AI: 101 (When, why, and what the hell?)
AI is going to change the world. It’s bigger than the internet. All of our jobs will disappear.
And so, the headlines continue. Everybody who’s anybody has made a meaningful quote about AI, and every technology business has jumped on the AI bandwagon with the same ready-or-not alacrity they embraced delivering cybersecurity services.
But you’ll have to excuse us if we’re going to take a bit more time to think about this. Because AI poses significant new challenges and complexities, we want to take the time to process the implications, not just pick it up and run with it.
If you’re feeling equally cautious about AI, you are not alone. In their (well-worth-a-read) feature article from April 2024, “Despite the Buzz, Executives Proceed Cautiously With AI,” Reworked raises the same concerns and cautions.
So, backing up a bit, let’s take a 101 approach to AI and start at the beginning.
What does AI even mean?
We all know that AI stands for Artificial intelligence. It’s a term from 1955 coined in a proposal for a two-month, 10-man study of artificial intelligence. The ‘AI’ workshop took place a year later, in July and August 1956, which is generally considered its official birthdate.
Today, AI describes the simulation of human intelligence processes by machines (mainly computers), and can be seen in expert systems, natural language processing (NLP), speech recognition, and machine vision. To note: AI is frequently confused, by vendors and users alike, with machine learning (aka ML). But whereas AI mimics human intelligence, machine learning identifies patterns and then uses that information to teach a machine how to perform specific tasks and produce accurate results.
So, how does AI work? Basically, AI systems ingest large amounts of labelled training data. AI analyses the data, identifies relationships and patterns within it, and uses what it learns to make predictions about future states. Much as a human brain will access everything it knows at any given point and make a (hopefully) rational and informed decision about what happens next, so will AI.
What do you need to run AI?
AI needs three things to work: 1. data – and lots of it; 2. specialised hardware; and 3. specialised software.
Let’s talk about the importance of hardware, though, as without access to this, you have nothing. What do you need to know? The process of using a trained AI model to make predictions and decisions based on new data is called AI inference. While you can, at a pinch, run AI inference tasks on a well-optimised CPU, you really need the parallel processing grunt power of a GPU (graphics processing unit) for the compute-intensive task of AI training.
GPUs play an important role in data centres as they deliver the performance needed to accelerate AI and ML tasks, facilitate video and graphics processing, and run scientific computing and simulation applications.
Given the importance of AI training, it’s probably no surprise that leading vendors have advised that the demand for high-end, AI-ready GPUs has exceeded supply. The wait time for the average buyer can exceed 260 working days – which is over a year.
(The silver lining is that you have more time to define your AI strategy rather than rushing in).
Where are we in the AI hype cycle?
Gartner uses ‘Hype Cycle’ to describe “innovations and techniques that offer significant and even transformational benefits while also addressing the limitations and risks of fallible systems.” So, what’s hot now, what’s coming, and when can we expect these innovations to become mainstream – or fail to follow through on their promise?
In looking at Gartner’s 2023 AI Hype Cycle for AI, they call out two kinds of GenAI innovations that dominate. Firstly, the innovations that will be fuelled by GenAI, and impact content discovery, creation, authenticity and regulations, as well as automate human work and the customer and employee experience.
The next is innovations that will fuel GenAI advancements. This includes using AI to build new GenAI tools and applications. In effect, it’s using innovation to create more innovation – so it’s a popular use case for business startups.
The hype cycle illustrates user expectations of AI and how and where it will be used against where Gartner sees these innovations in 2-10 years (see graph). So, while the ‘innovation trigger’ and ‘peak of inflated expectations’ are crammed full of use cases and solutions, some will fall by the wayside while others will surface and go on to be productive.
The problem being?
While the list of potential use cases is exciting, Garter’s Hype Cycle does show that AI isn’t going to deliver business value (or even last the distance) in every instance.
Yes, you may get a head start on the competition by being an early adopter, but you may also become that cautionary tale shared in hushed tones in GenAI blogs and headlines of the future.
Forbes certainly reached that same conclusion in its article, ‘AI Reality Check: Why Data Is The Key To Breaking The Hype Cycle.’ Here, Forbes discusses whether GenAI reached its ‘peak of inflated expectations’ in August 2023, and many companies then came face-to-face with the reality of extracting genuine and meaningful value from AI.
Earlier, we listed access to data as a must-have for AI to work. And in its article, Forbes agrees, referring to its research, which firmly points the accusing finger of failure and disappointment at data silos. Nearly 75% of respondents who had implemented AI pilot projects in their organisations said that data silos were the primary barrier to enterprise-wide AI integration. “The number one thing keeping GenAI initiatives from reaching their fullest potential inside large corporations,” says Forbes, “is data.”
What the hell are we going to do with AI, anyway?
Worryingly, ADAPT’s CIO Edge Survey from February 2024 says that 66% of Australian CFOs say their organisations are unprepared to harness AI. 25% are non-committal, and only 9% say they’re AI-ready. AI-ready or not, 48% of the CIOs surveyed say they haven’t even defined any clear use cases for AI.
This leaves us asking, where to from here? How do you ensure that your investment in AI not only delivers business value through the availability of data, hardware, and software but that you are ready to use it and can justify the investment?
In part two of this topic, we discuss some real-world use cases in action and suggest some of the hard questions you should consider before committing to the shiny new thing that is AI.
