I walked into CES 2024 in Las Vegas and was immediately hit with the fact that AI is now the default narrative for consumer electronics. Every product category had an AI twist, from televisions with AI upscaling to laptops with AI chips and smart home devices with LLM integration. The AI feature became the headline regardless of the product, and it was clear that this was the direction the industry is headed.
One of the most significant technical developments at CES 2024 was the emergence of the NPU, or neural processing unit, in mainstream laptop processors. Intel's Meteor Lake, AMD's Ryzen 8040, and Qualcomm's Snapdragon X Elite all ship with dedicated NPU silicon, which handles AI inference tasks without needing to route to the GPU or CPU. This means lower power consumption and lower latency for on-device AI workloads, a significant improvement over previous generations.
The NPU is such a key component that Microsoft later formalised it as a requirement for AI PC labelling in their Copilot+ PC category, announced in May. This move solidifies the NPU's position as a crucial part of the AI ecosystem, and it will be interesting to see how this plays out in the market.
Samsung and LG were both touting their AI-powered TVs, with features like upscaling standard video to 8K resolution and optimising colour and contrast scene by scene. They're using convolutional neural networks running on embedded chips to achieve this, and while the marketing may be stretching the definition of 'AI', the technology itself is real and impressive.
For instance, I saw Samsung's QLED 8K TVs in action, and the upscaling capabilities were indeed impressive. The TVs use a combination of machine learning algorithms and traditional video processing techniques to upscale lower-resolution content to 8K. While the results weren't always perfect, it was clear that the technology has come a long way. I also noticed that the TVs' AI-powered features, such as automatic picture mode switching, were quite useful in real-world scenarios.
The Rabbit R1, a $199 AI assistant device, generated a lot of conversation at CES 2024. It has an action model that can operate apps on your behalf, which resonated with the post-ChatGPT idea of what AI devices should be able to do. Whether the R1 delivers on this promise remains to be seen, but it's clear that there's a hunger in the market for devices that can take action rather than just provide information.
One thing that was noticeable across all the AI-labelled hardware at CES 2024 was the gap between the impressive hardware capability and the software that's available to take advantage of it. NPUs in laptops are available before the operating system and application ecosystem are optimised to use them, and AI televisions can run neural network inference but the scenes where it makes a visible difference are narrower than the marketing suggests. For example, I spoke to a developer who was working on an AI-powered video editing application for laptops with NPUs. They mentioned that it was taking longer than expected to optimise the app for the NPU, due to the complexity of the neural networks involved.
The hardware inflection point in 2024 is real, but the software that makes it valuable is still on a slower schedule. This is not necessarily a bad thing, as it gives developers time to catch up and create applications that can take full advantage of the new hardware capabilities. However, it does mean that the full potential of these devices may not be realised for some time.
As I looked around at all the AI-powered devices on display at CES 2024, I couldn't help but feel that we're at a turning point in the industry. The question is, what will it take for the software to catch up with the hardware, and how will that change the way we interact with our devices?