Wednesday, May 22, 2024

How Much Lag Time Before Widely-Useful Apps Develop for AI PCs?

Microsoft’s new Copilot+ artificial intelligence personal computer illustrates a common pattern in technology deployment, namely that hardware or platform capabilities have to come first before widely-used applications can be developed to take advantage of the new platforms. 


That can be seen in mobile network generations, where it can take five years or more for new applications taking advantage of the new platform to be developed, and longer for even popular innovations to be widely embraced. 


An optimistic estimate--showing app availability and not widespread use--suggests it takes two to three years for a new capability to be available on a new mobile platform, even in a relatively difficult or limited user experience, and has relatively low adoption.


Mobile Generation

Year of Introduction

Example Widely-Used App

Lag Time

3G

1998

Mobile Web Browsing

2-3 years



Mobile Email

2-3 years

4G

2010

Mobile Streaming (Music, Video)

3-5 years



Video Calling

3-5 years

5G

2019

High-Definition Video Streaming

2-4 years



Mobile Cloud Gaming

3-5 years


That same process tends to unfold for other innovations as well. The new CoPilot+ AI PC will require 16 GBytes of RAM and 256 GB of storage, for example. The device also will require an integrated neural processing unit with performance rated at 40 trillion operations per second. 


Right now, the only application that really takes advantage of that processing power is the “Recall” feature, that indexes and retains (on the PC) an image of each page a user has opened, for about three months of activity. 


The feature is touted as allowing users to find content they have viewed recently. Early on, most users might not find that one use case compelling enough to cause them to buy a new PC. But, over time, use cases could develop that do provide high value, even if we do not know what they are, yet. 


The point is that, for most users, it might be some time before the utility of an AI PC is obvious. 


Study

Platform

Timeframe for New Applications

Key Observation

"The Rise of Killer Apps: How Killer Apps Emerged from Past Technological Shifts" by Hyejin Kim

Mobile Phone Camera

3-5 years

Points to the launch of the iPhone's App Store in 2008 as a turning point for mobile camera apps.

"The Innovator's Dilemma: When New Technologies Cause Great Firms to Fail" by Clayton M. Christensen

Various Technologies

Varies

Emphasizes the disruption caused by new technologies and the challenges established firms face in adapting.

"AI Revolution: Road to Superintelligence" by Tegmark, Max

Artificial Intelligence

5-15 years

Discusses potential timelines for AI advancements but acknowledges significant uncertainty.

"Gartner Hype Cycle: Special Report" by Gartner

Various Technologies

Varies

Introduces the concept of the Hype Cycle, which illustrates the phases new technologies go through, including a trough of disillusionment before widespread adoption.


That has been the case for prior innovations related to the PC, such as the graphical user interface. That also was the case for the hobbyist phase of the personal computer evolution. 


Study Title

Platform

Timeframe for New Applications

"Dealers of Lightning: Xerox PARC and the Dawn of the Computer Age" by John Seabrook [Book]

Graphical User Interface (GUI) with Xerox Alto

10-15 years

"From Mainframes to Micros: A History of the Personal Computer" by Paul E. Ceruzzi [Book]

Personal Computers (pre-Apple II)

3-5 years

No comments:

When Will it Make Sense to Build a Custom Generative AI App?

Yes, it is becoming easier for larger enterprises to build custom generative AI applications. As always, that will not necessarily mean it i...