TechCrunch is reporting that Apple has acquired AI startup Lattice Data for $200 million. Lattice claims to focus on transforming unstructured data–what they refer to as dark data–into structured data. Unstructured data is often found in text, images, video and other media and is growing rapidly in the digital age. Some estimates going back two decades estimate that 70-90% of all data is unstructured.
Machines typically do not have the ability to make sense of unstructured data unless it is tagged, organized by humans, or run through human-built machine models that compare it to a library of known things. The idea is that by structuring data, machines can then process it and provide value more easily to humans or other computers. Lattice Data comes out of a Stanford project called Deep Dive which focused on automating this process by using statistical inference. A short overview of Deep Dive by MacArthur Genius Grant Recipient and Stanford Professor Christopher Re says this:
[Deep Dive] takes the position that for some tasks, machine-based solutions can be more efficient, more accurate, and have far greater coverage than human annotators. To accomplish this, DeepDive has a radically different design than a traditional database system: probabilistic inference is its core operation.
A Lot of AI is Focused on Interpreting Unstructured Data
Lattice Data may have amazing technology to justify the $200 million price tag. However, keep in mind that a lot of AI you hear about deals with unstructured data. Driverless cars review image data to determine whether to stop at a red light or avoid a collision with a pedestrian. Voice assistants take unstructured data such as speech or text apply AI-driven Natural Language Processing and Natural Language Understanding to determine meaning and intent. The big data craze was initially about managing large quantities of structured data. The AI movement is often about having machines solve problems hidden inside unstructured data. This is what everyone is working on and there are many approaches being used.
Will Lattice Data Help Apple Siri or Just Protect Apple in Patent Fights?
One thing we know about Lattice Data is that it has a prominent team behind it and TechCrunch reports that 20 engineers are now working at Apple. The competition for AI-based voice assistants is fierce and pits Apple against Amazon, Google, Microsoft and Samsung. Apple’s Siri seems lacking in performance today compared to voice assistant offerings from these rivals. Despite recent patent applications, Apple’s Siri is less capable today than Alexa, Google Assistant, Cortana and Bixby respectively. Improving Siri to better understand conversational meaning and intent is critical for the company to keep pace.
The other angle to keep in mind is that Apple may not have immediate need for Lattice Data technology but is merely amassing intellectual property to protect itself from the inevitable patent fights down the road. Apple has emerged from a decade of litigation with Google, Samsung and others and understands the value of having a strong patent portfolio.