As generative AI continues to dominate the headlines, it’s exhausting typically to seek out precise working enterprise use instances among the many hype. Writer is a San Francisco startup that’s working to create generative AI writing merchandise with the enterprise in thoughts. Right this moment, the corporate introduced a brand new functionality for its Palmyra model that generates textual content from pictures, together with graphs and charts, they name Palmyra-Imaginative and prescient.
Might Habib, firm co-founder and CEO, says that they made a strategic resolution to focus on multimodal content material, and with the ability to generate textual content from pictures is a part of that technique. “We are going to be focused on multimodal input, but text output, so text generation and insight that is delivered via text,” Habib informed TechCrunch.
By following that guiding star, the corporate determined to investigate pictures, moderately than produce them (not less than for now). She reserves the correct to create charts and graphs sooner or later from knowledge, however that’s not one thing they’re doing in the meanwhile. This specific launch is targeted on producing textual content from these sorts of pictures.
The corporate makes use of a a number of mannequin method to provide the Palmyra Imaginative and prescient outcomes, the place every mannequin has a particular job to do in figuring out what’s within the picture after which producing the textual content with 4 nines of accuracy, in line with Habib.
This has plenty of use instances together with an eCommerce web site producing textual content from 1000’s of adjusting pictures to populate the web site with the most recent merchandize with out having a human sustain with each change, or deciphering key takeaways from charts and graphs routinely. One other instance is a compliance checking. As an example, a pharmaceutical firm may use Palmyra-Imaginative and prescient to carry out an automatic FDA compliance test towards advert copy, ensuring that the advert is compliant with FDA rules as outlined in an related doc, as within the instance beneath.
Palmyra-Imaginative and prescient instance for pharma firm checking advert towards a doc with FDA necessities. Picture Credit: Author
Lastly the product can interpret and summarize handwritten notes into textual content, however Habib says that it requires coaching the mannequin for particular person use instances corresponding to medical or insurance coverage, in order that the accuracy is there.
Habib says that she doesn’t suggest utilizing these instruments and not using a human evaluation as a part of the workflow. She believes that is completely important as a result of any mannequin can hallucinate (make issues up) or just get information flawed, and it’s essential to have folks checking the outcomes. Whereas they at all times suggest this to each prospects, and most perceive it at this level, she believes that it’s finally going to require a extra automated workflow to make it occur constantly throughout prospects, one thing she says they’re working in direction of.
The corporate has raised $126 million thus far, per Crunchbase data. and is at the moment speaking to the massive cloud infrastructure platforms about partnering as they try and scale the corporate. Its most up-to-date spherical was a $100 million Series B final September led by Iconiq.
The newest Palmyra launch with the picture to textual content capabilities is out there beginning as we speak.
