On Smartphones, AI Already is Nearly Ubiquitous

Anyone trying to model artificial intelligence usage is immediately faced with a number of problems. AI already is used to support natural language processing, image processing on phones, recommendations, customer service queries, search functions and e-commerce. 

And some would argue the embedding of AI into popular smartphone processes started in 2007, meaning the widespread consumer use of AI on smartphones capable of natural language processing and camera image processing has been underway for at least 16 years already. 

In that sense, some might argue the use of AI on smartphones already has reached levels in excess of 95 percent. And that analysis ignores other areas of common consumer use, such as search, e-commerce, recommendation engines and social media, for example. 

Still, that example probably strikes most people as overstating the present application of AI. What most people likely have in mind is a future where virtually all popular web-related or app-related interactions embed AI in their core operations, so that AI becomes a foundational part of any experience.  

That might also imply that AI “usage” could grow much faster than any other discrete application or technology, since it would be part of nearly-all app experiences. 

All that shows the importance of defining what we mean by “AI use” and when such use is said to have started. In some discrete use cases, such as NLP and camera processing on smartphones, AI might plausibly be said to have reached 95 percent adoption by consumers. 

Generative AI and large language models, on the other hand, are still at the beginning, and arguably have not yet reached anywhere close to regular use by 10 percent of internet users. 

‘Beyond that, AI is a cumulative trend, representing different types of function and eventually to be used in virtually all popular apps and hardware. 

So we already face the problem that AI is not like earlier app adoption. 

Popular internet-using applications generally have taken two to five years to reach 10-percent usage by all internet users, for example. That is generally an inflection point where usage then grows to become a mass market trend. 

The difference with AI is that it will be embedded into core operations of virtually every popular app, hardware and software. So the “adoption or use of AI” will have a cumulative effect we have not seen before, with the possible exception of the internet itself. 

Still, it took roughly 12 years for internet usage to reach a level of 10 percent of people. With AI, embedded into virtually all major forms of software and hardware, adoption should be faster than that. Just how much faster remains the issue. 

The AI advantage is that if we set 2022 as the AI equivalent of 1995 for the internet, AI already begins with a higher start point, as it is used widely for smartphone image recognition, natural language queries, speech-to-text, recommendation engines and e-commerce. 

Unlike virtually all prior innovations, AI starts with higher usage from the inception, and is a multi-app, multiple-use case trend. 

So it might make more sense to set start levels for AI much earlier: 

• Recommendation engines, 1990s

• Image processing on smartphones, 2000s

• Search, 2000s 

• E-commerce, 2000s

• Social media, 2000s

Natural Language Processing, 2009

Looked at in that way, and looking only at AI use on smartphones, the AI trend has been underway since at least 2007. 

Year	Smartphone Image Processing	Natural Language Processing
2007	Apple's iPhone introduced face detection for unlocking.	Siri, Apple's virtual assistant, is first introduced.
2009	HTC's Desire integrates Google's Goggles, an image recognition app.	Nuance's Dragon Dictate for iOS is released.
2011	Samsung's Galaxy S II introduces image stabilization and HDR photography.	Apple's Siri expands its capabilities to include voice commands for various tasks.
2012	Google Camera app introduces features like panorama mode and HDR+.	Google Now, a personal assistant, is launched.
2013	HTC's One M8 features a dual-lens camera for depth-of-field effects.	Apple's iMessage gets voice recognition for dictation.
2014	Google's Pixel smartphone introduces computational photography with features like HDR+ and Night Sight.	Google Assistant is introduced.
2015	Dual-lens cameras became more common.	Apple's Siri gains the ability to control smart home devices.
2016	Artificial intelligence (AI) starts to play a more significant role in smartphone image processing.	Google Assistant continues to evolve.
2017	AI-powered facial recognition becomes widely used in smartphones for security purposes.	AI-powered facial recognition becomes widely used in smartphones for security purposes.
2018	AI-powered image editing tools become more sophisticated.	Google Assistant expands its capabilities to include language translation.
2019	AI-powered augmented reality (AR) apps begin to gain traction.	Google Assistant becomes more integrated with other Google products.
2020	AI-powered chatbots become more common in smartphone apps.	Google Assistant gains the ability to make phone calls and send text messages.
2021	AI-powered health and fitness tracking apps become more sophisticated.	Google Assistant gains the ability to interpret and respond to natural language conversations.
2022	AI-powered language translation becomes more accurate and real-time.	Google Assistant becomes more integrated with smart home devices.

The point is that the expression “AI usage” by the general public and most internet users is already problematic. In some ways, such as smartphone image recognition and natural language processing, AI already is nearly ubiquitous. In other areas use cases are nascent.

