Have you heard about ChatGPT?
That’s a rhetorical question. Of course you’ve heard about ChatGPT. Every media site on the planet has written multiple stories about ChatGPT at this point. Most of them have used ChatGPT to help write some of those stories (full disclosure: ChatGPT was not used in the writing of this article). After some opening acts featuring artificial intelligence (AI) image generators like DALL-E 2, Stable Diffusion and Midjourney, the text-based ChatGPT took the world by storm at the end of last year.
Even though everyone knows what ChatGPT is by now, here’s a quick summary to help set the stage for some further analysis. ChatGPT is a product from OpenAI, which also delivered DALL-E 2. These tools, along with many others that have been released or are currently being developed, use new programming models based on the “transformer” algorithm released by Google in 2017. This algorithm delivered a huge breakthrough in deciphering text, which led to the general class of AI models known as large language models (LLMs).
By building a LLM and training it on huge datasets available from the internet, companies are delivering products that define a revolutionary new wave of AI. Getting a computer to write has gotten even more attention than getting a computer to draw since there are far more applications for written communication. The whole thing has reignited the debate around robots taking over our jobs.
Asking the Right Questions
From a business perspective, the question on the mind of most decision makers is: “Where can we use this?” Some applications seem obvious; anywhere text is being generated is a tantalizing candidate for tools like ChatGPT. Then there are other applications that might not currently be centered on text but could be re-engineered to use the underlying algorithms.
The better question, though, might be: “How will we use this?” Artificial intelligence isn’t the type of technology that can be purchased as a standalone product and plugged into a data center. Along with powerful capabilities, AI has unique characteristics that will impact any implementation. As IT pros evaluate opportunities for their organizations, they should understand these characteristics and provide context to the decision-making process.
3 Characteristics of Artificial Intelligence
- Like most emerging technologies, AI is an enabling technology. Instead of providing value on its own, it provides value as a component of other applications. As an example, think of something most people have seen already: Email. Most email applications today have an auto-complete feature, where an AI algorithm predicts and suggests the next words or phrases as someone is typing. An organization wouldn’t focus on the specifics of the AI in this case—it would focus on the overall functionality of the email program and the additional value that AI might provide.
- The most significant difference between AI and previous software is that AI produces outputs based on probability. Traditional software is deterministic. That means that a set of inputs will always produce the same output. AI is probabilistic. An algorithm receives inputs, then calculates an output based on combining those inputs with the data used for training. It’s basically giving a guess, and that guess can provide unexpected insight, but it can also be incorrect. Plans for using AI must include plans for overseeing the output to make sure mistakes are caught and corrected.
- AI is vulnerable to one of the sneakiest truths about technology: The best tech doesn’t always win. There can be many reasons for this, but as technology gets more widespread in business and society, one of the main reasons is the difficulty in changing behavior. Think back to the email example. When the program gives a guess about what comes next, the user has to consider the suggestion, decide if it’s appropriate and hit the tab key or something similar to accept the suggestion. That’s a lot different than just continuing to write. Is there enough improvement in the end product to justify the learning curve? The answer to that might not be clear in a product demo.
The Need for AI Skills
These three characteristics all factor into the big debate mentioned earlier: Will AI eliminate a wide range of existing job roles? Even though AI capabilities will continue to grow, it seems far more likely that AI will change a wide range of existing job roles. AI-based programs have the potential to heavily automate tasks or produce unique insights, but they might never get past the point of needing some degree of oversight (think of how self-driving cars seem to have hit a limit).
So, when we talk about “AI skills,” we’re not just talking about the ability to code an algorithm, build a statistical model or mine huge datasets. We’re talking about working alongside AI wherever it might be embedded in technology, whether that’s help desk ticketing systems, network routers or cybersecurity monitors. There will be growing demand for new AI specialists, but there will be much greater demand for AI expertise across the many tech roles that currently exist.
Stay up to date with CompTIA and tech skills by subscribing to CompTIA’s IT Career News. As a thank-you, we’ll give you 10% off your next purchase.