Jasper Shows How a Generic GPT-3 Model Gets Customized for Enterprise Users
Beyond a generic AI writing assistant
When ChatGPT came out, some people quickly suggested that OpenAI’s software partners that offer writing assistants were doomed. Jasper AI, the highest-profile company in the category, was often highlighted as a company that might be in trouble due to OpenAI suddenly competing in a similar market.
Shane Orlick, president of Jasper AI, didn’t see it that way. In an interview on the Vociebot Podcast he said:
“I think it’s great at showing the magic trick…Ask it anything and it will give you an answer…It’s amazing for donig that. I think it’s been great for turning people onto generative AI and understanding the power of this. All of the sudden I think our market is 20 times larger that it was before ChatGPT came out in November.”
He also indicated that the company was working on provisioning multiple AI models for customers, including several fine-tuned versions of GPT-3 and other models. And he said that the company would soon have APIs with features beyond what OpenAI offers. ChatGPT just made more people aware of what large language models can do. Companies like Jasper AI were suddenly setting daily records for new customer signups every week in January.
Differentiated vs Commoditized
The people calling Jasper and the other AI writing assistants doomed when OpenAI began to offer the premium paid service ChatGPT Plus were expressing a fundamental misunderstanding of the market. First, many assumed that AI writing assistants merely put a user-friendly interface on top of GPT-3. Others assumed there was not much to offer beyond what ChatGPT was doing.
Some software vendors are simply slapping a new UI on top of GPT-3. But many others are differentiating on top of base models and offering options beyond what OpenAI offers. Orlick commented:
“Think of Jasper as this application layer that’s sitting on top of many different models, including our own models. We’re able to pick and choose which pieces of those models we want to support in order to create the right piece of output for our customers.”
Eric Schwartz from Voicebot summarized the Jasper for Business features, saying:
Jasper for Business leverages a version of Jasper Chat, a conversational version of its generative AI engine. The suite includes AI tools specifically built for a business and incorporates information about the company and its products, goals, and history. This Jasper Brand Voice aims to fine-tune the standard large language model (LLM), overcoming complaints about generic AI outputs. Instead, the responses account for the business as a unique entity, including their house style of writing and tone.
The Jasper Everywhere feature complements the custom AI by enabling a business to incorporate Jasper into its existing online presence. That means a browser extension for Google Chrome and Microsoft Edge that can perform in any channel. Users will see a purple cursor in their email, social media, content management systems, and other apps. It streamlines access to Jasper’s features far more than the existing pop-up window and the need to copy and paste the text. Jasper is also launching its first API, providing an option for businesses that want to integrate generative AI more directly.
Enterprise Access to LLM Features
The new generative AI products highlighted yesterday from Veritone represent packaged applications that address specific enterprise tasks. They provide a new UI, workflow, and other features that complement or replace existing business processes. Jasper’s first product, the AI writing assistant, follows a similar pattern.
Jasper Everywhere is designed to enable new capabilities. Those could complement the existing web presence, represent a new way to render website content, or serve as an API that existing enterprise applications can tap into to support new features.
It is unclear today how this market will play out. Will enterprises connect directly to the raw large language models (LLM) offered by OpenAI, or will they opt for tailored models and services provided by companies like Jasper AI? I suspect it will be a mix.
Some companies will have in-house capabilities sufficient to roll out their own integrations and features. At the same time, others will want more services wrapped around the LLMs and prefer not to be locked into dependency on a single model provider. It may be that the key value provided by Jasper AI and other model-independent providers is helping other enterprises avoid model lock-in.