人工智能 垂直领域 行业_垂直与水平AI创业公司

人工智能 垂直领域 行业

AI Startups continue to emerge at a rapid pace. Since 2014, more than 15,798 AI Startups have received an investment of at least $400k, according to the Stanford AI Index. AI Startups have also attracted record investments of $26.6bn in 2019, as stated by CB Insights. Given the projected exponential growth of data, both investors and entrepreneurs long to have an impact through AI.

A I Startups继续Swift发展。 根据斯坦福大学人工智能指数(Stanford AI Index)的数据 ，自2014年以来，超过15,798个AI初创企业获得了至少40万美元的投资。 正如CB Insights所说，人工智能初创公司在2019年也吸引了创纪录的266亿美元投资。 鉴于预计的数据呈指数增长，投资者和企业家都渴望通过AI产生影响。

Yet, not all AI Startups are created equal. This article examines the differences and commonalities between Vertical and Horizontal AI Startups. The insights are based on discussions with AI experts from a Silicon Valley-based VC, a Berlin-based AI Venture Studio, and an AI Entrepreneur.

但是，并非所有的AI创业公司都是一样的。 本文研究了垂直和水平AI启动之间的区别和共性。 这些见解基于与来自硅谷的VC ，位于柏林的AI Venture Studio和AI企业家的AI专家的讨论。

To understand Vertical and Horizontal AI Products, let’s dive into the topic of AI product development first.

要了解垂直和水平AI产品，让我们首先深入探讨AI产品开发的主题。

开发AI产品 (Developing AI Products)

Every AI-based product follows roughly the same workflow. For each Machine Learning project, the team has to gather and prepare data, develop a model, and deploy and monitor its performance. The CRISP-DM model describes the Machine Learning workflow below.

每个基于AI的产品都遵循大致相同的工作流程。 对于每个机器学习项目，团队必须收集和准备数据，开发模型以及部署和监视其性能。 CRISP-DM模型在下面描述了机器学习工作流程。

Uni Dresden 德累斯顿大学

Understanding the Data Science workflow is important to distinguish between Vertical and Horizontal AI Startups. While Vertical AI Startups complete the entire workflow to deliver a product, Horizontal AI Startups focus on enabling the general Data Science workflow.

了解数据科学工作流程对于区分垂直和水平AI启动非常重要。 垂直AI初创公司完成交付产品的整个工作流程，而水平AI初创公司则专注于启用通用数据科学工作流。

垂直AI初创企业 (Vertical AI Startups)

A Vertical AI Startup delivers a product based on AI, owning the entire workflow to solve a very targeted customer need. These startups develop a complete product from end-to-end, starting with understanding the business case to monitoring product performance.

垂直AI初创公司提供基于AI的产品，拥有整个工作流程以解决非常有针对性的客户需求。 这些初创公司从端到端开发完整的产品，从了解业务案例到监视产品性能。

Vara, for example, is a Vertical AI Startup that identifies breast cancer on x-rays. Vara is the first company that has received federal approval to use its AI solution to assist radiologists. The company is a prime example of a Vertical AI Company that completes the full Data Science workflow to deliver an innovative AI-based product.

例如， Vara是一家垂直AI初创公司，可以通过X射线识别乳腺癌。 瓦拉(Vara)是第一家获得联邦批准使用其AI解决方案来协助放射科医生的公司。 该公司是Vertical AI Company的一个典型例子，该公司完成了完整的Data Science工作流程以交付基于AI的创新产品。

National Cancer Institute on 美国国家癌症研究所的 Unsplash 照片

AI is a powerful tool that helps companies achieve superhuman performance on specific tasks. AI opens up new possibilities to deliver new products, improve existing products, or automate tasks. Vertical AI Startups optimize their predictions towards a specific use-case. If they’re successful, they can perform better in this use-case than humans or competitors.

