In a statement posted on the Korean newsroom of the company, Samsung Kim Dae-hyun anticipates “generated AI that fits the level of users and operational technology for enhancing use experience and usability and people’s security technology.” The new news is the planned introduction of knowledge graph technology to enhance the experience of AI.
Knowledge graphs have been around for some time now, or at least this should bring relief to those who possess a historical and chronological perspective of less than a decade. And the electronic vault was the first to refer to encyclopedias integrating diverse amounts of information around Google. Nevertheless, Generative AI had seemingly given a different view of the edifice, even though the objective is largely the same: to collect data about the ‘what’s who’s and where's of a particular domain or task, organize information from various places, and create links between them. This allows knowledge graphs to provide context and layers to other, more statistically oriented machine intelligence possibilities, such as machine learning. ‘and act as intermediaries between humans and systems, for example, producing explanations that are readable for humans.’
For example, I get up in the morning with the expectation that I will have an AI assistant who will present the day’s plans and do everything else I need, like it is the most natural act. I believe this way of employing AI would eventually become a normal practice in our day-to-day activities rather than remain a futuristic dream. The implication is more creativity, bearing in mind specific requirements and preferences of a given group of users. We are advancing system integration that will facilitate knowledge graph technology, one major technology of bespoke AI, and build it alongside generated AI for user-case services.’
While that aspect may be new, the emphasis on hybrid AI remains consistent. Hybrid AI refers to 'a technology that combines the use of AI on-device and on the cloud for a safe and fast approach’. Nowadays, on-device intelligence can be utilized for the quick provocation of information while privacy remains intact. However, the information stored in the cloud makes use of advanced computation technologies and enormous data storage and processing capabilities. In different settings and circumstances, the most suitable and efficient use of AI more effectively than ever before can be achieved.
On-device and cloud AI systems can be implemented individually or concurrently, depending on the functional technical expertise needed. It has been the assumed boundary with regards to safety and confidentiality concerns. “Safe AI” is a must not only for all the services utilizing AI but also for the AI-powered services that have a personal touch. We offer personal data-based artificial intelligence for the ease of the users, but it should be risk-free in terms of exposure of private information.”
To put it straightforwardly, the more such activities involve the individual in the private activity, the more potentially sensitive a user that must be processed on-device with lower processing and higher security, such processes are likely to be.
In that way, we had expected that Apple would have taken some similar turns to the situation with the on-device AI, or would have been even more restrictive and refused on-device AI only. Instead, however, it has chosen a radically different path. Private Cloud Compute, this time it claims, comes with'revolutionary privacy and security defense.’ Once again, in layman’s terms, this entire apparatus is the Apple silicon solution from a device to the cloud with the sole aim of declaring that ‘the personal user data sent to PCC is for the users only; even Apple cannot access it.’ Describing PCC, iMaker states that it is... the most elaborate security system in the history of humanity created for cloud iterations of deep learning.
There is a clear dichotomy now. Either pace-making advances in AI are taking place through a sociotechnical system characterized by a fluid, hybrid architecture, or there is the personal phone extension in a cloud that guarantees more security and privacy but comes with hefty costs in terms of enabling such technologies and speed of progress, particularly the use on a large scale of third-party artificial intelligence augmentations.
To disrupt the new structure, Apple has made available bonuses of up to $1 million, which, although a drop in the ocean, did generate headlines as it was meant to. So far, we have not witnessed any serious device-bound or cloud-based AI offerings attacks from the likes of Google, Samsung, or Apple. In due course, they will come. What remains quite ambiguous, however, is how this would draw the lines between Samsung, Apple, and the more cloud-oriented Pixel offering. Let’s not forget, still picture no other features of device lockdown battles does the Android vs. iPhone security angle fight engage from any other perspective.
In the past, I have made recommendations to Samsung concerning an adjustment to the PCC presented by Apple. What we have seen now, however, looks more like a reaffirmation of the hybrid AI strategy. In the meantime, it does appear that Samsung understands that both Google and Apple pose threats for the next generation of AI-based gizmos and hates all the delays regarding its Android 15. Now it is left to the respective audience. For the existing owners of Galaxy phones, the options that will be available in 2025 are becoming more attractive and intense. It is time to make the decision.
Comments