Samsung’s AI phones still have no true rivals, as Google fights app bad actors

I’ll float a question, for you to ponder – why has no phone maker been able to piece together a semblance of a challenge to Samsung’s runaway lead with ‘AI phones’? It was in January when Samsung gave us their first pitch at ‘AI smartphones’. The Samsung Galaxy S25 Ultra, the Galaxy S25+ and Galaxy S25, a troika that is perhaps the most powerful Android smartphone portfolio (in more ways than one). Other phone makers have very capable flagships too, but Samsung’s width of the portfolio remains unmatched. Samsung’s AI pitch, called Galaxy AI, turns up with a broad spectrum of utility much beyond the camera and the photo gallery. My experience points to a simple fact – the way Galaxy AI is envisioned, there is a daily utility aspect very much in place.

Circle to Search (it is convenient, and addictive to use), live translation of phone calls, a real-time interpreter, an AI ready Notes app and a recorder with AI generated transcription (for many of us, negates the need for another app and subscription; I did make a switch from Otter, for the duration), some of the still unique propositions that phones. Nothing we had seen baked properly into phones (functionality often available as separate apps) till the Galaxy S25 series came along. This underlined a definite intent. And if I may point out, not seen since either. That is, to say the least, a bit surprising. Back to the original point – why has no phone maker caught up with similar AI pitch since?

Quick access to our ‘AI phones’ coverage…

Read: Samsung Galaxy S24 Ultra review: AI bets, human nature and a generational shift

Read: AI’s chip wars are just getting started

I feel there are two reasons. One, Google and Samsung agreeing to a period of exclusivity for Gemini Pro and Gemini Nano for the Galaxy S25 phones. That advantage has worked well for the latest flagship portfolio (last year’s phones are getting updates too; scope broadens). Secondly, rival phone makers would need to find some level of uniqueness to go beyond the Galaxy AI’s composition, lest they simply follow the same template. Which runs the risk of annoying potential buyers. There’s flexibility in Samsung’s approach. Gemini Pro, a larger AI model, underlines Generative edit within the Gallery app, transcription and summaries in Voice Recorder and Samsung Notes. It sits alongside Gemini Nano, which is specific for on-device tasks such as assist in Messages. Samsung’s updates for these apps since, individually and as part of larger monthly updates, has only improved reliability and ease of use. My observations draw from the significant time I spent with the Galaxy S25 phones, particularly the Galaxy S25 Ultra.

Many phone makers are trying to compete with Samsung’s really broad Galaxy AI pitch, but most of their efforts with AI integration tend to be limited in scope. Often, that hints at some level of comfort with familiarity, which gives them a certain cushioning with experimentation.

• OnePlus recently added an AI Eraser mode to the gallery app on some of its smartphones. I tested this, and the results are impressive – erase and replace is much neater and smoother than I expected, and also experienced in some more illustrious implementations such as Google Photos. That bodes well for the broader AI features they’ve rolled out for China-specific phones for now.

• Vivo’s upcoming X Fold 3 foldable smartphone is expected to integrate a 7-billion parameter AI model on-device, and a 70-billion parameter model working on the cloud, for generative AI functionality that’ll make its way into the phone. Late last year, the smartphone maker had talked about an ‘AI matrix’ system, for a wide range of functionality.

• Google’s Pixel 8 Pro has added Gemini Nano as the foundation for summarise in Recorder, smart replies in Gboard, improving document scans and a range of photo and video editing features. Circle to Search is no longer exclusive to Samsung now, with the Pixel 8 phones now integrating that.

More analysis about AI and phones…

Read: A spark that illuminated the light: For AI phones, ecosystem aligns its pieces

Read: Android flagship battles resume as Vivo X100 Pro settles a photography yardstick

Read: OnePlus 12 is a refined flagship, bravely walking the path of conventionality

Read: Smartphone makers are wasting no time in the race for innovation and loyalty

Things may change at some point (there is a big “if” here), with availability of Microsoft’s Phi-Mini 3 AI model, the company’s smallest model thus far. They claim its performance and capabilities are similar to OpenAI’s GPT-3.5, complete with 3.5 billion parameters and can run on-device compute on lower power consuming hardware. But what we are talking about here is a perfect world scenario. Whether Microsoft chooses to focus on smartphones at all with its frugal AI model, is anyone’s guess. But a Gemini vs Phi-Mini 3 battle in the smartphone ecosystem, may be quite fun. In the meantime, Samsung will just have to confirm its plans for a paid subscription tier for Galaxy AI sometime in 2025 or later. Whether that plan is on.

SAFE?

