Artificial Intelligence Development Management - Best Practices
Best Practices in Managing Artificial Intelligence Software Development Units
By Peter Ludzen
There are many companies now engaged in the work of developing artificial intelligence. What is it like to be a manager in one of those companies? We interviewed a developer at a top development company (yes, it's probably the one you're thinking!) to understand more about ai development best practices.
But, first, let's look at what companies are developing ai (source): Deepmind, Google, Facebook, OpenAI, Baidu, Microsoft, Apple, & IBM - Watson. Your company not here? Comment or email us at info @ javarosa.org to include it. If you need to back up a step to learn what AI is, watch this video.
Here's the short Q&A:
- What kind of ai projects are you working with? To avoid leaking the name of my company and exactly what we're doing, I'll just say, we're working on state of the art ai virtual assistant.
- What are the obstacles to universal adoption of the tech you're working on? The virtual assistants on the market today are still in the nascent phase, and natural language processing still lacks a sense a true conversational ability, and is clumsy at best. Companies like Watson are making it possible to build a chatbot with Watson, but it's a bit desperate because over a dozen platforms exist if you want to know how to build a chatbot. While chatbots are not VAs, this trend shows how saturated the field has become. It's a race, but I think another company will emerge with a tech that jumps beyond what we're all working with.
- What kind of management style is in place to manage the dev team? We're actually lucky to have our AI team managed by a seasoned ai dev. We all respect him, and he doesn't get bulldozed by our CTO who is more conventional. I think it could be better though, because there is still a lot of petty politics.
- How could the management be better? AI management should focus more on research, and less on making gains to satisfy quarterly reports. That way we're only making short term gains without experiencing jumps in innovation necessary to dominate in the area of AI-VA. So, while we offer a widely adopted VA, we can't pioneer without sending the entire process back to research. There should be a hybrid, where we can toggle between research and commercial application.
- What is the future of management in ai software development? We actually use a lot of dev management methods inspired from agile, and run sprints, etc. However, at the end of the day, it's just a bunch of meetings where people don't speak up and ask for help, and everything still gets delayed. So, even with the great management methods at work, we're still human. I read an article about AI actually replacing managers in Psychology Today (https://www.psychologytoday.com/blog/wired-success/201611/why-artificial-intelligence-will-replace-managers) . Obviously we have to get beyond Hawking's concerns about letting AI have that much control, but I could see where we could ascend into the stratosphere of AI dev if we could get beyond the petty politics of being managed, and get down to doing some real work.