So how does this whole AI generated content and scoring work?

This has been the year of the LLM (large language model). There have been so many new ones being put out, and even companies that previously were using LLM’s like GPT-3 (e.g., Copy.AI) are now promising that they’re going to be building their own LLM’s that will automatically generate copy.

The amount of data available to train models has never been greater, and people are going all out training larger and larger models. They basically feed the computer examples of text and the computer learns from those examples and is able to generate similar text on its own.

It sounds simple enough, however the reality is much more complicated.

SlayerAI is slightly different. We don’t generate any text, we just read text and determine if it’s engaging or not. To do this, we study millions of data points… both the text and its’ engagement rate (whether that be replies, retweets, opens etc), and then we train a model to try and recognize high performing text, without seeing its’ engagement rate.

The models we’ve trained are very good at this. We look at different target audiences and can predict 4/5 times whether or not text will be highly engaging.

There are two keys to this process…. One is how we ingest and manage our data sources, organizing them by interest and maintaining tags for how well performing the text is. The second is how our AI works and learns. It incorporates natural language processing into its’ algorithms so it can actually tell whether or not text is high performing or not. Pretty incredible stuff.

We have curated millions of data points from around the web, and email campaigns, so you never have to worry about having a poorly performing campaign again.

If you’d like to try the product, simply click login at the top right and create an account!