Author, the three-year-old San Francisco startup that has, since its inception, centered on constructing out a safe generative AI platform for enterprises to assist them throughout capabilities, introduced it has raised $100 million in Sequence B funding at present, in a spherical led by ICONIQ Progress with participation from WndrCo, Balderton Capital, Perception Companions, and Facet Ventures.
Notably, Author’s enterprise purchasers together with processional providers firm Accenture and funding big Vanguard additionally joined within the spherical, signaling sturdy confidence in Author’s choices.
The cash will assist Author proceed to construct out extra new options and AI brokers for its “full stack” of enterprise-friendly software program functions and connections round its three proprietary AI fashions: Palmyra Small (128M), Palmyra Base (5B), and Palmyra Massive (20B), which may already assist enterprises do all the pieces from create content material and emails to transcribe audio and summarize it.
Author’s Palymra fashions help with capabilities throughout enterprises
Whereas the B2B software program market has for the final six months been completely inundated with new generative AI instruments and functions — a lot of them making use of huge language fashions (LLMs) equivalent to OpenAI’s ChatGPT or Meta Platforms’ Llama 2, Author goals to face out with its in-house trio of Palmyra fashions.
The corporate mentioned its press launch at present that it’s focusing on folks working in “help, operations, product, gross sales, HR, advertising and marketing, and extra,” capabilities.
Among the fashions are tailor-made for particular industries like healthcare and are totally auditable—a novel promoting level in a subject typically criticized for its “black field” nature.
Educated on non-copyrighted, enterprise-friendly information sources
Author says its Palmyra fashions are educated on upwards of “1 trillion tokens of formal and enterprise writing containing no IP or copyrighted content material,” in line with an e-mail from Might Habib, CEO and co-founder of Author, offered solely to VentureBeat by a spokesperson.
What concerning the efficiency? There too, Author has room to brag: the corporate’s Palmyra “ranked first in three necessary checks,” utilizing the Stanford HELM benchmarking system for AI fashions, “scoring 0.609 on Large Multitask Language Understanding (MMLU), 0.896 on BoolQ, and 0.413 on Pure Questions, outperforming fashions by OpenAI, Cohere, Anthropic, Google, Microsoft, and extra.
“The final mile of high quality is the toughest,” Habib mentioned within the launch.
Why are the fashions named Palmyra? The title comes from “the traditional metropolis situated in current day Syria, the place Queen Zenobia rebelled in opposition to Rome and established the Palmyrene Empire,” in line with Author’s weblog submit on the mannequin household, an evocation of feminine management that honors others equivalent to Habib.
Greater than basis fashions
Nevertheless it’s not simply the fashions that Author is providing: the corporate additionally gives a “Data Graph” that helps the fashions join with and reference an organization’s proprietary information (controllable by prospects), and a complete layer of functions together with chatbots, prebuilt templates for various enterprise options, and “composable,” person interface (UI) choices, which means, you don’t should be an AI professional to get arrange with Author’s LLM Palmyra.
The software program layers do the heavy lifting, and implementation is as straightforward as drag and drop in lots of instances, equivalent to with Author’s “Recaps” software that lets enterprise workers drag and drop audio recordings of conferences, webinars, calls — something — and robotically summarizes them with key factors.
That signifies that, along with competing with different basis mannequin suppliers for enterprises, Author can be fielding apps that compete with specialty software program suppliers like Gong and Otter.ai.
Safety and information privateness
The outcomes are spectacular, positive, however why ought to a potential enterprise buyer belief Author’s Palmyra over extra well-known LLMs equivalent to ChatGPT (which not too long ago launched an enterprise tier), Cohere, Llama 2, or others?
Along with having been educated on a “clear” and enterprise-focused dataset, Habib says that Palmyra additionally gives the correct mix of safety and privateness enterprises are in search of, proper out of the field.
“Author LLMs retains buyer information non-public and by no means makes use of it for mannequin coaching,” Habib wrote to VentureBeat. “Author additionally gives prospects the choice to self-host their very own LLM.”
That reality is necessary, coming after OpenAI CEO and co-founder Sam Altman posted on X (previously Twitter) earlier this 12 months that OpenAI “by no means trains on something ever submitted to the API or makes use of that information to enhance our fashions in any manner,” and advised CNBC anchor Andrew Ross Sorkin that the corporate had modified its Phrases of Service to make this express.
Palmyra can be Soc-2 Kind II, PCI and HIPAA (Well being Insurance coverage Portability and Accountability Act) licensed — similar to Anthropic’s Claude 2, which can be mentioned to be HIPAA licensed.
Glad enterprise prospects
Alongside the funding announcement, which Author plans to make use of to proceed increasing its choices, the corporate touted the advantages its purchasers are already realizing with its platform.
The lingerie/intimates retailer Adore Me reported a 40% improve in non-branded search engine optimisation site visitors utilizing Author’s Palmyra, whereas the robotic course of automation platform firm UiPath says it minimize its content material creation time in half.
Author says its Palmyra enterprise platform has confirmed its worth throughout a number of verticals, displaying a 10x income progress within the final two years and a internet income retention price over 150%.
With the brand new funding, Author is poised to speed up its mission of embedding generative AI into the enterprise panorama. As AI continues to democratize entry to superior instruments and applied sciences, Author’s full-stack strategy, centered on each flexibility and governance, makes it a compelling selection for enterprises trying to harness the total energy of AI.
VentureBeat’s mission is to be a digital city sq. for technical decision-makers to achieve data about transformative enterprise expertise and transact. Uncover our Briefings.