We Will Overestimate what Generative AI can Accomplish Near Term

For most people, it seems as though artificial intelligence has suddenly emerged as an idea and set of possibilities. Consider the explosion of interest in large language models or generative AI.

In truth, AI has been gestating for many many decades. And forms of AI already are used in consumer applicances such as smart speakers, recommendation engines and search functions.

What seems to be happening now is some inflection point in adoption. But the next thing to happen is that people will vastly overestimate the degree of change over the near term, as large language models get adopted, just as they overestimate what will happen longer term.

That is an old--but apt--story.

“Most people overestimate what they can achieve in a year and underestimate what they can achieve in ten years” is a quote whose provenance is unknown, though some attribute it to Standord computer scientist Roy Amara. Some people call it the “Gate’s Law.”

The principle is useful for technology market forecasters, as it seems to illustrate other theorems including the S curve of product adoption. The expectation for virtually all technology forecasts is that actual adoption tends to resemble an S curve, with slow adoption at first, then eventually rapid adoption by users and finally market saturation.   

That sigmoid curve describes product life cycles, suggests how business strategy changes depending on where on any single S curve a product happens to be, and has implications for innovation and start-up strategy as well. 

source: Semantic Scholar 

Some say S curves explain overall market development, customer adoption, product usage by individual customers, sales productivity, developer productivity and sometimes investor interest. It often is used to describe adoption rates of new services and technologies, including the notion of non-linear change rates and inflection points in the adoption of consumer products and technologies.

In mathematics, the S curve is a sigmoid function. It is the basis for the Gompertz function which can be used to predict new technology adoption and is related to the Bass Model.

Another key observation is that some products or technologies can take decades to reach mass adoption.

It also can take decades before a successful innovation actually reaches commercialization. The next big thing will have first been talked about roughly 30 years ago, says technologist Greg Satell. IBM coined the term machine learning in 1959, for example, and machine learning is only now in use. 

Many times, reaping the full benefits of a major new technology can take 20 to 30 years. Alexander Fleming discovered penicillin in 1928, it didn’t arrive on the market until 1945, nearly 20 years later.

Electricity did not have a measurable impact on the economy until the early 1920s, 40 years after Edison’s plant, it can be argued.

It wasn’t until the late 1990’s, or about 30 years after 1968, that computers had a measurable effect on the US economy, many would note.

source: Wikipedia

The S curve is related to the product life cycle, as well. 

Another key principle is that successive product S curves are the pattern. A firm or an industry has to begin work on the next generation of products while existing products are still near peak levels. 

source: Strategic Thinker

There are other useful predictions one can make when using S curves. Suppliers in new markets often want to know “when” an innovation will “cross the chasm” and be adopted by the mass market. The S curve helps there as well. 

Innovations reach an adoption inflection point at around 10 percent. For those of you familiar with the notion of “crossing the chasm,” the inflection point happens when “early adopters” drive the market. The chasm is crossed at perhaps 15 percent of persons, according to technology theorist Geoffrey Moore.

source 

For most consumer technology products, the chasm gets crossed at about 10 percent household adoption. Professor Geoffrey Moore does not use a household definition, but focuses on individuals. 

source: Medium

And that is why the saying “most people overestimate what they can achieve in a year and underestimate what they can achieve in ten years” is so relevant for technology products. Linear demand is not the pattern. 

One has to assume some form of exponential or non-linear growth. And we tend to underestimate the gestation time required for some innovations, such as machine learning or artificial intelligence. 

Other processes, such as computing power, bandwidth prices or end user bandwidth consumption, are more linear. But the impact of those linear functions also tends to be non-linear. 

Each deployed use case, capability or function creates a greater surface for additional innovations. Futurist Ray Kurzweil called this the law of accelerating returns. Rates of change are not linear because positive feedback loops exist.

source: Ray Kurzweil  

Each innovation leads to further innovations and the cumulative effect is exponential. 

Think about ecosystems and network effects. Each new applied innovation becomes a new participant in an ecosystem. And as the number of participants grows, so do the possible interconnections between the discrete nodes.  

source: Linked Stars Blog 

Think of that as analogous to the way people can use one particular innovation to create another adjacent innovation. When A exists, then B can be created. When A and B exist, then C and D and E and F are possible, as existing things become the basis for creating yet other new things. 

So we often find that progress is slower than we expect, at first. But later, change seems much faster. And that is because non-linear change is the norm for technology products.

Gigabit Services are Right on Schedule According to Edholm's Law and Nielsen's Law

U.S. home broadband customers buying gigabit tiers of service grew 35 percent year over year in the third quarter of 2022, according to Openvault. At the moment, more than 15 percent of U.S. home broadband accounts use gigabit connections. 

