Begin’s Graph of People & Objects
Begin technology's purpose is to help businesses provide their customers with a fantastic online experience and provide everyone access to AI-enabled technology.
To do just that, we designed a graph that can describe people and how they interact with the world around them. It's all about the people, after all. Isn't it?
This graph has many benefits to the businesses and their consumers in the short and long term:
It is the most efficient method to launch machine learning models to production currently available in the market.
Mid and Long Term:
The graph gets more and more intelligent the more businesses use it, and the more consumers are served through it, meaning everybody's machine learning models grow exponentially smarter over time.
As the graph grows, it unlocks co-learning opportunities between businesses, securely and without directly sharing data (see privacy in the graph).
To make this graph possible to use instantly, we designed a platform around it that allows companies to integrate and evaluate machine learning experiences for their customers very easily.
So how it works?
See a sample of information in the graph above. Notice the following:
- The graph captures how a person interacts with objects and their sentiment about these interactions. Did they like this book?
- The graph has several algorithms that allow you to:
- Recommend future relationships. e.g. ( based on a person's reading history, what books might they like? )
- Predict future relationships. e.g. ( based on this person's reading history, how much will they like the "Life of pie" book if presented to them? )
- Identify fake items and relationships. e.g. Is this person a fake profile in my app? Is this a fake purchase by a bot?
- You can also build your custom algorithms on top of the graph, such as "can you guess the price a person might be willing to pay for a service?"
- Groups work together, i.e., imagine a team of people deciding on purchasing a service. You can predict such sentiments and decisions too (more to come on this soon).
- All data is entirely private. The graph's knowledge grows over time across all businesses. The data that generates the knowledge remains on devices where people interact with objects. Begin's SDKs mask all information in the graph through mathematical signatures that cannot be reversed to the original data, making all businesses' and consumers' data private.
- All algorithms work in a decentralized fashion, allowing for speed and online learning without building cloud pipelines or pre-engineering datasets.
- We've come to learn that this design performs better than traditional machine learning methods every time.
- The graph is designed to scale infinitely to hundreds of millions of people and their interactions without becoming data centres or cloud heavy, because any subset of computations in the graph can be decentralized at any time.
- Ultimately, if granted access by service providers and graph operators, a consumer may be able to write their own algorithms and be their own scientist to enhance their experience through service optimization.
- You can have a graph by describing your customer's interactions with objects in your business and installing Begin's SDKs.
- You can use the graph's algorithms to predict relationships, detect fake profiles, serve recommenders or launch your custom algorithms on top of the optimization infrastructure.
- All algorithms continue to learn automatically through the graph dynamics, including ones you write by yourself (more details coming soon).
- You can optionally opt to grow your graph's knowledge by collaborating with partners securely who may have more knowledge about your customer or domain. You do that by simply linking the two graphs using secret linking algorithms available to you.
- Begin SDKs do not copy any of your customer data or your datasets. All graph interactions and data are processed locally on end-user devices where knowledge is merged mathematically in a privacy-preserving format across the network.
- You can operate a subset of the graph privately to service your businesses, companies, and more.
Using the platform is a simple 3 steps process.
- Describe your objects and interactions schema, and upload it to your Begin’s application account (request one, if you don’t have one).
- Set up an objective, for example “predicting a relationship” to tell if a user will like a new book or subscribe to a new service.
That’s it, you’re live and learning and you can monitor the performance of your models in your dashboard. Typically it takes about 2 hours to be able to start using the model to predict new relationships in this case.
The graph, algorithms and SDKs are technology designed and owned by Begin AI Inc. Any algorithms you write on top of the graph is owned by you unless you decide to publish it and share it with other graph customers.
Your data remains yours. Begin never copies your data from its source of origin, all objects and interactions are processed locally on the devices they originate using the SDKs.
If you choose to collaborate with other members of the graph to make your algorithms smarter, the agreements you’ll have with these collaborators govern your relationship with them - contact us for more details on collaboration on the graph.
Relevant resources and some techniques used at Begin:Why Decentralized LearningProcess on Device, Train CentrallyFederated LearningKey Terms