The good folks over at Google got my attention on a set of principles dubbed SAFE, which they say helped block as many as 2.28 million policy violating apps from being published on the Play Store, more than 333,000 bad actor accounts for policy violations and more than 200,000 app submissions rejected because these apps unnecessarily requested for permissions to user data such as background location access or reading SMSes on the phone. What really is this SAFE, you ask? It isn’t a simple term, but a collective – (S)afeguard our Users, (A)dvocate for Developer Protection, (F)oster Responsible Innovation and (E)volve Platform Defenses. Try remembering that.

Nevertheless, the idea is clear – make the Google Play platform more active and responsive to what exactly is happening with app updates, permissions and new applications. “We have also strengthened our developer onboarding and review processes, requiring more identity information when developers first establish their Play accounts,” the tech giant makes its attempts clear, in a statement. This builds on from Google, Meta and Microsoft’s App Defense Alliance (ADA; other members include McAfee, OnePlus, Trend Micro and Eset) that had, late last year, moved under the umbrella of Linux Foundation.

An important development that Google’s data points out and illustrates a risky (though often glossed over, for the sake of choice and countering monopoly) side of third-party app platforms that alternative app stores promise – Play Protect real-time scanning on millions of Android devices globally, detected more than 5 million new, malicious off-Play apps installed from sources other than the Play Store.

PAY

In my column this week, I spoke about two issues that you may have come across, but perhaps not exactly registered as big enough problems. Or maybe you did. We’ve collectively talked up and continue to every month (rightly so too), the volume of payments that have become “digital”. Along the way, two initiatives emerged to further push this momentum – a second layer of authentication for subscription payments across platforms (be it Netflix, Apple App Store or Google Pay) that would be made using credit cards, and secondly, the RuPay credit cards plugging into the UPI (or unified payments interface; I’m sure you’ve encountered this). Here is where we are at.

Read (Premium): The gaps in India’s digital payments ecosystem need plugging

• It was a good move to bring a layer of security for subscriptions that can often be difficult to exit. But now, it has become a nightmare to use a credit card or debit card for subscriptions and recurring payments on the Apple App Store or the Google Play Store, be it for any app or service. Netflix struggles from time to time with credit cards. As does Spotify. As well as F1 TV Pro. Apps or platforms aren’t to be blamed.

• The move to get RuPay credit cards on to the UPI network was expected to enable bigger ticket purchases. The QR codes are already everywhere, and instead of the bank account, you would have another avenue. However, a UPI transaction made using a RuPay credit card attracts a charge called MDR, or merchant discount rate, which is essentially a processing fee that the card issuer levies on the merchant for that transaction. Currently, a charge of around 2%. For merchants who are used to UPI being “free”, this isn’t palatable. RuPay cards on UPI, are stuck.

Data tells me credit card transactions need to be smooth. In FY 2024, India’s credit card spends surfed around 27% year-on-year, to the tune of ₹18.26 trillion. In comparison, in FY24, there were 117.6 billion UPI transactions, totalling over ₹199.89 trillion, according to official numbers. Therefore, both issues need to be looked at, with regards to gaps in implementation. Is there something in the guidelines that online platforms have mis-read, and therefore the complication with subscription payments using credit cards? Can the MDR for card transactions on UPI be reworked, to balance the needs of funding for the system and what merchants are used to? We’ll just have to wait. Lest we lose the momentum.

KNOW

• A few days ago, top AI companies (you’d have heard most of these names) agreed to a new set of standards to counter the worsening issue of deepfakes, alongside child abuse, becoming quite rampant on social networks. The principles include a re-look at the training data used for AI models, as well as the need for watermarking AI generations and developing new detection solutions that’ll prevent generative tools from creating AI-generated child sexual abuse material, or AIG-CSAM. On board are 11 companies, including Meta, Google, Anthropic, Microsoft, OpenAI, Stability.AI and Mistral AI.

Read: As the world grapples with deepfakes, AI companies agree to a set of principles

• Apple is preparing for the big iPad refresh on May 7 (mark your calendars). I have a feeling that you’re in for a more than cursory generational specs upgrade for the new iPads. Anyone’s guess whether Apple is ready to unleash the OLED screens on tablets just yet, but this could be the first time an iPad refresh leads the line for a new series of Apple Silicon. But what is more certain than anything else is, the next chapter for Apple’s AI strategy. My understanding is, iOS 18 and iPadOS 18 arriving this year, will reconfigure a lot of the functionality and apps to embed an intelligent assistant. They already have the hardware in place, particularly the Macs and iPads from the last few years (for clarity, since Apple Silicon). The next chips will simply take that game forward quite significantly.

Also read: Apple MacBook Air 2024 ushers in a new era as a resounding response to AI PCs

The following article was originally published in this week's HT Wired Wisdom. Subscribe here.