name. Rename to something more descriptive.
Set up Source & Destination
Adding AI & Blockchain Components
From the components bar click search for Instill Model and Numbers Protocol and drag it onto the pipeline editor
Create and Configuring Connectors
For Connectors to be functional you will need to input in required API keys or tokens. On each component click “Create Connector” and fill in the required information. We will need the Capture Authorization token for the Numbers Protocol connector. Since we will be using the Instill Model Connector we do not need any additional information, however if you select other AI connectors such as OpenAI, you will need to gain access to an API key.
Set up the connections
Configure your pipeline like shown above. Text-to Image takes in text input which is fed into Stability AI to output an Image. This Image is stored along with metadata inputs onto the Numbers blockchain. Since we are interested in also recording our prompt we can also feed the text input Numbers. As a final output we are interested in seeing our final image and the metadata stored on-chain
Configuring the on-chain records
To ensure the on-chain commits are comprehensive you will need to set the values. Do so by navigating to the Numbers Protocol connector component and clicking on “more”. Doing so will open up a window with commit parameters such as license, mining_preference, generatedBy, etc. Fill these with the appropriate values and be mindful that as on-chain records they will be immutable. To learn more about each of these parameters and what they represent in terms of digital provenance see here.
Activate
Click “save” in bottom right to save and initialize our pipeline. Open up the pipeline properties window to view details on how to trigger the pipeline
This pipeline creates an API endpoint that we can initialize by sending a POST request. Before we do that we will need to set up a free API token by navigating to Settings > API Token
Executing the Pipeline with Instill Editor
The pipeline can be executed with an API call utilizing your preferred API platform or sending a curl request. For simpler execution of our pipeline we can utilize the in-editor “Run” function. Enter in the inputs (prompt and other defined parameters) in the start operator and click “Run”. If the execution is successful we will see our results in the end operator. For our pipeline our outputs are the final image and the asset-profile link. The asset-profile link is where we can view the assets full history and most importantly its on-chain records.
Viewing the Asset Profile
Reviewing On-Chain Records
From the commit table click on the transaction opens up the transaction record on Numbers Blockchain. We can click on UTF-8 under Call Function to display the metadata that was recorded on-chain.
For a better view of the metadata we can click on the Asset Tree link from the asset profile commit table. The Asset Tree is a formatted file standard containing the assets information. It contains information like license, creator, encoding format, etc.
That’s It!
In this walkthrough we were able to set up a fully operating data pipeline that takes a text prompt and generates an image utilizing Stable Diffusion AI model. Unlike other AI generated images, it was also registered with a unique ID onto the Numbers Blockchain creating an on-chain record defining the images license, creator, and most importantly how it was created. In other words, we have created a fully transparent and traceable WEB3.0 AI Generated asset that is able to standout in a sea of untraceability.
Next Steps
If you are interested in learning more about Numbers Protocol check out our documentation. Also consider joining our Telegram and Discord Community to get the latest news on Numbers Protocol. If you’re a developer and have questions, we would love to hear from you! We have more developer related content coming soon. So stay tuned!
About Instill
Instill AI, founded on June 11th, 2020, is an innovative startup dedicated to democratising access to AI. We specialise in developing AI and data infrastructure to offer a streamlined, no-/low-code solution for users seeking to rapidly extract insights or build AI-first applications from varied data types such as text, images, audio, or video.
We are the primary developers and maintainers of the Instill Core open-source project, which encompasses Instill VDP and Instill Model. These integral components form the backbone of AI-first application development via low-code/no-code interfaces. Leveraging the success of Instill Core, we introduced its cloud service, Instill Cloud, in May 2023.
As a provider of no-/low-code tools, we empower our users to efficiently handle and manipulate all kinds of data — unstructured, semi-structured, and structured. By integrating our services into their existing data stack, users can unlock the full potential of their data reservoir and seamlessly benefit from AI advancements.
We are a small and ambitious team of passionate engineers/researchers sharing an unconventional culture fused by DevOps, MLOps and academic research labs. We are hands-on and love to automate everything. We care about our products and deliver to the highest standard.
Website | Discord | Linkedin| Facebook | Twitter
About Numbers Protocol
Numbers Protocol is the first asset-centric cross-network protocol and handles 6 billion gateway monthly traffic from 190 countries and is already being used and trusted by art, music, NFT platforms, metaverse, and is developing an application for AI companies to trace digital media provenance. Products including:
- Capture App: The world’s first blockchain camera.
Allows users to secure photo and video provenance with on-chain integrity records, providing a transparent, immutable history of digital content. - Numbers API: Developers and enterprises can implement Numbers API to register photos and access their Web3 addresses and certificates.
- Numbers Search Engine: Google for digital media.
Boasting the highest multi-network asset coverage, reverse-image search, dedicated asset profiles and AI-powered similar image results. Provenance of digital media can be found– ideal for general queries and vetting.
The rise of AI has brought an unprecedented amount of data being generated and processed. Ensuring data traceability has become an increasingly important topic of discussion in the tech industry. Through collaboration and the development of innovative solutions like Numbers Protocol, we can ensure that the use of AI is transparent and responsible.