As I mentioned a few times in this blog, I’ve been working with and teaching about artificial intelligence since the start of my career. This drove much of my interest in cloud computing because AI was not economically viable or accessible until “the cloud” came along.
Interest in AI and its applications inflected about five years ago. Then the pandemic happened and some budgets shifted to speedy cloud migrations. Now that things are returning to normal, AI is back. Most enterprises grasp the fundamental possibilities of AI and are looking to weaponize the technology for their own business.
The technology got way more impressive along the way. Generative AI, for example, went from PhD dissertations to an accessible and free reality with the advent of generative AI services such as ChatGPT.
Generative AI is a type of artificial intelligence that generates new and unique outputs, such as text, images, or audio based on input data and learned patterns. This can include tasks such as text generation, image synthesis, and music composition.
A wide variety of inputs can be made through a chatbot or an API that results in an impressive response. The responses are so impressive that I’ve been fielding calls from journalists who are writing stories on AI replacing workers. It’s a question I’ve heard for the past 20 years but now there’s a modern twist. Colleges and universities have new concerns about college students using ChatGPT or similar services to write papers for them. AI creates output that plagiarism detection systems can’t quickly identify because it’s not plagiarism.
AI ethical and bias concerns could emerge from certain types of learning data. Will those biases cause unintended negative consequences, such as an automated pattern that denies loans to certain groups of people?
I’m hearing the following core questions: What types of human tasks can AI replace now or soon? Should I be planning a career change to a job that AI can’t automate? Is it safe to be a cloud computing architect, cloud developer, cloud operations engineer, devops engineer, cloud project leader, etc.? Those are the job titles of most of you reading this article. Are you at risk?
I think the reality is we’re well on our way to replacing many human tasks with AI-fueled automation. It’s just what happens as technology progresses, and it’s nothing new. Technology is why we no longer have dozens of people in a field to pick crops in the fall. We can check out of the supermarket without interacting with a human. Our cars and trucks can drive themselves.
One thing that I’ve been frustrated with is how the whole IT design and deployment process lacks helpful automation. Oh sure, we have tons of tools, processes, methodologies, and other assets to speed up our journey to an optimized cloud architecture and deployment. However, they don’t make critical decisions for the architects. Cloud architecture must typically be determined through deep analysis and judgment, which only come through experience. More importantly, creativity and innovation are still required—things humans bring to the table.
Of course, humans make many architectural mistakes, such as picking the wrong platform, tools, and services. Humans create architectures that are wholly underoptimized and fail to return value to the business. I addressed that issue recently.
If we turned solution creation over to AI, perhaps we’d get better decisions. Imagine if the AI system had training data that simultaneously reflected the knowledge of thousands of good cloud architects. Such an AI system could effectively process that knowledge into solutions based on provided business and technology requirements. It may not get you the final answers required to build something, but it could get close enough that it could remove a great deal of the work and potential mistakes.
The most likely path is that tactical AI tools will continue to appear. These tools will focus on specific architectural areas, such as network design, database design, platform selection, cloud-native design, security, governance, use of containers, etc. The output should be as good as, if not better than what we see today because these tools will leverage almost perfect data and won’t have those pesky human frailties that drive some architecture designs—emotions and feelings. Of course, some of these AI tools exist today (don’t tell me about your tool) and are progressing toward this ideal. However, their usefulness varies depending on the task.
Tactical AI tools must still be operated by knowledgeable people who understand how to ask the right questions and validate the designs and recommendations the tool produces. Although it may take fewer people to pull off the tactical component design of a large cloud architecture, the process will not likely eliminate all humans. Remember, many of these mistakes occur because enterprises have difficulty finding skilled cloud pros. Tactical AI tools may also help fix that problem by better syncing the demand and supply of talent.
It’s easy to predict how this will progress, and it’s nothing earth-shattering. Design, development, and deployment tools will evolve. They will provide more value and usefulness. Fewer people may be needed overall, but these tools will require talented operators to work correctly. They will focus primarily on the tactical design of cloud architecture components, such as network and security.
So, we still must rely on humans, such as yours truly, to build sound cloud solutions. The smaller problems may have AI solutions, but the larger problem of poetically automating cloud architectures remains a problem not yet solved.
I suspect we’ll still need human cloud architects and solution designers for some time to pull all of this together and ensure that we’re deploying optimized solutions that return the most value to the business. I don’t see that being something that can be replaced entirely by AI, but I’m under no illusion that it’s unlikely forever.
But let’s ask ChatGPT: “Will AI replace cloud computing architects?”
I agree.
