In a nutshell: despite their rapid initial popularity, AI chat tools like ChatGPT that use LLMs (large language models) will follow the same hype cycle and adoption curve of any new technology. Read on to balance the hype with a bit of technology wisdom and a touch of humour.
True technology adoption moves surprisingly slowly
When was the last time a new technology revolutionized everything within days, weeks, months, or even a year or two of being introduced? What are the new technologies we’ve recently seen that we expected to have massive changes?
People love pointing to the iPhone’s impact on society. But the iPhone took many years to reach massive sales numbers. It also represented decades of evolution in mobile computing technology, touchscreen technology, etc, relying upon the penetration of reasonably fast and high-bandwidth mobile data coverage to enable the multiplicity of apps we all now use. Apple introduced the first iPhone in 2007 and sales hit their peak in 2020. The first iPhone was basically a touchscreen phone, and it took over a decade for most websites to go mobile and for the critical mass of apps and services to become mobile-friendly. This means this new technology took 13 years to reach full adoption.
What about something more recent, like blockchain?
Today, in 2023, Bitcoin is 14 years old. But while we are still discovering valuable uses of blockchain, the technology has not revolutionized our daily lives. Only about 4% of the global population has participated in cryptocurrency ownership so far, with high concentrations of ownership. About 0.1% of Bitcoin miners controlled 50% of mining capacity according to a 2021 report. A significant amount of cryptocurrency activity supports illegal activities on black market websites, rather than supporting mainstream commerce.
AI chat tools will encounter challenges, many of their own making
The main point here is that chat bots that leverage LLMs will not significantly change most of our lives or our work immediately. Instead, over time, we will find ways to systematically take advantage of these new technologies, and the technologies will evolve significantly.
Generally speaking, we can expect ChatGPT and its ilk to follow Gartner’s technology hype lifecycle. While we hear in the news about examples of chat-based LLMs helping this person or that team increase their productivity by X percent, it may take many years to see systematic uptake. This is especially true as the weaknesses, threats and requisite regulation of the technology are brought to bear – a trend we’re already starting to see.
- The Biden Administration is seeking input on AI safety measures.
- Italy has banned ChatGPT.
- Major banks and other companies are banning or restricting use of ChatGPT.
- New stories emerge daily about generated lies and inaccuracies, like ‘hallucitations’, where LLM chat tools cite made-up news stories. In some cases, these hallucitations (paywall) impugn the sterling reputations of public figures through defamatory false accusations.
This growing body of problems and bans, along with the general unknowns about the new technology, will lead to a slowdown in usage, or at least in the wide-scale official adoption of the technology.
This doesn’t mean we should ignore it, but rather that we should treat the coming couple of years as a time of testing and exploration, as we do with any new technology.
We should be asking a few key questions about AI chat
There are several broader questions we should be asking around AI and other new technology.
First, what is the existing data and technology at work that we are not yet taking advantage of? For example, Microsoft’s Copilot tool is marketed as marrying LLM AI with the work social graph (a mapping of relationship connections among a group of people). But most companies and technology providers have massively underutilized existing work social graphs over the last 15 years, leaving tons of value on the table. For example, we’ve had the technology to map social connections and apply related filters to finding/delivering more relevant information but have rarely executed on this capability. We didn’t need LLMs to leverage this information, so why are we starting now? And will there be real value here, or is Microsoft just riding the LLM hype wave?
Second, yes, we all should be duly suspicious of artificial intelligence. The main reason is that very few people actually understand how any particular instance of it works – and this is always dangerous. The second reason is that AI and machine learning tend to reinforce the biases and mistakes inherent in their learning data sets.
Third, with all new technology we should be asking how it will help humans become better humans, and how we can design the technology to augment human ingenuity rather than replace it. Just like manufacturing machines and email, AI chat will mostly take over manual human effort but will typically still need expert human oversight. For example, a friend of mine recently used ChatGPT to create an API script. ChatGPT wrote the basic code for him, but he needed to provide the correct initial prompt, which was rooted in a deep understanding of the programming model and the business needs. He also had to review the script and make edits. ChatGPT just did the boring work of writing lines of elementary code.
Fourth, all new technology is subject to human nature. Some people will try to use it for good, others for evil. Some will try to enact violent acts with it, while most of us will try to make money from it. How, through regulation, can we push the technology towards doing more good than harm, helping the masses and not just the well-heeled and early adopters?
Let’s all take a breath, do it right, and enjoy the ride
Final thought: how do we have fun with it? We are all part of this grand adventure, an enormous group of conscious animals, living on earth, a planet that sustains life against infinite odds, spinning through the universe, smaller than a speck of sand on a beach. We get to be part of this unfolding and unpredictable drama of human life, technology and the natural world. Let’s try to enjoy this ride, and the quirky new technologies that pop up as single notes in the ongoing and incomprehensible symphony of the universe.
For example, a friend of mine won a local chilli cook off by asking ChatGPT for a recipe that would win a chilli cook off. That’s ingenious – and hilarious.
Talk to us about…
Categorised in: Artificial intelligence and automation