Как выбрать пилотный ИИ проект

Автор: Andrew Ng

Источник: HBR

Artificial intelligence (AI) is poised to transform every industry, just as electricity did 100 years ago. It will create $13 trillion of GDP growth by 2030, according to McKinsey, most of which will be in non-internet sectors including manufacturing, agriculture, energy, logistics, and education. The rise of AI presents an opportunity for executives in every industry to differentiate and defend their businesses. But implementing a company-wide AI strategy is challenging, especially for legacy enterprises.

My advice for executives, in any industry, is to start small. The first step to building an AI strategy, drawn from the AI Transformation Playbook, is to choose one to two company-level pilot AI projects. These projects will help your company gain momentum and gain firsthand knowledge of what it takes to build an AI product.

5 traits of a strong AI pilot project

Tapping the power of AI technologies requires customizing them to your business context. The purpose of your one or two pilot projects is only partly to create value; more importantly, the success of these first projects will help convince stakeholders to invest in building up your company’s AI capabilities.

When you’re considering a pilot AI project, ask yourself the following questions:

Does the project give you a quick win? Use your first AI pilot project to get the flywheel turning as soon as possible. Choose initial projects that can be done quickly (ideally within 6-12 months) and have a high chance of success. Instead of doing only one pilot project, choose two to three to increase the odds of creating at least one significant success.

Is the project either too trivial or too unwieldy in size? Your pilot project does not have to be the most valuable AI application as long as it delivers a quick win. But it should be meaningful enough so that a success convinces other company leaders to invest in further AI projects.

In the early days of leading the Google Brain team, I faced widespread skepticism within Google about the potential of deep learning. Speech recognition was much less important to Google than web search and advertising, so I had my team take on speech as our first internal customer. By helping the speech team build a much more accurate recognition system, we convinced other teams to have faith in Google Brain. For our second project, we worked with Google Maps to increase data quality. Each successful project increased the momentum in the flywheel, and Google Brain played a leading role in turning Google into the great AI company it is today.

Is your project specific to your industry? By choosing a company-specific project, your internal stakeholders can directly understand the value. For example, if you run a medical devices company, building an AI+Recruiting project to automatically screen resumes is a bad idea for two reasons: (1) There’s a high chance someone else will build an AI+Recruiting platform that serves a much larger user base and will outperform what you could do in-house and/or undercut you; (2) This project is less likely to convince the rest of your company that AI is worth investing in than if your pilot project applied AI to medical devices. It is more valuable to build a healthcare-specific AI system — anything ranging from using AI to assist doctors with crafting treatment plans, to streamlining the hospital check-in process through automation, to offering personalized health advice.

Are you accelerating your pilot project with credible partners? If you are still building up your AI team, consider working with external partners to bring in AI expertise quickly. Eventually, you will want to have your own in-house AI team; however, waiting to build a team before executing might be too slow relative to the pace of AI’s rise.

Is your project creating value? Most AI projects create AI value in one of three ways: reducing costs (automation creates opportunities for cost reduction in almost every industry), increasing revenue (recommendation and prediction systems increase sales and efficiency), or launching new lines of business (AI enables new projects that were not possible before).

You can create value even without having “big data,” which is often overhyped. Some businesses, such as web search, have a long tail of queries, and so search engines with more data do perform better. However, not all businesses have this amount of data, and it may be possible to build a valuable AI system with perhaps as few as 100-1000 data records (though more does not hurt). Do not choose projects just because you have a lot of data in industry X and believe the AI team will figure out how to turn this data into value. Projects like this tend to fail. It is important to develop a thesis upfront about how specifically an AI system will create value.

Setting up your AI project for success

So what do these traits look like in practice?

AI is automation on steroids. A rich source of ideas for AI projects will lie in automating tasks that humans are doing today, using a technology called supervised learning. You will find that AI is good at automating tasks, rather than jobs. Try to identify the specific tasks that people are doing, and examine if any can be automated. For example, the tasks involved in a radiologist’s job may include reading x-rays, operating imaging machinery, consulting with colleagues, and surgical planning. Rather than trying to automate their entire job, consider if just one of the tasks could be automated or made faster through partial automation.

Before executing on an AI pilot, I recommend clearly stating the desired timeline and outcome, and allocating a reasonable budget to the team.

Appoint a leader: Choose someone who can work cross functionally, and bridge both AI and your industry’s domain experts. This will make sure that when the project succeeds, it will influence the rest of the organization. Their goal is not to build an AI startup. Their goal is to build a successful project that will influence the rest of the company’s beliefs and state of knowledge about AI as a first step toward building future projects.

Conduct business value and technical diligence: Make sure that, if executed successfully, the business leaders agree that this project will create sufficient value for the business. But also make sure that the project is feasible. Technical diligence can take a small number of weeks, requiring a technical team to examine what data you have and perhaps even carry out small-scale experiments.

Build a small team: I have seen numerous pilot ideas that were executed with about five to 15 people. The exact level of resources varies wildly per project, but scoping projects that can be done with a small team ensures that everyone can know everyone else and work cross-functionally, and perhaps also makes the allocation of resources more painless. While there are some projects today that require 100+ (or even 1000+) engineers to do well, such a high level of resourcing is likely not necessary for your pilot AI project.

Communicate: When the pilot project hits key milestones, and especially when it delivers a successful result, be sure to give the team an internal platform — ranging from talks, to awards, to even external PR — to allow their work to become known inside the company. Making sure the project team is recognized by the CEO and is visibly successful will be a key part to building momentum. If you have an AI technology team working with a business team, make sure also that the business team receives plenty of credit and rewards for the success. This will encourage other business teams to jump into AI as well.

Having led Google Brain and Baidu’s AI Group, which were respectively leading forces for turning Google and Baidu into great AI companies, I think most companies can and should become good at AI. Your goal should not be to compete with the leading internet companies, but rather to master AI for your vertical industry sector. And remember: the first step is to select the right pilot projects and execute on them.

Источник

Data Scientist # 1

Машинное обучение, большие данные, наука о данных, анализ данных, цифровой маркетинг, искусственный интеллект, нейронные сети, глубокое обучение, data science, data scientist, machine learning, artificial intelligence, big data, deep learning

Данные — новый актив!

Эффективно управлять можно только тем, что можно измерить.
Copyright © 2016-2021 Data Scientist. Все права защищены.