PATENT DROP: Meta will hear you out
Discover more from PATENT DROP
PATENT DROP: Meta will hear you out
Plus: Zoom’s AI push; eBay watches you shop
Happy Monday and welcome to Patent Drop!
Today, we’re checking out a patent from Meta to learn your voice; a filing from Zoom for better AI training; and eBay’s plans to watch you scroll.
Let's dive in.
#1. Meta’s recording studio
Meta may want to add your voice to the stockpile of data it has on you.
The company is seeking to patent a system for “user identification with voiceprints” for social media networks. As the title implies, a voiceprint is essentially an amalgam of a user’s voice that Meta would use to identify or authenticate a user, comparing a user’s audio input to the collected voiceprint.
In particular, a voiceprint could be a part of a two-factor authentication system, with a user’s voice being one of the steps alongside a password or other biometric scan. Meta says this is more secure than using solely a biometric, such as a thumb or retina scan.
But authentication isn’t the only way that Meta would use your voice. Meta said in the filing that these voiceprints could allow the company to provide “customized content to the identified users.” For example, if Meta’s system detects “one or more people speaking, and the people speaking may be identified as users,” then the system will populate with content to “match the interests of the identified users, and may include advertisements, news feeds, push notifications, place tips, coupons, or suggestions.”
So, if Meta’s system hears you speaking, it may pull up specific types of ads on your Facebook feed, or content you typically gravitate towards on your Instagram.
Users would be allowed to opt in or out within privacy settings, Meta noted. Users can also specify that voiceprints can only be used for specific purposes, such as sending voice messages or using other voice-operated features, and can’t be shared with a “third-party system.”
Meta isn’t the first company we’ve discussed on Patent Drop that wants to get to know your voice: PayPal is seeking to patent AI-based methods of voice authentication for customer service call scenarios.
Voice biometrics aren’t new, gaining popularity in recent years alongside the movement for password-free security, said Aubrey Turner, executive advisor at Ping Identity. Adding voiceprints as part of a security strategy makes the process of multi-factor authentication more convenient, Turner said, as this method allows companies to “meet the customer where they’re at.”
Like any other method of cybersecurity, however, a 10-foot wall generally leads to a 12-foot ladder from attackers. In this case, that 12-foot ladder is AI voice dupes, said Turner. “The ability to spoof and impersonate with (AI) has sort of balanced out some of the interest in voice authentication,” he said.
But voice authentication can be part of a cybersecurity strategy if that strategy doesn’t stop after logging in, said Turner. No form of cybersecurity is fool-proof, but when multi-factor authentication is paired with monitoring for “risk signals,” such as looking at device type and location of log-in attempts, or what a user is looking to do in an account, companies can prevent a good deal of cyberattacks.
“Particularly post authentication, the context of a transaction and any number of other signals that we can leverage as part of that … is how you balance that ‘10-foot wall, 12-foot ladder’ scenario,” said Turner.
Though voice biometrics can be done responsibly, secure data management protocols are vital when implementing any kind of biometric authentication, said Turner. Meta, however, doesn’t have the best track record: The company faced a record $1.3 billion fine in the EU, and paid $725 million to settle a class action lawsuit over Cambridge Analytica (though it did not admit wrongdoing in the situation). Privacy experts have also raised concerns with the data handling policy of Threads, the company’s recently-released Twitter competitor which hit 100 million users in a matter of days.
“They've had their hand slapped more than once,” said Turner. “Privacy is not something to overlook when it comes to voice authentication.”
#2. Zoom’s training ground
Zoom wants to make sure its AI is being trained the right way.
The company filed to patent a “context similarity detector” for training AI models. This system aims to solve the problem of “data context mismatch,” or when a model is trained on data that doesn’t relate to its purpose, which can slow down development of AI models.
Zoom’s detector essentially aims to sniff out which training data sets are worth using. The system identifies which data sets will work by taking in two data sets – one that the system is testing, and another that is known to be of the right context – combining them, and coming up with a “context similarity score.”
If the score is above a certain threshold, then the datasets are similar enough to use as training material for the AI model. Zoom said that this system takes the manual labor out of comparing two datasets, which can involve visually inspecting a ton of qualitative data, or listening to hours of audio samples if the AI involves audio processing (which, given that Zoom’s business relies on audio-visual tech, it’s AI likely would.)
“The manual process can be difficult, inconsistent, or time-consuming … when trying to determine context-similar training and test datasets in audio environments, where training datasets can be from a variety of disparate data sources,” The company noted in its filing.
That Zoom is looking for ways to efficiently train AI models – specifically AI that handles audio and video processing tasks – is a no-brainer. In March, the company unveiled AI features for Zoom IQ, the platform’s “smart companion,” in partnership with OpenAI, including message and email composition features and a meeting summary generator. Zoom also announced a feature called Intelligent Director, which aims to use AI for better video quality in hybrid meetings, in late June.
In May, the company also invested an undisclosed sum in AI startup Anthropic, with plans to integrate the company’s chatbot, called Claude, across its platform, starting with its Contact Center product for business accounts. In a June press release regarding expansions to Zoom IQ, Zoom noted that it utilizes its own proprietary large language models alongside those of other companies in what it calls a “federated approach to AI.”