人工智能是一种强大的工具，可以帮助公司在特定任务上实现超人的绩效。 人工智能为交付新产品，改进现有产品或自动化任务开辟了新的可能性。 垂直AI初创公司针对特定用例优化其预测。 如果他们成功了，他们在这种用例中的表现将比人类或竞争对手更好。

Vertical AI Startups include domain experts in key positions such as product roles. For example, it is highly likely for a company that builds an automated cancer detection product to have in-house radiologists and/or oncologists. — Sahar Mor, Entrepreneur in stealth mode

垂直AI初创企业包括关键职位(例如产品角色)的领域专家。 例如，一家制造自动化癌症检测产品的公司很可能会拥有内部放射科医生和/或肿瘤科医生。 —隐身模式下的企业家Sahar Mor

However, it is harder for Vertical AI Startups to adjust their product to serve adjacent use-cases. For Vara, which excels at identifying breast cancer, it is not trivial to switch the task to detect brain cancer. AI models perform well in their narrow tasks. Vara could train a new AI model to detect brain cancer from a new dataset, but still reuse their existing model development pipeline.

但是，垂直AI初创公司很难调整其产品以适合相邻的用例。 对于擅长识别乳腺癌的瓦拉(Vara)来说，切换检测脑癌的任务并非易事。 AI模型在狭窄的任务中表现良好。 Vara可以训练新的AI模型来从新数据集中检测脑癌，但仍可以重用其现有的模型开发流程。

InRasmus Rothe, Co-Founder and CTO at Merantix

Merantix的联合创始人兼首席技术官Rasmus Rothe

Additional examples of Vertical AI Startups are driverless car company Waymo and financial crime prevention company hawk.ai. Altogether, for Vertical AI Startups it is of paramount importance to focus on solving a single use-case that is large enough to ensure growth. As it takes time and data to train a new model for a new use-case, the initial use-case has to be impactful enough to support a profitable company.

垂直AI初创公司的其他示例是无人驾驶汽车公司Waymo和金融犯罪预防公司hawk.ai。 总而言之，对于垂直AI初创公司而言，最重要的是专注于解决一个足以确保增长的单个用例。 由于为新用例训练新模型需要时间和数据，因此初始用例必须具有足够的影响力以支持盈利的公司。

Let’s evaluate the specifics of Horizontal AI Startups next.

接下来，让我们评估“水平AI启动”的细节。

横向AI公司 (Horizontal AI company)

A Horizontal AI company enables Data Scientists to complete the Data Science workflow more efficiently. Thus, the customers of Horizontal AI Startups are generally Data Scientists or development teams.

水平AI公司使数据科学家能够更高效地完成数据科学工作流程。 因此，水平AI初创企业的客户通常是数据科学家或开发团队。

Horizontal AI Companies pick specific or multiple parts of the Data Science workflow and improve them. A Horizontal AI Company could for example focus on helping Data Scientists monitor their models. Horizontal AI Products are use-case agnostic and are used to develop any AI-based product.

横向AI公司从数据科学工作流程中选择特定或多个部分并加以改进。 例如，水平AI公司可以专注于帮助数据科学家监控其模型。 水平AI产品与用例无关，可用于开发任何基于AI的产品。

Saturn Cloud is a Horizontal AI Startup that aims to make Data Scientists more effective. It provides a fully managed service that enables data scientists to work at scale using the tools they know best: Python, Jupyter, and Dask. The tool is generic so that it can be used for any AI-based product.

土星云是一家水平AI创业公司，旨在提高数据科学家的效率。 它提供了一项完全托管的服务，使数据科学家可以使用他们最熟悉的工具(Python，Jupyter和Dask)进行大规模工作。 该工具是通用工具，因此可以用于任何基于AI的产品。

“Across dozens of industries, we are hearing how automating data science workflows is enabling more efficient teams and better experiences for their end customers. At the same time, Python is soaring in use and rapidly dominating the data science landscape — so adopting a suite of Python-native tools for analytics and automation makes sense for most data teams.” — Hugo Shi, co-founder of Saturn Cloud and previously founder of Anaconda.

“在数十个行业中，我们听到了自动化的数据科学工作流程如何为最终用户带来更高效的团队和更好的体验。 同时，Python的使用量猛增，并Swift占领了数据科学领域-因此，对于大多数数据团队而言，采用一套Python原生工具进行分析和自动化是有意义的。” — 土星云 ( Saturn Cloud)联合创始人，此前Anaconda的创始人Hugo Shi。

Horizontal AI Solutions promise to become more useful in the future, as the amount of data created grows exponentially, according to an IDC forecast. Data is the fuel for AI-based products, thus the more data available, the higher the potential to create valuable data-based products. Additionally, the market for Data Scientists continues to grow steadily. According to a study by IBM, the demand for Data Scientists will increase by 28% to 2.7m in the USA in 2020. Thus, the pool of target customers continues to grow.

