AI Agency Language I/O Closes USD 5m Series A; Expands Focus Beyond Chat, Email

AI代理语言I/O关闭500万美元A系列;将焦点扩展到聊天,电子邮件之外

2021-03-24 01:25 slator

本文共697个字,阅读需7分钟

阅读模式 切换至中文

US-based AI agency Language I/O announced, on March 23, 2021, that it had closed a USD 5m funding round. Led by Massachusetts venture capitalists Bob Davoli (Gutbrain Ventures) and Bruce Clarke (PBJ Capital), the round also drew previous investor Golden Seeds, and new VCs such as Omega Venture Partners, Michael Wilens, Tom Axbey, and Eric Schnadig. Language I/O CEO, Heather Morgan Shoemaker, said the Series A, which closed on February 5, 2021, brought total funds raised to USD 5.62m. She declined to disclose the valuation. Shoemaker told Slator the funds will go toward expanding the team (“We are already hiring data scientists / ML engineers to build out our existing machine learning platform”) and marketing and sales activities. The CEO added that she and Chief Business Officer, Kaarina Kvaavik, “went from jointly owning over 95% of the company to jointly owning just over 50% of the company. But we maintain majority ownership for now.” Investors Bob Davoli and Tom Axbey will join Shoemaker and Kvaavik on the Language I/O board, which will elect a chairperson in April. Language I/O (LIO) has 22 full-time employees, many located in Wyoming where the company is headquartered. They also have a large team in the Boston area, a few FTEs located in California, and a couple outside the US. The data scientists they are in the process of recruiting will work on what Shomaker called “our self-improving glossary, [which] automatically detects terms in company chats / emails that require company-specific translation.” She said that while LIO is currently focused on customer support — that is, “enabling companies to use their monolingual customer support agents to chat and email with customers in any language” — the new funds will be used for “expanding our core tech beyond customer support.” According to the CEO, Language I/O is an AI agency in every sense — which, based on Tomasz Tunguz’s investment thesis as shared with the SlatorCon San Francisco 2019 audience and a 2020 Forbes article cited by Shoemaker — relies on AI to scale rather humans, replacing some jobs with software. What LIO does is basically to customize machine translation engines for companies by homing in on so-called “problematic segments.” Shoemaker explained: “Today, to provide NMT [neural machine translation] that generates accurate translation for a specific business, most of the language industry trains a single model for each language pair. This requires a large corpus of human-translated content and linguistic expertise to train and iterate on the model. “We determined early on that this was not a scalable solution. Instead, we integrate with numerous NMT engines, provided that they are GDPR compliant, etc. — and we don’t train them; we hit the general engine and impose a company’s preferred translations of problematic segments atop the general translation.” The CEO further said, “Our