Also, more than half of home broadband accounts buy service in the 200 Mbps to 400 Mbps range. That group grew 100 percent year over year. 

A little more than a year ago about half of households were buying service in the 100 Mbps to 200 Mbps range, showing that Nielsen’s Law and Edholm’s Law of bandwidth supply continue to operate. 

source: Openvault 

Edholm’s Law states that internet access bandwidth at the top end increases at about the same rate as Moore’s Law suggests computing power will increase. Nielsen's Law essentially is the same as Edholm’s Law, predicting an increase in the headline speed of about 50 percent per year. 

Nielsen's Law, like Edholm’s Law, suggests a headline speed of 10 Gbps will be commercially available by about 2025, so the commercial offering of 2-Gbps and 5-Gbps is right on the path to 10 Gbps. 

source: NCTA  

Headline speeds in the 100-Gbps range should be commercial sometime around 2030. 

How fast will the headline speed be in most countries by 2050? Terabits per second is the logical conclusion. Though the average or typical consumer does not buy the “fastest possible” tier of service, the steady growth of headline tier speed since the time of dial-up access is quite linear. 

Gigabit tier subscribers hit an inflection point last year. The rule of thumb is that any successful and widely-bought consumer technology enters its mass adoption phase when about 10 percent of homes are users. For U.S. gigabit adoption, that happened in 2021. 

Some might attribute the Covid pandemic and work from home as driving the change, but adoption rates would have taken off in 2021 in any case, as predicted by the 10-percent-of-homes adoption theory. 

It also is easy to predict that 2 Gbps to 4 Gbps is the next evolution, as speeds at the top end continue to increase by 50 percent a year. Ny 2025 we should start seeing the first 10-Gbps services deployed at scale.

Experts Say Metaverse Will Not be Common in Consumer Life in 2040. Why?

Experts surveyed by Pew Research believe that augmented and mixed-reality applications will dominate full virtual reality environments in 2040. But half of the experts also believe the “metaverse” will not be common in the lives of most consumers by that point. 

source: Pew Research 

This will be unwelcome news for many metaverse proponents. But it is historically realistic. 

Major technology transitions typically take much longer than proponents expect. One common facet of new technology adoption is that change often comes with a specific pattern: a sigmoid curve such as the Gompertz model or Bass model. 

S curves explain overall market development, customer adoption, product usage by individual customers, sales productivity, developer productivity and sometimes investor interest. It often is used to describe adoption rates of new services and technologies, including the notion of non-linear change rates and inflection points in the adoption of consumer products and technologies.

In mathematics, the S curve is a sigmoid function. It is the basis for the Gompertz function which can be used to predict new technology adoption and is related to the Bass Model.

Such curves suggest a longish period of low adoption, followed by an inflection point leading to rapid adoption.

That leads supporters to overestimate early adoption and vastly underestimate later adoption. Mobile phone adoption, and smart phone adoption, illustrate the process. You might think adoption is a linear process. In fact, it tends to be non-linear.

Also, the more fundamental the change, the longer to reach mass adoption. Highly-useful “point technologies” such as telephones, electricity, mobile phones, smart phones, the internet and so forth can easily take a decade to reach 10-percent adoption. Adoption by 40 percent of people can take another decade to 15 years. And adoption by more than 40 percent of people can take another decade to 15 years. 

source: MIT Technology Review 

That suggests a 30-year adoption cycle for a specific innovation that has high value to be used by 40 percent to 70 percent of people. Something such as metaverse, which is far more complicated, could easily take 30 years to reach 40 percent of people in ordinary use. 

That might mean at least a decade before metaverse apps are in common use by 10 percent of people. Even then, use cases are likely to be dominated by gaming, business communications and video entertainment. 

source: Robert Patterson 

The sigmoid function arguably is among the most-important mathematical expressions one ever encounters in the telecom, application and device businesses. It applies to business strategy overall, new product development, strategy for legacy businesses, customer adoption rates, marketing messages and  capital deployment, for example. 

The sigmoid function applies to startups as well as incumbents; software and hardware; products and services; new and legacy lines of business. 

source: Innospective

The concept has been applied to technology adoption in the notion of crossing the chasm of value any technology represents for different users. Mainstream users have different values than early adopters, so value propositions must be adjusted as any new technology product exhausts the market of early adopters. Early adopters can tolerate bugs, workarounds or incomplete on-boarding and support experiences. They tend to be price insensitive. 

It always takes longer than one expects for a major new innovation to become ubiquitous. Metaverse, being a complicated development, might take longer than any point innovation.