Twitter's latest acquisition uses AI to control the setting for other AI algorithms. Source: Whetlab.

Last week, Twitter (NYSE:TWTR) announced the acquisition of Whetlab -- a machine learning start-up run by five super-smart dudes. When they're not using Whetlab's technology to discover the next great beer recipe -- seriously -- they use it to develop better AI algorithms for each company they work with.

Whetlab calls it "AI for AI," which means it's sort of meta-AI that can replace a roomful of Ph.D.s developing machine learning algorithms for companies like Twitter and Facebook (NASDAQ:FB). And Facebook literally has dozens of people working on AI.

Twitter said it will use Whetlab's technology to accelerate its own machine learning research, which it uses to surface more relevant content. The first example of Twitter's efforts in that department are the "While You Were Away ..." tweets that started popping up early in 2015. But Twitter is going to take a more algorithmic approach to curating content later this year with Project Lightning, and machine learning algorithms can be applied to numerous other aspects of its products as well.

Content first
From the way Wall Street reacted to Twitter's latest earnings results, you'd think the company was in crisis. While the revenue miss was significant last quarter and user growth continues to slow, it's not like the company is really struggling. Revenue still grew 74% year over year, and Twitter has over 300 million active users and claims 500 million unique visitors to the site each month who don't log in.

Still, those 500 million logged-out visitors, along with the billion or so people who signed up for Twitter and subsequently abandoned the social network, represent a significant opportunity the company has largely been unable to exploit. Chief Financial Officer Anthony Noto said attracting those users to the platform requires the company to emphasize consumption first. To do that, it needs to display its best content to anyone visiting the website.

That's an application perfect for machine learning. Based on users' initial reactions to content and their level of engagement, Twitter can show certain tweets to other users through the new logged-out experience and While You Were Away. Project Lightning, which provides curated media-rich timelines for live events, will be heavily reliant on algorithms to surface quality tweets for the human curators to inspect.

All of these efforts are intended to create a more engaging Twitter in which users can easily discover interesting content. Twitter boasts about the amount of content on its platform, but finding that content isn't particularly intuitive or easy. Correcting that with more curated content will enable Twitter to retain more users, which will lead to better user growth in the long run. And more users generate more ad revenue.

Targeting ads should be a close second
Twitter's shortcomings last quarter fell on the company's direct-response advertisements, which rely on conversions for revenue. The shortfall indicates the company's ads aren't converting as well as Twitter expected, which means it's not doing a great job targeting its users. To that end, Twitter announced the acquisition of TellApart to help with cross-device retargeting.

Machine learning could help improve targeting in several ways. First, it can help Twitter target certain kinds of advertisements to certain users based on their previous responses to ads. Second, it can better understand what users are tweeting about through language processing and object recognition algorithms, which is one of the top applications of machine learning today. Twitter is likely already working on AI for understanding its users better and improving targeting, but Whetlab and its employees might accelerate those efforts.

Competing with the big boys
Facebook's research and development budget is more than four times as large as Twitter's, and Twitter is already spending almost half of its revenue on research. In order to compete, Twitter must be selective about how it spends its money.

It can't go off funding AI algorithms that produce tiny realistic images of planes and animals regardless of what long-term applications they might serve. It needs more focus on things that can have a measurable impact on its business, either through more users or more ad revenue -- preferably both. The acquisition of Whetlab, and the subsequent shutdown of its product so no competitors can access its technology, is a smart choice to improve its AI team.