Ring’s New AI Search Feature Allows Easy Video Scanning—Results Vary
The home-surveillance company owned by Amazon will provide users the option to search video for particular items and activities.
LIZ HAMREN, THE CEO of Amazon’s Ring camera division, frequently thinks about whether her husband took the package of frozen products that is consistently sent to their residence. She could bother him, or she could review the recordings from their Ring.
In the past, this alternative forced her to painstakingly scroll through a video timeline in Ring’s app to determine if and when the box was collected and by whom. However, in the past few weeks, Hamren has managed to allow AI to handle the searching. She enters “package today” into the app and immediately sees the footage of her husband fulfilling his obligation, presuming he has.
Ring’s recently announced Smart Video Search feature will be available to all Home Pro subscribers, who pay $20 a month or $200 a year, starting on November 5. The feature was evaluated by the press in recent days. It didn’t provide an answer to the frequent question in our household: “Did you remember to secure the door?” However, queries for “dog defecating,” “landscapers,” and “roller skating” yielded primarily correct outcomes.
Hamren, who manages three additional Amazon home tech divisions, states that Ring prevents searches for content deemed “offensive, inappropriate, or harmful.” She refused to provide more details beyond mentioning that names of weapons and “descriptive terms about individuals or circumstances” are included in those prohibited.
Our inquiries for terms that some might view as sensitive, such as “Black,” “turban,” and “gun,” produced no results even though there was footage that corresponded to those descriptions. “Disabled” saw individuals rolling by in wheelchairs—and moms maneuvering strollers. Ring announces it will now prevent searches for “disabled” because of potential abuse.
Nonetheless, the search feature represents a significant enhancement for Ring, which has gained popularity in home surveillance over the last ten years with its video doorbells and an expanding array of security devices. The problem has been that countless Ring users are left with an extensive archive of recordings—possibly as much as six months’ worth—and few tools to easily sift through the data. Hamren, a past hardware executive at Microsoft and Meta who started in March 2023, aimed to provide users with insights without requiring them to tinker excessively. That was made possible by AI.
Hamren’s time at Ring has also resulted in a notable change in the company’s approach to prospective clients. “Tough-on-crime” advertisements featuring dubious individuals vanished. She scaled back collaborations with police, which had allowed Ring to create tools for users to effortlessly share videos with law enforcement without needing a warrant.
Ring’s current marketing showcases charming clips with pets and kids, and users are encouraged to post playful videos to their online communities, rather than just the serious ones. The emphasis has shifted from combating crime to alleviating anxiety by providing customers detailed visibility into events occurring in front of the camera.
Searching is merely the beginning. According to Hamren, the advanced AI models driving search will ultimately allow Ring to alert users about particular occurrences, like a bear roaming in the backyard or unusual puddling on the patio. They will also get recaps of recent events. “Ultimately, our goal is to equip you with intelligence and context regarding current events,” she states. Houses are merely one target on her radar. In the initial half of 2025, Ring intends to introduce a subscription plan tailored for small and medium-sized enterprises. Hamren states that subscribers will have improved capabilities to monitor various sites and provide employees access to specific recordings.
Search Experiments
Hamren states that Smart Video Search has been under development for more than a year. The AI system underwent training with a method known as contrastive learning to identify packages, individuals, vehicles, animals, weather conditions, and activities like running. “She states, ‘The issues that truly matter to people.'” According to Ring, the company conducts training with public videos as well as those provided by employees and their acquaintances. Rivals such as Nest and Wyze are exploring or trialing comparable search features.
Ring is placing search features within its highest-priced subscription as it demands considerable cloud computing resources to operate, and the firm, which achieved profitability last year, aims to maintain this status, according to Hamren. She states that concentrating on what users are willing to pay creates a “very clear relationship.”
Users enrolled in Ring’s Home Pro plan can participate in the search feature and select which of their cameras to incorporate—it is compatible with all models. However, search functionality is ineffective when users enable end-to-end encryption, providing greater privacy for their videos. Hamren mentions that Ring is striving to eliminate that restriction, but she didn’t provide a specific timeline. According to Ring, searches are not recorded in users’ accounts, but anonymized search queries might be utilized to enhance the feature.
Ring’s search operates similarly to looking for a video on YouTube. Outcomes can be sorted by importance or chronological order. The system can identify categories of vehicles (such as “police” or “minivan”), but don’t anticipate it to distinguish a Kia from a Mazda or recognize a particular license plate. “Right now, our focus isn’t on reading license plates,” says Hamren. The company does not intend to provide facial recognition. Despite being more widely used by governments and businesses, facial recognition still raises worries regarding mass surveillance and the risk of misidentification. Hamren states that Ring would rethink its stance depending on the user demand for the technology.
In trials, Ring’s search performed reasonably well on queries such as “backpack,” “scarf,” “speeding cars,” “ice cream truck,” and “shopping cart.”
However, for numerous searches, the results contained clear misidentifications. The software understood headlights appearing from turning vehicles as “fireworks.” “Blonde-haired women” provided individuals matching the description as well as golden retrievers. To the AI, individuals with a hand close to their face appeared to be “smoking.” Hamren anticipates that feedback will result in revisions.
Other kinks were stranger. “Purple hair” featured an individual in a purple sweater and a bus adorned with purple paint. Ring couldn’t recognize a decorative pumpkin but linked it to a landscaper in an orange vest who was gathering leaves into a white bag. “Rollerblades” accurately identified a person skating, but “roller blades” (with a space) resulted in no relevant information, as Ring linked “blade” to weaponry.
All our attempts to make the camera identify a person swinging a wooden baseball bat or holding a hammer were unsuccessful. The “Hammer” results featured a hedge trimmer and a leaf blower. “Water gun” mentioned two cases of individuals using water guns, whereas “gun” did not.
Hamren mentioned she was struck by how a search for “volleyball” in her assessments found an individual wearing a standard volleyball outfit. However, Pressf was unable to have Smart Video Search identify the name of a football team on a jersey since the feature does not directly detect text.
What we found most fascinating about the search tool is the combination of uneasiness and chance of being aware of events during the extended periods when there is no need to actively check the footage. We gained profound understanding of the different methods a specific dog uses to relieve himself. The quantity of nights a family of raccoons, who appeared in a search for “bat,” wandered by was just as entertaining. It’s uncertain how these facts soothe anxiety or make someone feel a greater sense of connection. Search undoubtedly paves the way for an increasing number of Ring videos to gain popularity—and users are expected to find it beneficial as it evolves. Hamren admits that exploration hasn’t satisfied all her curiosities so far. The mystery of how one of her backyard tiles became stained is still unsolved. It seems that her camera was improperly placed to record the enigmatic event. Currently, an issue that AI is unable to resolve.
