Writer deploys large in-house language models to enhance corporate copy.

There’s a lot of buzz right now about how creative AIs like ChatGPT and Bard can improve different aspects of the web, but companies targeting narrow verticals are already finding success. Writer is just that, and Enterprise Copy has announced three new big language models to power its assistant.

The company allows its customers to customize these models according to their own content and style guidelines; From this point on, the AI ​​can dictate, write, or edit content that meets internal criteria. More than just catching typos and recommending preferred wording, the author’s new models can assess style and write content themselves, with a little fact-checking when they’re done.

But the real draw is that everything from fine tuning to handling can be done internally, at least on the two Palmira series models.

“No corporate leader wants to be fodder for someone else’s foundation model, including ours,” CEO May Habib said in a press release. “We provide customers with all the benefits of an AI application layer without any of the risks of other AI applications and business models. Enterprise leaders want to invest in solutions that provide their own LLM.”

Palmira comes in three sizes: 128 million, 5 billion, or 20 billion units, respectively, for Small, Base, and Large. They’re trained in business and marketing writing, not Reddit posts and Project Gutenberg, so there aren’t many surprises to begin with. Then you load the mammoth with the last ten years of annual reports, financials, blog posts, and more to make it yours. (This and any derived data do not filter to the author for clarity.)

After writing my enterprise and marketing copy, I can say that this is not the most exciting of applications. But what it lacks in glamor it makes up for in practicality: companies have to do a lot of this kind of writing and editing, and they actually tend to pay for it. The author has been involved in several development and productivity suites, so not much friction has been added.

The business model is similar to other generative AI companies: you get it all set up and edited for free, then you pay a penny per thousand tokens, which gets you about 750 words. (This article is over 500, as a quick reference.)

Alternatively, if you have the computer, you can install the small or base models for free.

A few dozen companies have been using the models since late last year, and we haven’t heard of any serious problems since the early days when Microsoft and Google tried to popularize generative AI… so that’s a good sign. This is the success I talked about earlier. While ChatGPT is certainly impressive, it’s hard to say whether it can really be used as a generalist or dilettante AI. The next year or two will see more targeted plays like Writers, as Microsoft and Google kick the tires on their latest toy.