How to manage “AI” development and R&D team in big corporations without “casting spells” and “doing rain dances”?

As my main work is related to AI and brains for robots, ( and I write in a blog about AI use cases in different industries, some big corporations involved in traditional industries ask me to help them solve problems with their AI research & development.

Usually, these problems are related to “over-budgeting” of R&D or issues with goal setting, or inefficient management in R&D teams.

Here are the most common issues and takeaways from my experience:

#1. “AI” is not magic and should be developed and delivered within time limits and budgets.

Almost all tasks (roughly 90%) performed in areas of Computer Vision and Machine Learning are not related to “real” R&D (creating algorithms or significant improvements).
For a successful implementation, it is usually enough to find a relevant, previously published paper and implement it well.
So there is no need to call “shamans” who will cast the spell and tell you about the “magic” of AI, but who won’t tell you when and how the problem can be solved. Hire only those specialists who can in a highly technical way describe the entire process involved in reaching the goal and who do not start reinventing the wheel instead of implementing a popular correct method.

#2. For the hardest R&D tasks in “AI” people are your primary asset, for other tasks rely on data.

When working on standard functions, you can hire average professionals, whose task will most likely be to create a well-marked and cheap dataset, implement a properly defined approach and build a system for correct data collection and subsequent algorithmic training. In this case, datasets and data itself, which was created and collected in the process of using the system (i.e., through its additional training, etc.) are the main assets of the business.

If the development accounts for 10% of your tasks and its implementation requires serious R&D, then your main asset is the people who implement the solution. Therefore, the majority of your and your competitors’ budgets will be spent not on implementation, but testing hypotheses. In this case, it is essential to know precisely which paths lead to results, and which paths do not.

Of course, when creating such projects, you need to reinforce your competitive advantage with an efficient logging system, data collection system, and a subsequent training system.

#3. The hardest part is the integrating of AI solutions into business processes. Developing a solution is the easiest part.

If a large corporation creates a well-functioning AI solution, they will soon be surprised to find out that the integration of the solution into their business process COSTS much more than the cost of creating an AI system.
Situations like this often happen due to the poor communication between those who are directly involved in the business process and those who innovate.
Encourage AI developers to get out of their lab and go into the “field” as soon as possible to interact directly with the expected users of the system!

#4. Most AI solutions are useless, because, internally, people don’t use them

AI system developers typically do not like to go out into “the real world.” For example, in agriculture, often when an AI system is created, it is implied that the AI developer used “STANDARD machine learning approaches on some pictures and got some result.”

Everything seems to be just fine – an AI system was successfully created after all. When the results from using a system like this get into the hands of an expert (for example, an agronomist), they happen to be absolutely useless to him.

Of course, the “let’s invite an AI shaman to create an AI system using standard machine learning methods” approach is cheap and can give useful results, but, typically, in my practice, the results are going to be useless. To get really valuable results, the AI system developers need to dive deep into the subject area and understand all the nuances. This way they can create neural networks, which as a RESULT will be able to compete with the existing neural networks of specialists in this specific field who have studied it as their major for 5-6 years and then practiced it for ten years.

#5. 50 to 70% of problems/tasks can be solved more efficiently without using “AI”

Nowadays, a lot of AI startups are trying to offer silver bullets to a variety of tasks that have been done effectively using well-known methods that do not require AI (simple statistical methods, for example).

Therefore, before developing any AI system, it is necessary to calculate the cost of its creation and OPERATION and to make sure that it is predictably more effective than currently existing solutions – if the system can only provide an improvement that is within a margin of error, then there is no point in creating it.

Instead of the conclusion

Overall, I would recommend large corporations to stop “worshiping” AI “shamans” and begin applying the same decision-making and management methods to AI projects as they do to others.
There is no “magic” in modern AI (CV / ML, Deep Learning); therefore AI projects should be carried out according to planned budgets and schedules: =)

Vitaliy Goncharuk

vactivity @