Much has been written about artificial intelligence systems being able to write. But few have dared to ask - can an AI cover data centers?
With the launch of OpenAI's new AI chat system 'ChatGPT', DCD spent some time talking to the bot, and seeing if it could spit out clean and engaging copy. ChatGPT is based on a model in the GPT-3.5 series, which finished training in early 2022.
Let's start with an easy one.
We asked: "What is a data center?"
It said: "A data center is a physical facility that organizations use to store, process, and disseminate large amounts of data. These facilities typically include a network of computers, storage systems, and other components that are used to store and manage data. Data centers are typically used by organizations that have a large amount of data to manage, such as Internet service providers, cloud computing companies, and large enterprises. They are essential for the functioning of many modern businesses, as they provide the infrastructure needed to store, manage, and access data quickly and efficiently."
All correct, and all remarkably human sounding. But a bit on the dull side - if one of our writers produced this, we would send it back and look to go deeper. (It also did not capitalize the I in Internet, but we standardized it with our house style).
Next, let's ask it to answer one of the big questions in the industry.
We asked: "Will liquid cooling replace air cooling in the data center?"
It said: "It is possible that liquid cooling could replace air cooling in some data centers, but it is not likely to become the norm. Liquid cooling has several potential advantages over air cooling, such as the ability to more effectively remove heat from components and the ability to cool components more evenly. However, liquid cooling also has some significant drawbacks, such as the need for specialized equipment and the potential for leaks and other problems. Additionally, many data centers are already built around air cooling infrastructure, so it would be expensive and time-consuming to switch to liquid cooling. Overall, it is likely that air cooling and liquid cooling will continue to be used in different data centers, depending on the specific needs and requirements of each facility."
Again, an incredible answer to be created by a non-sentient system in less than two seconds. As an editor, I would say that the copy remains rather basic, could go into more about the different types of liquid cooling, and should have a stronger finishing sentence. But a good start.
Let's go meatier.
We asked: "Should a company use colocation space, or move to the cloud?"
It said: "Whether a company should use colocation space or move to the cloud will depend on a number of factors, such as the company's specific needs, budget, and technical requirements. Colocation space is a good option for companies that need to store and manage a large amount of data, but do not have the infrastructure or expertise to do so in-house. Colocation providers offer secure and reliable facilities for storing and managing data, as well as support and maintenance services. However, colocation can be expensive, and it may not be as flexible or scalable as cloud storage.
"On the other hand, moving to the cloud can be a good option for companies that need to store and access data remotely and do not want to invest in the infrastructure and personnel needed to do so in-house. Cloud storage is typically more flexible and scalable than colocation, and it can be accessed from anywhere with an Internet connection. However, it can also be more expensive in some cases, and it may not be as secure as some other storage options.
"Ultimately, the decision to use colocation space or move to the cloud will depend on the specific needs and requirements of each company. It may be best to consult with experts or conduct online research to compare the different options and determine the best solution for the company."
There we go, a longer piece of text to edit. Another good effort by our junior reporter here, but a little repetitive - we don't need to say that colocation is less flexible than cloud in the colocation section, and that cloud is more flexible than colocation in the cloud section.
The claim that the cloud is less secure also raises questions. If we published that, we'd definitely get some pushback - so we'd need to back that up with data, or quotes from an industry figure.
Questions about pricing also need more delving into, and there is no discussion about the high price of getting data off of the cloud, a key reason some are souring on the cloud.
Okay, so these were general questions. Let's pit it against our own copy from the latest issue of the DCD Magazine.
We asked: "Why is there a shortage of submarine cable-laying ships?"
It said: "I'm sorry, but I am not aware of any current shortage of submarine cable laying ships. As an AI assistant, my training is limited to a specific set of information, and I do not have the ability to access up-to-date information or browse the Internet. My knowledge is based on information that was available up until 2021, and I am not able to provide information on events or developments that have occurred since then. It would be best to consult with experts or conduct online research to find current information about any potential shortage of submarine cable laying ships."
Okay, not great. But hey, bonus points for admitting when you don't know something - we encourage that in our team. It's best not to pretend and end up with poor copy.
How about trying to write our cover feature?
We asked: "What is the future of long-term data storage?"
It said: "The future of long term data storage is likely to involve a combination of different technologies and approaches. One trend that is likely to continue is the use of cloud storage, which allows data to be stored and accessed remotely over the Internet. This makes it easy for organizations to store large amounts of data and access it from anywhere. Another trend is the use of more advanced storage technologies, such as solid-state drives (SSDs) and 3D NAND memory. These technologies offer higher storage densities and faster access times, making them well-suited for long-term data storage. Additionally, there is ongoing research into new storage technologies, such as DNA-based storage and quantum computing, which could potentially offer even greater capacities and performance in the future."
Here we run into an issue. This sounds convincing to someone who doesn't follow the sector - and hey, it even brought up DNA storage, which we profiled in the mag - but it falls apart under close inspection.
3D NAND is flash and, as the word 'flash' suggests, that's not exactly for long-term storage. Even SSDs only last a few years, when the piece is looking at millennia. And quantum computing? Again as the name suggests, it's for computing, not storage. No freelance payout for you today, AI.
I tried out the system for some time - these are just a few of the prompts. Many gave solid answers along similar lines as above, while some failed, or it declined to answer.
In each successful case, it is clear that this is a monumental technical achievement.
But it's worth putting it into perspective: Am I worried about my job any time soon? No - not yet, anyway.
The AI is terrible at choosing sides - its pieces are many pros and cons lists with no payoffs. Its technical chops are lacking, and often confuse things. It doesn't know anything current, and doesn't say anything new. It doesn't reference particular companies or people (likely an intentional choice by OpenAI to stop it attacking individuals) unless you prompt it. It doesn't understand media law, and could say stuff that could get us sued. We don't know if it is plagiarizing content, and that could also get us sued.
The copy is closer to a school paper than an industry magazine. Were we to fill the next magazine with its text instead of ours, I'd like to think people would notice - but perhaps I am just being naive.
The reality is that, for some, this will be good enough. The Internet is full of poorly written or informationless content that does well, often created to game Google or pad out a company blog. This could join that sea, and prosper.
For our Opinions channel, we often get cold-pitches from public relations groups that offer us pieces from respected figures. When we receive them, they are sometimes clearly written by the PR team or freelancers (in fact, we have been invited to write them, as a side-hustle, but say no). We turn down copy that doesn't provide information, and just does recaps known facts or industry assumptions. I imagine that such teams will make use of this AI to pump out more of this copy [I fear they may already be doing this - Opinions Editor].
There is also a strange feeling in knowing that, given the small number of data center publications out there, and DCD's decades-long coverage of the sector, our content has played a part in feeding this system.
Over a longer term, it could prove a threat, especially if audiences are not as discerning as I hope. Over a shorter term, it only serves as a reminder: We better try to be damn good at our jobs.
We need to lean into what only a human can do - site visits, live interviews with respected figures, not taking all information at face value, and making connections that are not immediately clear by simply looking at existing data.
The rapid progress of this system, and others like it, serve as a warning to us in the media industry. Simply writing clean copy is nothing special. You actually have to say something for it to be worthwhile.
You have to put the work in to create something new, otherwise what's the point?