An Intellyx BrainCandy Brief
Alegion knows labels may be problematic when applied to people, but supremely useful when applied to data. The ability to understand how and why humans rapidly categorize audiovisual and language data, and the cognitive bias they bring to the process, is beneficial for training AI and machine learning scenarios.
Even if computers can process real-world signals millions of times faster than humans, they still need help with judging and interpreting what humans infer from that data. This is why Alegion has built up an audience of both human and computer responses to images, sounds and text that seeks to break down how sensory data is categorized along several dimensions.
We tend to think of this sort of innovation as being applicable to ‘reading faces’ for social apps and situational awareness for self-driving cars. This tech could help there, but why stop with areas that are already rife with captures of billions of specific industry datapoints?
What if you wanted AI-based software to learn if crops are ready for picking, or if cells look cancerous, or if a baby bear is likely to have a mother nearby, or if the whirring sound of an engine requires a part to be replaced?
Audiovisual interpretations are not simply answered, so Alegion applies a sliding scale to apply more of the wisdom of experts or crowds when data objects seem unfamiliar to their systems, and less human intervention when the system is contextually familiar with the input.
Training of machine learning systems is never 100% automated, but if you want to get to the “ground truth” of that data, human intervention shouldn’t be left out. The result is prepared training data that could feed AI/ML applications we (or, ‘they’) haven’t thought of yet.
©2019 Intellyx LLC. Intellyx advises companies on their digital transformation initiatives, and helps vendors communicate their agility stories. At the time of this writing, none of the organizations mentioned in this article are Intellyx customers. To be considered for a Brain Candy article, email us at pr@intellyx.com.
