Six AI Lessons From A Past Technology Wave

We are all aware that AI is the big topic of the last year plus. Some people will tell you this technology wave is different than those in the past. They may be right. But today, I want to take you for a stroll down memory lane a bit to see what lessons we can learn from a past wave of technology.

"History doesn't repeat itself, but it often rhymes." - Mark Twain, maybe

Our little stroll in search of rhymes takes us back to 2011. Do you remember the state of smartphones in 2011? It was WILD! Let me refresh your memory:

•  LTE and WiMAX (two 4G technologies) were released in the US to supplant the 3G technologies that helped push smartphones into the mainstream.
• The White iPhone 4 sent the internet into a frenzy before the iPhone 4S release smashed sales records.
• Windows Phone was not only a thing, but you could also pick from multiple devices.
• Nokia had phones with the Symbian operating system for sale.
• The Samsung Galaxy Note had a whopping 5.3" display and was considered so big by phone standards it was called a "phablet."
• There were two 3D-enabled phones, the HTC Evo 3D and the LG Optimus 3D, each approaching 3D differently.
• The Sony Experia Play had slide-out physical game controls.
• HP was even in on the act with the HP Veer running WebOS.
• There was also a bevy of Android phones from the likes of Samsung, HTC, Motorola, Sony, and LG.
• And Blackberry was still hanging on with the Blackberry Bold Series that it launched in 2008.

I was a Verizon customer at the time, and I dropped my hard-earned cash for the HTC Thunderbolt, an Android phone that had kickstand built into the back of it. It was also the first phone to use LTE on Verizon's new 4G network and I was giddy at getting download speeds of 20 megabits per second (Mbps) on a network that had virtually no traffic. Of course, that phone also had the worst battery life of any phone I've ever owned. I felt lucky to get 6 hours of life between charges. Let's just say it was a love/hate relationship. More on the Thunderbolt later.

These wild days of innovation, successes, and failures were not unlike today's Generative AI environment. Today, there are no shortages of AI models or applications employing them to change the way we approach all kinds of things in our business and home life. Technology waves happen repeatedly and while we're all riding this one, we can look back for some lessons.

The Blackberry Lesson

In 2010, the Blackberry platform owned 43% of the US smartphone market share. By 2012, it was down to 15%. Blackberry had built an empire during the first decade of the new millennium based on enterprise business users. Those users had specific security requirements and primarily used their phones for voice calls and email. But the iPhone made the smartphone attractive to the masses and the masses wanted something else in their smartphone. They wanted a touchscreen. They wanted to listen to their favorite music on demand. They wanted a good camera. And they wanted apps. Blackberry ignored the shift in the market when the smartphone went from a power-user luxury to ubiquitous.

AI was once the domain of large enterprises who could afford a team of data scientists. But a shift in the AI market has occurred. Today, AI is being used by enterprises, but also by businesses of all sizes, students, and entrepreneurs. You no longer need a team of data scientists to deploy AI. There will be lots of failures along the way, but the companies who choose not to figure out how to harness the power of AI will be left behind.

The Phablet Lesson

The Samsung Galaxy Note, the original "phablet" had a 5.3" display and was 5.8" tall. The iPhone 15 has a 6.1" display (measured diagonally) and is 5.8" tall. That's right, today's "average" phone is a phablet by 2011 standards. The Galaxy Note was a niche device that could perform the functions of a smartphone and replace your tablet. But as screen resolutions and touch input improved, data got faster, and technologies like swipe-typing emerged, our phones took over more and more the tasks we once relegated to tablets. Soon tablets became the niche devices while our phones continued to swell in size.

What seems like a niche use-case today may become the standard as technology evolves and our use of it evolves. Many implementations of AI will be ridiculed, some deservedly so. But the Galaxy Note is a reminder that what seems ridiculous today may feel very comfortable tomorrow, as long as we use bigger pockets!

The Android Lesson

Operating Systems are a great analogy to today's Large Language Models. You have big names behind closed-source models just like you had iOS, Blackberry OS, and Windows Phone. Then you had the Android Open Source Project (ASOP). Yes, most phones running Android today (and even in 2011) use Google's Android OS which is not open source, but ASOP was the base and it made building your own custom OS not only possible, but widespread. Fire OS, ColorOS, OxygenOS, OriginOS, MagicOS, CyanogenMod, and LineageOS, among others, were all developed using AOSP.

Today, there are dozens of LLM models, some of which are open-source and many others that are not. While many enterprises will partner to gain access to the closed-source models, many will create viable products using the open-source models as their starting point. Today, Google’s Android (based on open-source code) holds 70% of the global market share.

The 3D Lesson

2011 brought 4 new technologies to the world of mobile phones. While 4G, NFC, and fingerprint scanners were innovative solutions for existing problems in 2011, taking 3D images and video was clearly just a party trick. While 3D was certainly a differentiator in the market, the potential use cases were also edge cases that required you to use one of these specific phones.

When considering how and where to deploy AI, focus on solving common challenges that occur with frequency. Doing so greatly increases the likelihood of widespread adoption. Edge case-based solutions will likely be viewed as a gimmick and those tend to flame out quickly.

The Windows Lesson

For the briefest of moments, Microsoft had a mobile OS that rivaled iOS in both beauty and performance. It began using hardware partners much like Google did with Android. But then it decided to walk along two paths. Microsoft bought Nokia to be its in-house hardware team. Yet they still wanted to also license to other hardware partners. I know what you’re thinking – Google does this today and they’re still successful with it. The difference is that Nokia was king of the economy market and the team Microsoft acquired wanted to focus on the lower price-point market. Meanwhile Microsoft executives were keen to take down (or at least match) Apple in the premium market. Those differing goals led to internal strife that eventually spilled out into the public.

When your team has a solid understanding of who they are and what their goals are, you have a much better chance at success no matter what you're trying to do. If your AI ambitions are not entirely clear, the internal struggles could rival the technology struggles.

The Thunderbolt Lesson

This last one is quite personal to me as you might guess. That phone got the tech nerd in me so excited. The Thunderbolt featured a 1 Ghz Qualcomm Snapdragon processor, 4.3" WVGA display, 8-megapixel camera - all pretty pedestrian in 2011. But the LTE connectivity - that was anything but pedestrian! Coming from 3G loading times on Verizon's crowded network to their new 4G network that only The Thunderbolt could access felt like hopping in a Ferrari and driving full speed on the autobahn with no other vehicles around to get in the way. But it was also like the Ferrari only had a 3-gallon gas tank. That early LTE radio sucked the life out of the battery. You could drain it in about 2 hours if you hopped on YouTube. Ironically, my daily needs didn't need 4G speeds. I was not a power user in any way. A pedestrian 4-cylinder Toyota was more than enough for my daily needs and more aligned with my budget.

As each new Large Language Model touts its billions (probably trillions by the time you read this) of parameters, keep in mind what you're using it for. Smaller models may be more than sufficient to accomplish your needs. Smaller models require less RAM and less computing power to do their job, so they're less expensive to use. If you just want a call summary for a 5-minute call to be pushed to a CRM, you don't need the biggest, fastest LLM on the market to do it. And if you’re leaning on technology partners for your AI strategy, they, too, should be using the right technology for the right tasks!