We’ve all heard the news: AI will be here in 2010! No wait, 2015! Absolutely by 2020… right? As it turns out, 2023 is the year, and now there’s no doubt that we’re fully living in a world where concepts like AI, ML, and ChatGPT are a part of our lives. These ideas can be anything from scary to exciting, sparking optimism and pessimism across the world. How do you even begin to piece together your point of view on AI?
Take a deep breath — AI isn’t here to replace your job today, but can already help you do your job better. To better understand the what and why of AI in this particular moment, I sat down for a chat with Contentful co-founder and Chief Technology Officer Paolo Negri. He shared with me his take on the latest developments and implications of this far-reaching technology.
The realm of NOW
Now that advancements in cloud computing have made AI-based software a practical reality, we’re already starting to see a new wave of innovation. Models, or systems trained to generate AI content, are getting introduced daily. Software built on these models, from sites to chat with AI to tools to automatically generate video, audio, and text, is appearing all over the place. It can be hard to keep up with how quickly industries are innovating and businesses are forming.
According to Paolo, “AI isn’t a replacement for humans, but can help us solve real problems we face with creativity and productivity.” Let’s take the blank page problem. Think of any time you’ve had to write — you start with a blank page, struggle to turn your ideas and topics into an outline, and stress until the article is done. The focus becomes getting the work done, and your energy and time are lost just getting started.
With AI, the blank page problem becomes a lot less impactful. “Rather than just staring at the page, you can now feed your topics and ideas to technology like ChatGPT, and get back content that you can use to spark your creativity,” Paolo said. “AI doesn’t replace the work of creating — it frees your creativity and gives you momentum.”
So what is AI? What about ML?
Let's get into some definitions. I asked Paolo to give us a helpful rundown of all of the buzzwords we’re hearing about AI. Let’s start there — “AI, or artificial intelligence, refers to the idea that computers can use their knowledge to replicate human intelligence,” he said. “Machine learning, or ML, is a type of artificial intelligence which allows the computer to use rules a human defines to learn faster and teach itself without constant intervention.”
You may also hear a term called NLP, natural language processing. AI can use NLP to interpret human speech and text and translate it into instructions that the computer can follow. This is how tools like Amazon Alexa work, using NLP to interpret your commands and turn them into instructions that a smart speaker can understand.
Going behind the scenes, AI’s knowledge comes from a dataset powered by a model. “A dataset is the information that AI references to learn,” Paolo said. “The model uses an algorithm, or a set of rules, to tell the computer how to navigate the dataset.” Humans write the rules, they provide the data, and they decide how an AI processes information. This is what keeps the intelligence artificial, and can also introduce bias or inaccuracies that make the results potentially untrustworthy.
If you’ve spent any time with AI news, then you’ve also heard names like OpenAI and ChatGPT. What are these, what do they do, and why do they matter?
“OpenAI and ChatGPT are at the forefront of this current AI movement,” Paolo continued. “Alongside hundreds of new companies, products, and technologies, ChatGPT has allowed more people than ever to actually use AI.”
ChatGPT, as the name suggests, is an AI that anyone can “chat” with. You can ask questions, develop complex prompts, or even ask for lyrics and poems, and the software interprets your prompt and delivers a response in near real time. ChatGPT has many versions — most recently GPT-4 — each with varying quality, accuracy, and ability. The latest versions can even interpret video and image content!
How does it all work?
Let’s look at a tool like traditional search. “Traditional search tools weren’t based on AI. They took the search you made and broke it down into individual keywords. The search engine then scanned its index of keywords to find the content most likely to be relevant,” Paolo explained.
Let’s use an example search — “How far can an emu run in a day?” This is a simple question at first glance, but is a hard question for traditional search engines to answer because most content will refer to how fast, and not how far, an emu can run. You might see a few links to pages that mention a lot of the keywords, but the search isn’t actually interpreting the question to get the right answer.
With AI products like ChatGPT, a computer can interpret your question differently. Paolo walked me through the process: “The power of AI technology like ChatGPT is that it breaks your prompts into individual words, and replies with other words that are highly related based on its dataset,” he said. This means that this version of AI can better interpret the spirit of your question, and can lead to more relevant answers.
Let’s run the same question (“How far can an emu run in a day?”) through ChatGPT. It takes each word in the question, and uses its wide dataset (i.e. “vast amounts of data from the internet written by humans”) to look for patterns, build knowledge about the question, and connect information, much like a human would. What results is a generally more helpful answer, but one that may be completely incorrect if the AI selects the wrong relevant content.
“AI offers a different way to interpret information that we provide. It’s not a replacement for human intelligence, and still relies heavily on humans who write the rules and programs,” Paolo said. “The potentially concerning parts of AI/ML come from how systems can be designed for bias or can produce hallucinations — answers that appear factual but are actually incorrect — or can use copyrighted content to produce its answers.”
Otherwise, AI is a novel tool, and not a solution, for consuming and interpreting information. It helps take complex and large datasets like the internet and interpret them more like a human would, providing back quick answers that more naturally fit how humans think. Humans still need to think critically about these tools, questioning the dataset, the model, the rules, and even the output before accepting anything as fact.
Using AI to enhance your business?
So you’re ready to go run your business through ChatGPT, right? You may need to rethink your plan before you start, and look for ways for AI to augment, and not replace, how your business works.
“We’re really seeing a moment where jobs and companies across the board can benefit from the AI tools available today,” said Paolo. “Writing and editing teams can already use tools like Jasper.ai and Writer to generate on-brand ideas and content. Creative teams can use AI image and branding software to generate specific images, use AI to transform and redesign, and even ensure assets meet brand guidelines automatically. Data teams can enhance their analysis skills with AI support to get to valuable insights faster (and with less code).”
One of the more exciting opportunities highlighted by Paolo is the use of AI in the product development process. “AI tools like GitHub Copilot can help software engineers generate code, transform code into other languages, produce complex content like API specs automatically, or even help engineers test and improve their code,” he said. AI isn’t meant to replace your engineers or do the work for you, but instead to save time, simplify, and undertake operations that the human mind simply can’t.
Let’s put it in practical terms — engineers can spend less time (and money) on solving their own blank page problem, and spend more time doing what they do best — solving customer problems. AI becomes a partner to provide the hints, optimizations, and automations that free up mindshare to focus on important tasks.
Tips for responsible AI use
While generative AI technologies have the potential to boost productivity for your company, there are some important steps you should take to make sure you don’t expose yourself and your customers to unnecessary risks.
If you want to use any new AI tool – even a free version of the tool – it’s a good idea to speak with your manager or legal team. If they approve it, initiate your company’s standard software procurement process to ensure you are using AI responsibly. Why should you be so cautious? For several reasons:
Vendors may impose terms and controls differently between free and enterprise level accounts. Using free or personal tier accounts on your company systems may inadvertently put your data and the data of your customers at risk.
AI-generated results that originate from “publicly available sources,” including open source code, might be protected under copyright, trademark, or privacy laws, or may include license restrictions.
Some AI tools may have broad access to your company data (e.g., code repositories, Slack chats, etc.) under certain contexts which must be carefully assessed during the software procurement process. They may collect or store your company data inappropriately, or even use it to fine tune results that expose your confidential information and/or intellectual property.
Company information which may not seem sensitive on its own could be used in conjunction with other data to provide results unfavorable to your business.
The conversation about artificial intelligence and content creation doesn’t begin and end in a single blog post. We’ve only just started!
Stay tuned for more perspectives and updates from Contentful about how we will help you prepare for — and respond to — this fascinating field.