The training data that guides INK writing comes from a variety of sources, which makes our AI writing more performant than any other solution on the market.
The Base Model
The base model of INK is a very large language model which has read a large portion of the internet in mid-2021.
Since the internet contains both the best and worst of humanity, we filter all user inputs and AI outputs for possibly harmful or discriminatory language.
Based on your intent, INK is trained on expert examples from top copywriters.
For example, INK knows when to write a headline or intro just because that's where your cursor is. In that case, our base model is augmented with expert examples that teach our base model what a fantastic headline or intro looks like.
INK reads the content you wrote in your article so far. It will learn both from the information and the style (your unique tone and voice).
While the base model does not know what happened since mid-2021, INK learns updated facts from your content, and then use that to write about it.
Tl;DR: INK learns from what you wrote so far. Add a paragraph of new facts in your article to teach our Base Model the latest.
INK can learn new information from your content, for example:
The FIFA world cup of 2022 took place in Qatar. It was the first world cup in the Arab world. There were 1.5M visitors, who arrived on 1,300 daily flights. The world cup stadiums have 26 more doors than the White House, and use 84,000 tones of structural steel. At $220B, this is the most expensive World Cup ever.
Based on your SEO keywords, our patented technology decodes the patterns of how to perfectly optimize your content for your audience, and then teaches the AI how to write.
- What information is important to your users but you didn't cover yet
- What information the Google algorithm thinks you're missing to rank
- What exact emotions your audience is craving for
- What precise reading difficulty is ideal for your specific audience
- What tone & voice will your specific audience resonate with
We call this technology Natural Language Optimization (NLO). It has been in development since 2016, and is protected by multiple patents. Only INK knows how to write for your audience.
Even if the base model doesn't know recent facts, we devised a way to teach it new information sourced from the web.
Based on your keyword, INK goes to the web to read as much as it can about your topic. All this happens in real-time. In the process, it may learn a thing or two about the topic that neither you nor the base model know.