There’s a lot of buzz right now about how generative AI like ChatGPT and Bard will revolutionize many aspects of the web, but companies with narrower niches are already finding success. Writer is such a company, and has just announced three of his new large-scale language models that power the Enterprise Copy Assistant.
The company allows customers to fine-tune these models based on their own content and style guides, and from that point on AI creates, assists in, or edits the copy to meet internal standards. increase. Writer’s new model not only catches typos and recommends words you like, but it also lets you assess style and create content yourself, and even do a fact-check when you’re done.
But the real draw, at least for the two smaller models in the Palmyra range, is that everything from tweaking to hosting can be done under the hood.
CEO May Habib said in a press release: “We offer our customers all the benefits of the AI application layer without the risks of other her AI applications and commercial models. I would like to invest in
Palymra comes in three sizes. 128 million, 5 billion and 20 billion parameters for Small, Base and Large respectively. It doesn’t surprise me at first because I’m trained in business and marketing articles, not Reddit posts or Project Gutenberg. Then, load up on the last 10 years of annual reports, financial information, blog posts, and more, and make it yours. (For clarity, this and derived data are not filtered back to the Writer.)
Having written corporate and marketing copy, I would say this is not the most exciting application. But what it lacks in thrills, it makes up for in practicality. Companies tend to need and actually pay for this kind of writing and editing a lot, and since Writer is already built into many development and productivity suites, it doesn’t add much friction.
A mockup of a writer that generates a product description.
The business model is similar to other generative AI companies. All the setup and tweaking is free, and you pay 1 penny for every 1000 tokens and you get about 750 words. (This article is just over 500 as a quick reference.)
Alternatively, if you have the computing power, you can self-host a small or base model for free.
Dozens of companies have been using this model since the end of last year, and we haven’t heard of any serious problems like we saw on the first day Microsoft and Google tried to popularize generative AI. This is a good sign. This is the success we talked about earlier. ChatGPT is certainly impressive, but as a generalist or hobbyist AI, it’s hard to say what it could actually be used for. While Microsoft and Google are kicking tires with their latest toys, the next year or two will see more targeted play like Writers.