But it’s not all sunshine, rainbows and machine learning. Like many companies, Zoom faced layoffs of 15% of its staff in February after a period of pandemic-induced hypergrowth. While the company’s revenue is still growing, it only saw an uptick of 7% year-over-year in 2022, compared to 55% in 2021. Things weren’t looking any better in the latest quarter: Revenue grew just 3%, while costs soared 33%. Net income tumbled to $15.4 million from $113.6 million year-on-year.
With high-profile partnerships, AI productivity tools and a way to quickly and effectively train models as this patent describes, Zoom may be clinging to this tech as a way to maintain growth momentum.
SPONSORED BY VANTAGEPOINT
Remember 1991?
The World Wide Web had just been launched, the Cold War had ended, and an obscure grunge band named Nirvana was breaking into the mainstream.
You may not have been trading the markets back then, but many intelligent traders were using VantagePoint’s very first predictive analysis AI tool.
That’s what sets VantagePoint apart from others who are just now jumping in on the current AI wave – they’ve been developing their tools since the Soviet Union was around.
Now, it’s summer 2023. The lemonade is fresh. And the juice is always worth the squeeze when you know the best time to enter and exit trades.
This isn’t some one-off AI-driven solution that emerged overnight to ride the hype. Register and attend VantagePoint’s next free live AI stock market training, and you’ll receive a brand-new accuracy study immediately after.
#3. EBay’s keen eye
EBay wants to track your scrolls, clicks and zooms to build a personalized shopping heatmap.
The company wants to patent a system using computer vision to increase “efficiency in user interaction” by tracking interactions when online shopping. Essentially, eBay’s technique is another way to monitor user interaction using AI-based technology.
Using this system, eBay monitors user interaction with “digital content.” That can be anything, including results from a search query, a digital ad, or an AR or VR shopping experience. Using computer vision, this system identifies images and objects that a user seems to interact with, such as if they’re clicking or zooming in, and the characteristics of those objects.
For example, if a user is browsing in an AR shop and they zoom in on a dress, the system will take note of the style, color or material. That information, in turn, is used to categorize that user within a “user segment,” which helps eBay “control output of subsequent digital content to the user,” such as customizing recommendations and digital marketing.
EBay also notes that its computer vision system can be used for “missing item determination,” or determining if items are “missing from a particular context” when you’re using AR to shop. For example, if you show this system a photo of a living room with a blank wall, it may use what eBay calls “complete-the-look techniques” to suggest wall art.
EBay is leveraging computer vision the same way that many companies are using other forms of AI: to track user behaviors with the goal of hyper-personalizing targeted ads. Patents for AI monitoring tech from Walmart, Uber, Visa and more prove that companies are desperate to understand what actually makes a user click.
But eBays’ tech offers a more personalized solution, said Toby Awalt, VP of marketing at computer vision checkout company Mashgin. Because this tech relies on object image detection, rather than language-based tags or descriptions, it allows eBay to basically perform “advanced heat mapping through watching what somebody consumes.”
While eBay can — and likely will — use this for targeted advertising, this has the potential to greatly improve search queries by understanding the specific attributes of items a user commonly searches for, said Abhinai Srivastava, founder and CEO of Mashgin. Because most search engines rely on metadata and keywords associated with an image or product listing to decide what surfaces, the results sometimes aren’t as relevant as a user would like.
Computer vision like this could benefit any ecommerce company that relies on a search bar, Srivastava said. “The ability to have signals from the image impact the ranking … I would be surprised if at least the top players are not already doing it.”
What may make this tech difficult to put into practice is that eBay and other major ecommerce brands sell “everything under the sun,” Srivastava noted. Training AI to catch thousands of different attributes won’t be an easy endeavor.
One last consideration: eBay’s patent filing mentioned that it can use its computer vision technology with “digital images … scraped from a social media profile of the user, a photo sharing web page of the user, an email of the user, and so forth.”
While computer vision within eBay’s own platform can vastly improve search results and ad placements, using that tech on images scraped from a user’s social media may take it a bridge too far, said Srivastava. “The line between very relevant and very creepy is a fine one.”
Extra Drops
A few more tidbits before you swipe away.
Visa doesn’t want to keep you waiting. The company is looking to patent a system for “wait time estimation” using predictive modeling, which, as the name implies, predicts how long someone will be waiting in line at a store.
Have you ever wanted your video game character to look more like you? Sony has something for that. The company is seeking to patent a technique for “combining (a) user's face with (a) game character.”
Baidu wants to shepherd you through the mountains. The company is seeking to patent a system for constructing 3D maps that can show the fluctuations in terrain, such as mountains or large buildings.
What else is new?
Sen. Elizabeth Warren urged the SEC to investigate Tesla over Elon Musk’s Twitter takeover causing potential “conflicts of interest, misappropriation of corporate assets, and other negative impacts to Tesla shareholders.”
The House Judiciary Committee will investigate content moderation policies on Meta’s Threads, adding to its ongoing investigation of tech platforms’ policies.
Sony and Microsoft signed a 10-year deal to keep selling ‘Call of Duty’ on PlayStation after the Activision acquisition is closed.
Have any comments, tips or suggestions? Drop us a line! Email at admin@patentdrop.xyz or shoot us a DM on Twitter @patentdrop.
 | A guest post by
|