As generative AI continues to dominate the headlines, it can be difficult to find working business use cases among the hype. Writer is a San Francisco startup working to develop generative AI writing products with enterprises in mind. Today, the company announced new features for its Palmyra model that generates text from images, including graphs and charts. This is called Palmyra-Vision.
May Habib, the company's co-founder and CEO, said the company has made a strategic decision to focus on multimodal content, and being able to generate text from images is part of that strategy. Ta. “We're going to focus on multimodal input, but we're going to focus on textual output, meaning text generation and the insights provided via text,” Habib told TechCrunch.
Following that guiding star, the company decided (at least for now) to analyze images rather than create them. She reserves the right to create charts and graphs from the data at some point, but she is not doing so at this time. This particular release focuses on generating text from such types of images.
Habib said the company uses multiple model approaches to generate Palmyra Vision's results, with each model determining what's in an image and producing text with four-nine precision. I am doing a specific job.
This includes e-commerce websites that generate text from thousands of changing images to add the latest products to a website without a human having to keep track of every change, or capturing important points from charts and graphs. There are many use cases, including automatic interpretation. Another example is compliance checking. For example, a pharmaceutical company uses his Palmyra-Vision to perform automated FDA compliance checks on ad copy, ensuring that ads comply with FDA regulations outlined in relevant documents, as in the example below. You can check that you are compliant.
Palmyra-Vision example where a pharmaceutical company matches advertisements to documents with FDA requirements. Image credits: Writer
Ultimately, the product will be able to interpret handwritten notes and summarize them into text, but Habib said the model will need to be trained for specific use cases, such as healthcare and insurance, to ensure accuracy. says Mr.
Habib says he doesn't recommend using these tools without human review as part of your workflow. She believes this is absolutely essential. This is because any model can be hallucinatory (fabricated) or simply get the facts wrong, so it's important to have people check the results. They've always recommended this to all customers, and most people understand it at this point, but eventually running this consistently across multiple customers will require a more automated She believes there will be a need for workflow and says she is working towards it.
The company has raised $126 million to date, according to Crunchbase data. We are currently in discussions with major cloud infrastructure platforms to partner with them to help us scale our company. The most recent round was a $100 million Series B led by Iconiq last September.
The latest Palmyra release with image-to-text conversion capabilities is available starting today.