Interview
What are the end user demands for more intelligence across Edge applications? Te Edge is where many AI implementations are best served. In fact, research company ABI Research believes that ‘on-device AI inference capability is expected to reach 60% of all devices by 2024’ outstripping AI functionality that is performed centrally. One reason for this is latency; AI applications such as instant on, attention tracking, face framing and privacy require an immediate response - therefore there is no time for data to be collected, transferred to the cloud, analysed for a decision and that decision to be transmitted back to the Edge device. Terefore, effective AI inferencing must be done at the Edge. Another reason is bandwidth. Transmitting huge amounts of data
continuously requires similarly huge and costly bandwidth rendering centralised AI impractical for client compute AI functions as described. Edge AI is also preferred for security reasons. With FPGA- base Edge AI, the data never leaves the Edge device, in this case the laptop, so a user’s information is much more secure. Tis can also have implications for ethical AI questions where users are concerned about their personal details being transmitted via cloud services, which could become compromised. With FPGA-based AI solutions this cannot happen.
Please could you explain how Lenovo and Lattice have achieved smarter human-computer engagement, privacy, collaboration, and better energy management through this technology integration? Internal testing at Lattice has shown that a user attention sensing application running on a Lattice FPGA was up to seven times more power efficient (in terms of inferencing operations per second per watt) than competing ASICs. Tis results in extended battery life and a better user experience. Lattice also takes ethical AI extremely seriously. Te company
actively trains its models with a diverse dataset including various backgrounds, skin colour, facial features, and hair coverings, while only acquiring data from authorised sources. Trough these practices, Lattice AI/ML models are validated to ensure that there are no biases built into the model. Overall, the more aware a PC is of the user – the better the user
experience will be and the smarter the device will be perceived to be, whether that is in power consumption, privacy, readiness, or any number of functions which are continually evolving.
Please could you explain more about the trend towards edge computing, what this is and what devices are currently being developed for this area? Edge computing is becoming dominant as we embrace AI functionality in more areas of our lives. Again from ABI Research, ‘By 2025 silicon that goes into edge compute capabilities will account for more revenue than silicon which is used in servers’. Applications such as Siri and Alexa can be delivered at the Edge, not from the Cloud. Tis is oſten for reasons of bandwidth and latency, but security of private information is also a concern. Edge applications are much better performing in terms of speed of response and there are no Cloud bandwidth issues. Because all information can reside on the Edge device, there are far less privacy and security concerns compared to Cloud processing.
www.pcr-online.biz
What will the ‘new era’ of smart PC’s deliver in terms of technology capability? The evolution of the PC tells us that new PCs, laptops and tablets will continue to deliver more for the user. New functionality is being added with each new release based on user demand, and manufacturers must deliver, or risk being left with obsolete machines. The way to add functionality quickly is to design-in programmable FPGAs that can do tasks with unbeatable energy efficiency while the CPU performs the main processing tasks. FPGAs, therefore, provide a low power, flexible and highly scalable way of introducing new features quickly without compromising the performance of the main machine.
How will AI feature in this next revolutionary step for always-aware, on-devices? AI will be employed to further understand the user, enabling better video conferencing that takes into consideration the user space and enable users to work and study from anywhere. These experiences will provide better audio and focused video quality.
Please could you explain more about new computer vision capabilities and what this means for the end user? In terms of client computing applications, the vision capabilities are simply functions that use the device’s camera. In effect, this means that the device can see and interpret actions by the user, and other parties within its field of vision. This can be a very useful function for power saving but needs to be implemented sensitively. For instance, AI can use the laptop’s camera to save battery life by dimming the screen or signaling the computer to go into sleep mode. Computer vision capabilities can also be used to enhance collaboration between parties on a Zoom or conference call, and for security purposes (protecting against visual hackers). However, it is imperative that computer vision capabilities are implemented ethically. AI models must be developed to ensure the same experience for users with different skin colour and/or clothing items. With planned pre-awareness and proper AI training models, poor user experiences can be avoided.
Are there any further tech integrations planned for future generation releases of this product line? Without speaking for Lenovo, new technical innovations are constantly in development. Improvements in ‘always-on and always-aware’ accuracy are in progress. Better collaboration determined by improved sound and vision using AI is also being considered at Lattice, while new sensors are being evaluated to deliver better interfacing.
What trends in the user experience are you seeing increasing demand for? One area that has increased tremendously in demand in recent years is health and wellness. AI can actively monitor a user’s posture or time online and deliver reminders and notifications to take a break from the screen, or to get up and stretch or exercise. We predict that wellness and health functions will be emphasised in nextgen client compute devices in the very near future.
March 2022 | 43
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52