根据IDC的预测 ，随着创建的数据量呈指数增长，Horizo​​ntal AI Solutions有望在未来变得更加有用。 数据是基于AI的产品的推动力，因此可用数据越多，创建有价值的基于数据的产品的潜力就越大。 此外，数据科学家市场继续稳定增长。 根据IBM的一项研究 ，到2020年，美国对数据科学家的需求将增长28％，达到270万。因此，目标客户群继续增长。

IDC IDC

However, Horizontal AI products also face challenges. They have to seamlessly integrate into the target Data Science workflow and software stack. Additionally, “it is important to strive for market leadership when developing a Horizontal AI Product”, says Rasmus Rothe. For Horizontal AI Startups, it is hard to collect proprietary data and thus to build data defensibility. When defining which step of the Data Science workflow to automate, it is important to keep different requirements for different AI-based models in mind. Focusing on the most valuable features for the target group will set aspiring from successful Horizontal AI Startups apart.

但是，水平AI产品也面临挑战。 他们必须无缝集成到目标数据科学工作流程和软件堆栈中。 另外，“在开发水平AI产品时，争取市场领导地位很重要”，Rasmus Rothe说。 对于水平AI初创公司，很难收集专有数据并因此建立数据防御能力。 在定义要自动化的数据科学工作流程的哪一步时，务必牢记针对基于AI的不同模型的不同要求。 专注于目标群体最有价值的功能，将使成功的卧式AI初创公司脱颖而出。

While Horizontal AI companies often have more competition, their product is generally more acquirable because its tech can fit into the stack of a broader set of acquirers. — Jason Risch

虽然水平AI公司通常具有更多的竞争优势，但他们的产品通常更易于购买，因为其技术可以融入到更多收购方的堆栈中。 — 杰森·里施 ( Jason Risch)

Additional examples of Horizontal AI Startups are the data labeling company understand.ai, and Data Science workflow management startup neptune.ai. The potential for Horizontal AI Startups is vast. Yet, it is challenging to build a tool that outperforms competitors and reaches market leadership. Once you provide value to many Data Scientists, your Horizontal AI Startup will be successful.

水平AI初创公司的其他示例是数据标签公司谅解 .ai和数据科学工作流管理初创公司neptune.ai 。 水平AI创业的潜力是巨大的。 然而，构建一种超越竞争对手并达到市场领导地位的工具具有挑战性。 一旦您为许多数据科学家提供了价值，您的水平AI启动就会成功。

垂直与水平AI启动 (Vertical vs. Horizontal AI Startups)

AI Startups are different, and understanding their unique value propositions is central to building a successful AI Company. While Vertical AI Startups deliver products, Horizontal AI Startups empower AI product creation. T-shaped mixes of both startups types exist as well, but are subject to another article.

人工智能初创企业是不同的，了解其独特的价值主张对于建立成功的人工智能公司至关重要。 垂直AI初创公司交付产品的同时，水平AI初创公司则支持AI产品的创建。 两种初创企业类型的T形混合也都存在，但另一篇文章对此有所介绍。

It is important to be aware of the type of AI startup you’re building because that will ultimately impact your product roadmap.

重要的是要知道您要构建的AI初创公司的类型，因为这最终会影响您的产品路线图。

• Vertical AI startups need to focus on solving a single use-case that is large enough to sustain a growing company
  垂直AI初创企业需要专注于解决一个足以维持一家成长型公司的用例
• Horizontal AI startups need to reach market leadership in their chosen niche to provide value for many Data Scientists
  横向AI初创企业需要在其选择的市场中达到市场领导地位，才能为许多数据科学家提供价值

Thank you for your attention! After finishing this post, you have a better understanding of the characteristics of Vertical and Horizontal AI Startups. If you’d like to hear more about the intersection of Business, Project Management, and Data Science, follow me on Medium, LinkedIn, or Twitter.

感谢您的关注！ 完成本文后，您将更好地了解垂直和水平AI创业公司的特征。 如果您想了解有关业务，项目管理和数据科学的交叉学科的更多信息，请在Medium ， LinkedIn或Twitter上关注我。

翻译自: https://towardsdatascience.com/vertical-vs-horizontal-ai-startups-e2bdec23aa16

人工智能 垂直领域 行业