Google announced innovative AI upgrades for Android. But it has a dark side. Google’s AI will forever read and analyze your private messages. So what does this mean for you, how do you maintain your privacy, and when does it start?
The following is updated on 1/29. This article was originally published on 1/27.
It’s understandable that Google would be excited to bring Bard to Messages. His UI like his ChatGPT ready-made for a ready-made user base of hundreds of millions. “This is an AI assistant that can enhance the messaging experience, from facilitating communication to increasing creativity to providing information. It becomes your personal AI assistant within the messaging app,” says Bard. .
But Bard also analyzes the private content of messages “to understand the context, tone, and interest of the conversation.” Analyze the sentiment of your messages and “tailor your response to fit your mood and vibe.” And it “analyzes your message history with different contacts to understand relationship dynamics and personalize your responses based on who you’re talking to.”
And now comes the next privacy battleground for smartphone owners who are still coming to terms with app permissions, privacy labels, tracking transparency, and the voice-AI assistant eavesdropping scandal still fresh in their minds. To do. Google’s challenge is to convince users that this doesn’t open the door to the same kind of privacy nightmares we’ve seen before with user content and AI platforms.
Message requests to Bard raise another less controversial privacy issue. These are sent to the cloud for processing, used for training, and can be viewed by humans even if anonymized. This data is stored for 18 months, remains for a few days even if you disable AI, and can be deleted manually.
Such requests fall outside of Google Messages’ new default end-to-end encryption and are literally sending messages to Google itself. There’s no denying this, but it pays to keep it in mind. As with all generative AI chatbots, including ChatGPT, you should assume that anything you ask is non-private. did it Will come back to haunt you.
But message analysis is different. This is content that is (currently) behind an end-to-end encryption shield in a world where such private messaging is the new norm. The ideal here is on-device AI processing where the data never leaves the phone, rather than uploading content to the cloud where more processing can be done.
This is where the Android vs. iPhone battleground could come into play. Historically, Apple has been much more powerful when it comes to on-device analytics than Google, which has defaulted to the cloud for analyzing user content.
“Apple is quietly enhancing its capabilities.” F.T. As we reported this week, “Apple’s goal is to bring AI to the next generation of iPhones…by manipulating the AI it generates through mobile devices, using AI chatbots and apps on the phone itself, rather than using cloud services.” It’s like being able to run on the hardware and software of a data center. ”
Updated 1/29:
The latest update on Apple’s own efforts to bring generative AI to iOS suggests that its intentions to keep everything on the device may not be as solid as expected.
The code I just discovered 9to5 mac The new iOS 17.4 beta reveals that progress. “Apple continues to develop new versions of Siri that leverage large-scale language modeling technology, with some help from other sources.”
as bloomberg’s Mark Garman wrote last fall that “When OpenAI’s ChatGPT took off like a rocket, Apple largely sat on the sidelines…watching as Google and Microsoft rolled out generative AI versions of their search engines. …The only notable AI release from Apple was improved auto-recognition capabilities.” Correct system for iOS 17. ”
Unsurprisingly, one of the “sources” helping Apple is ChatGPT, according to 9to5Mac, which says “Apple uses OpenAI’s ChatGPT API for internal testing to help develop its own AI models. It seems there is.”
According to 9to5Mac, “iOS 17.4 code suggests Apple is testing four different AI models, including two versions. [Apple’s internal model] AjaxGPT (including those processed on-device and those not). Apple appears to be matching the results of its own AI model to ChatGPT, and part of it includes an iMessage interface, but not all of it is on the device.
The problem for Apple is that Google’s setup lends itself to a more performant edge/cloud architecture that powers generative AI like ChatGPT. Although analysis of message content is guaranteed to occur only on the device, the reality is that millions of users demand new features and are willing to sacrifice opaque privacy themes that are not easily understood in the process. I’ll make it.
This puts Apple in a tight spot. The philosophy that “what happens on your iPhone stays on your iPhone” is deeply ingrained. However, the reality is that there are limits to the AI processing available at the edge and cloud, determined by processing advances, hardware cost, battery life, heat generation, and general operational limitations.
Garman said iOS 18 is “considered internally as one of the biggest iOS updates, if not the biggest, in company history.” He points out that “the debate continues.” [within Apple] Learn how to deploy generative AI: as a complete on-device experience, a cloud-based setup, or something in between. While on-device approaches work faster and help protect your privacy, deploying Apple’s LLM via the cloud allows for more advanced operations. ”
This decision about where AI processing occurs could be the biggest change to Apple’s privacy principles. Because Apple controls its entire ecosystem, it can bend physics in ways no other company can. But this is a whole new level of complexity.
When Apple really starts rolling out AI integration into iMessage, we’ll see how the more vocal and privacy-conscious members of its user base react. The fierce backlash to his proposed CSAM analysis showed how controversial this could become if Apple somehow replicated Bard’s content analysis. So the initial focus will be on Siri-style requests and support, and I don’t see that happening anytime soon.
Bard says, “Google guarantees that all Bard analytics happens on your device, which means your messages aren’t sent to any servers. , giving you full control over how you use it.”
But I wonder if it would actually water down the warranty on the device. It makes sense to provide a more seamless interface between smartphones and the cloud. There is a good chance that analysis and actual content will get lost in the mix. Some way of explaining the risks to the novelty-seeking user group will need to be found.
You must decide whether any guarantees give you sufficient comfort to publish your personal content to Bird.
Please note.what is the difference canThings that cannot be done, such as compromising end-to-end encryption it’s not Policies are enforced, including policies regarding where content analysis occurs. We strongly urge you not to publish content too freely until appropriate safeguards are in place.
Bird agrees. “Although Google ensures on-device analytics, data accessed by Bard is technically collected, even temporarily.” regarding potential leaks, misuse, or hidden data sharing practices. Concerns arise. Bard’s scope of analysis and how data is used must be transparent. Users have the right to granular control over what data is analyzed, for what purposes, and for how long. ”
What happens this year will define the landscape much more than anything we’ve seen thus far. Both Google and Apple are eyeing their messaging apps as the primary UI for generative AI capabilities, suggesting this could be a real game changer.
This generative AI chat and messaging integration will forever transform text messaging platforms, and will soon usher in a new competitive landscape between Google, Apple, and Meta, whose smartphone ecosystems and apps power our lives. It will open up.
“The exact date is still unknown, but all signs point to Bard coming to Google Messages sometime in 2024. It could be a matter of weeks or months, but They come without,” Bard said. On the other hand, what we’ve seen so far is buried deep in the Beta His release and may change before release.
When you’re live, think twice before unlocking your message privacy settings. “Ultimately, the decision to use message analysis is up to you,” Bird says. Please carefully weigh the potential benefits and privacy concerns and make an informed choice based on your own comfort level and expectations. ”
The privacy discussion here goes beyond analyzing message history. Google’s introduction of Bard is just part of a shift from browser-based search to direct search, and we need to be increasingly cautious about the quality of results we get. Bard is not a chat with friends. This is his UI that sits on top of the most powerful and valuable advertising and tracking machine in the world.
On that point, Bard left me with a final thought that might be better directed at authors rather than users. “Remember: You have the right to demand clarity, control, and responsible AI development from the companies you trust with your data.”
follow me twitter Or LinkedIn.
