Lenovo has kicked off its Tech World gabfest by bending a flexible smartphone concept into its AI vision, which goes from the cloud to the wrist.

The Chinese kit-maker and its Motorola smartphone unit used the conference to show off "an adaptive display concept that molds to our consumer needs."

In non-marketing English that translates as a "FHD+ pOLED display that can be bent and shaped into different forms."

Here it is standing up.

And here it is clinging to a wrist.

There's no word on when or if this vision will become an actual product - although Lenovo asserted the concept devices are the result of the same research efforts that helped create the rollable smartphone and laptop it showed off at last year's Tech World.

Neither has been sighted since, although HP and LG have started selling laptops with foldable screens.

Lenovo reckons its concept machines are a perfect match for AI - the tech that is to conference keynotes in 2023 what warnings about your industry's imminent disruption by an Uber clone were to keynotes in 2017.

Motorola's vision is for a generative AI model "that runs locally on the device to allow users to extend their personal style to their phone. With this concept, users can upload or capture a picture of their outfit to produce multiple unique AI-generated images that reflect their style. These images can then be used as a custom wallpaper on their device."

Lenovo, meanwhile, is developing a personal assistant for both PCs and smartphones that "has a foundational model uniquely personalized to the user [and] also has a knowledge base that adapts to patterns and usage with natural voice or text interaction capabilities."

Siri? Google? Do you think this will escape the irrelevance suffered by Cortana or Bixby?

Lenovo is clearly adept at using cute concept devices to grab attention. But CTO Dr Yong Rui also articulated a role for the bendy phones in the context of the PC builder's plans to provide AI infrastructure.

Dr Rui explained Lenovo's "Hybrid AI framework" - the idea that three important types of AI model will emerge:

• Public foundational models will be trained on internet-scale data, be capable of many tasks, be publicly accessible and are already visible in services like ChatGPT.
• Private foundational models will be accessible only to certain users and be trained on an organization's data.
• Personal foundational models will run on users’ own PCs and smartphones.

Dr Rui mapped those models to Lenovo's products. Its hardware for hyperscalers can power a public model. On-prem hardware can handle private models, and shrink them down to a size at which they can run on PCs and smartphones. Lenovo PCs and Motorola smartphones.

If the private model needs to extend to an ERP system or other source of on-prem or cloudy data, Lenovo intends to make it happen.

Of course it does - Lenovo exists to sell more stuff, and recently stated a services business to complement its TruScale IT-as-a-service offering.

Lenovo is, of course, not a player in either GPUs or software.

It's therefore done the other thing that every tech conference in 2023 needs: stage a surprise appearance with Nvidia CEO Jensen Huang for a grip'n'grin with the host CEO - in this case Yuanqing Yang - to announce a partnership.

The alliance between Nvidia and Lenovo will see custom AI models built on the Nvidia AI Foundations cloud service, moved to on-prem Lenovo systems packing Nvidia GPUs and running Nvidia software.

Which perhaps demonstrates another important element of tech vendor strategy in 2023: that hardly any of it can happen without Nvidia. Of course the GPU-maker is also an industry bottleneck - because it can't crank out enough kit to satisfy demand immediately.

Thankfully, at least for buyers outside China, US sanctions prohibit the export of even modest GPUs to China.

Which ironically is great news for China-headquartered Lenovo - its customers might have to buy more servers to run more of those less-powerful GPUs. ®

https://www.theregister.com//2023/10/25/lenovo_ai_strategy_bendy_phone/