ability to employ AI to automatically detect problematic segments, generate the preferred translations for those segments, and impose them atop the general engine is unique and lies at the core of our solution. It’s also what throws us into the category of AI agency as we are employing AI to do the bulk of this work, not humans.” LIO works with single-language vendors (SLVs) and directly hires freelance linguists for certain languages they “chose to staff up ourselves,” Shoemaker told Slator. “All said, there are hundreds in our network for the aspects of our work that require human translation, such as the MTPE [a.k.a. PEMT] of knowledge-base content.” She said they do not focus on any specific market segment but maintain a diverse client portfolio (e.g., travel, diesel engines, gaming, robotics, consumer electronics, online gambling). Asked to describe LIO’s pricing model, Shoemaker said it is a tiered subscription with a cap on the number of words that a subscriber can translate through the Language I/O platform on an annual or monthly basis. The CEO said they saw “a sharp increase in usage during Covid as companies looked for a technology alternative to staffing up operationally” — to the extent that the company’s monthly recurring revenue doubled from August 2020 to today. Image: Language I/O Co-founders Kaarina Kvaavik (L), CBO, and Heather Morgan Shoemaker, CEO
总部位于美国的人工智能机构Language I/O于2021年3月23号宣布,它已经完成了一轮500万美元的融资。在马萨诸塞州风险投资家Bob Davoli(Gutbrain Ventures)和Bruce Clarke(PBJ Capital)的带领下,这轮融资还吸引了之前的投资者金种子,以及新的风险投资公司,如Omega venture Partners,Michael Wilens,Tom Axbey和Eric Schnadig。 语言I/O首席执行官希瑟·摩根·舒梅克说,2021年2月5日结束的A系列筹资总额达到562万美元。她拒绝透露估值。 Shoemaker告诉Slator,这些资金将用于扩大团队(“我们已经在雇佣数据科学家/ML工程师来构建我们现有的机器学习平台”),以及市场和销售活动。 这位首席执行官补充说,她和首席商务官卡丽娜·克瓦维克“从共同拥有公司95%以上的股份,到共同拥有公司50%多一点的股份。但我们目前仍保持多数股权,“ 投资者Bob Davoli和Tom Axbey将加入Shoemaker和Kvaavik的语言I/O董事会,该董事会将于4月选举主席。 语言I/O(LIO)拥有22名全职员工,其中许多人位于公司总部所在的怀俄明州。他们在波士顿地区也有一个庞大的团队,在加利福尼亚有几个FTEs,在美国以外也有几个。 他们正在招聘的数据科学家将致力于Shomaker所称的“我们的自我完善词汇表”,[它]自动检测公司聊天/电子邮件中需要公司特定翻译的术语。 她说,虽然LIO目前专注于客户支持--即“使公司能够使用他们只懂一种语言的客户支持代理以任何语言与客户聊天和电子邮件”--但新的资金将用于“将我们的核心技术扩展到客户支持之外”。 据这位CEO称,Language I/O在任何意义上都是一家人工智能机构--基于托马斯•通古斯在2019年旧金山SlatorCon论坛上分享的投资理论,以及苏梅克引用的2020年福布斯文章--该机构依靠人工智能而非人类来提升规模,用软件取代了一些工作。 LIO所做的基本上是通过专注于所谓的“有问题的片段”来为公司定制机器翻译引擎。 Shoemaker解释说:“今天,要提供为特定业务生成准确翻译的NMT[神经机器翻译],大多数语言行业都为每种语言对训练单一模型。这需要大量的人工翻译内容和语言专业知识来训练和迭代模型。 “我们很早就确定这不是一个可扩展的解决方案。取而代之的是,我们集成了众多的NMT引擎,前提是它们是符合GDPR的,等等--而且我们不训练它们;我们使用通用引擎,将公司对有问题片段的首选翻译强加在通用翻译之上,“ 这位首席执行官还说:“我们利用人工智能来自动检测有问题的片段,为这些片段生成首选的翻译,并将它们强加在通用引擎之上的能力是独一无二的,也是我们解决方案的核心。这也是我们被归入人工智能代理范畴的原因,因为我们使用人工智能来完成大部分工作,而不是人类。“ LIO与单一语言供应商(SLVs)合作,并直接为他们“自己选择的”语言雇佣自由语言学家,Shoemaker告诉Slator。“总之,在我们的网络中,有数百个工作方面需要人工翻译,例如知识库内容的MTPE(又称PEMT)。” 她说,他们并不专注于任何特定的细分市场,而是保持着多样化的客户组合(例如,旅游,柴油发动机,博彩,机器人,消费电子产品,在线博彩)。 当被问及LIO的定价模式时,Shoemaker表示,这是一种分层订阅模式,订阅者每年或每月通过语言I/O平台翻译的字数有上限。 这位首席执行官表示,他们看到“在Covid期间,随着公司寻找替代人员配置的技术,Covid的使用量急剧增加”--从2020年8月到今天,公司的每月经常性收入翻了一番。 图片:Language I/O联合创始人CBO Kaarina Kvaavik(左)和CEO Heather Morgan Shoemaker

以上中文文本为机器翻译,存在不同程度偏差和错误,请理解并参考英文原文阅读。

阅读原文