Gathering information with AI

Blog series “Technical documentation and the power of AI”

It is the talk of the day: Artificial Intelligence, also known as AI. Whether it’s at the coffee machine at work, or during dinner with the family – The rapid development of AI is what everyone is talking about. From the huge impact it already has on our lives to the undiscovered possibilities (and challenges) that lie ahead. How does this affect our work and daily life?

At Foxiz we’re full of questions and ideas. By now, AI is part of the job for our technical writers, illustrators, translators, and consultants. We would like to take you with us on our journey of discovery. The thing is, we want to tell you so much; one blog simply doesn’t suffice. That’s why we’ve written an entire series of blogs on the matter, in which we highlight the different aspects of our job, as well as the impact AI has. Follow along as we explore AI.

Nr. 2 – Gathering information with AI

As a technical writer it is essential to find reliable information on products or machines. When your job is to write about how a machine works and what the user or mechanic need to pay attention to, you must know what you’re talking about, yourself. How do you get that information? And can AI be of use in sourcing information?

Sources

The first source of information is, of course, the manufacturer. But what if they are unavailable, or you don’t speak the same language? As writer of the manual, it is important to be well informed: it is after all your job to explain how something works.

Looking up information on the internet is no easy job, especially not when it comes to technical data. There is a chance you get caught up in a search through all kinds of websites that may or may not provide up-to-date and reliable information. Filtering all this information is quite a time-consuming task.

Your search for information basically consists of two subtasks: 1) finding information and 2) checking information. AI can be a useful aid in both subtasks, as long as you implement it in a proper way. In this blog we’ll tell you all you need to know about these tasks. Finally, we’ll discuss reliability and safety: definitely not unimportant!

Finding information

You can converse with ChatGPT to obtain some general information and to explore concepts. With a bunch of smart prompts, you’ll quickly get an overview of the information that is available online. ChatGPT can also provide you with ideas on what else may interest your readers. It is, however, important to write your prompts properly. See the first blog in this series to learn more about this.

ChatGPT can lead you to tools and platforms that may prove useful to your search. You could, for example, end up on platforms with experts and scientists, or with tools for data analysis. It all depends on the kind of information you’re looking for. By now, there are quite a lot of AI tools. Which poses its own challenge: how to choose between all the possibilities? When choosing, there’s a few things to keep in mind. By far the most important ones being reliability and safety. That is why it is incredibly important to check the gathered information.

Checking information

Always be critical when it comes to the source of your information, even though AI presents it on a silver platter. Do not assume: ‘AI said so, that must mean it is true.’ Ask AI to provide sources of all provided information and check those thoroughly. Try to find as many primary sources and official documentation as possible. This way, you can be sure the information is of high quality.

Checking information is quite a time-consuming process, but luckily AI can help you check the reliability of sources. Below are three handy types of tools for this task:

1. Fact checking tools

There are specific AI tools designed for fact checking that can verify technical information using validated knowledge bases and technical databases. Wisecube AI Fact Checker and LEO AI are examples of such tools. They convert technical statements to ‘knowledge triplets’: a set consisting of a subject, predicate, and object that is compared to reliable sources. The perfect way to verify technical specifications and calculations.

1. Tools for source information analysis

These tools take a closer look at the metadata of former publications of a source. Examples are Ansys, Siemens Simcenter 3D and Autodesk. Which tool you use depends on your question and on how deep you want to dive into technology. Unfortunately, the tools can be pricey and, although they are quite extensive, they sometimes fail at processing and analysing complex structures. A tool for analysis is only as good as the information it was trained with. In other words, the tools may be powerful, but you need a lot of experience and money to use them effectively.

2. Tools for network analysis

These AI tools map sources and citations, giving you an overview of the spread of information and whether it comes from reliable channels. There’s the open-source tool Gephi, that you can use to analyse related technical documentation. Also check out UCINet, a tool used to gain insight into the collaboration between researchers in technical domains.

Tool reliability

You should not carelessly copy the information from the tools above. The effectiveness of reliability tools depends on how up to date their data is, as well as its quantity and quality. The tool will always present information as if it is 100% true, even though it might not be. AI can even ‘hallucinate’: it can wrongly interpret correct information and present faulty information it created itself as if it were true. Be aware of this. Lastly, not all tools are useful for all domains of information. Wisecube for example is especially useful when it comes to engineering, but less so when it comes to electronics and software.

In summary, AI can help to verify information, but human expertise remains indispensable. Always check information using reliable technical documentation.

Safety of AI-tools

How is data in tools secured? Who has access? Especially when working with sensitive information, it is good to know where the tool comes from and how safety and privacy are maintained. Don’t trust a tool based on a fancy logo and a website full of reassuring words. Check out the safety tips below:

1. Encryption

Consult the tool provider’s security documentation. Look for mentions of AES-256, TLS 1.3 or similar. Also keep an eye out for certifications, e.g. ISO 27001 or SOC 2. This way you know for sure that strict security measures have been taken for the tool.

2. Tools with sensible access management

Only use tools that grant access to users with certain functions (role based access). Also check whether the tool exploits API gateways. These ensure that communication with the service is kept secure.

3. Provider warranties

Read the privacy policy and terms of service. It should include that data will only be processed in encrypted form and will not be used for training. Always ask about the provider’s security policy.

What do you have to consider?

If you want to use AI tools in your search for reliable technical information, it is important to pay attention to the following:

Context awareness: AI does not always understand the nuances and context of technical subjects. Adapt your questions and prompts to account for this: clarify technical jargon or complex concepts and name your target audience. Who are you searching information for? What is their level of technical knowledge and what do they already know about your product?
Copyrights: AI plucks information from all texts ever written or created by someone. You’ll have to keep in mind possible copyrights. Also be aware that information may be confidential. That is why it is so important to check your sources!
Functionality: does the tool provide the information you need? Ask yourself exactly what kind of information you are looking for: technical explanations, data analysis, scientific literature, specifications, and so on. Also consider for what specific technical domain you need information and see if the tool is suitable for that.
User friendliness: how easy is the tool to use? Do you first have to read an entire manual or is it fairly intuitive? If figuring out how the tool works is going to take a lot of time, you may want to consider whether it adds value.
Accessibility: some tools offer full access to all the information immediately, but others offer only limited access or require you to buy a license first. Consider whether that investment is worthwhile.
Integration: can the tool ‘communicate’ with the documentation systems or other software you use? Easily transferable information makes your life easier. If not, you’ll still be up typing all night long…
Scalability: can the tool handle the amount of information you want to process? Check whether there is a data limit and how long it takes for the tool to process your data.
Actuality: not all data that AI-tools use is up to date. Check its sources and make sure that the websites still exist and get updated regularly.

Conclusion

AI offers powerful tools for finding and checking technical information, but you should always remain critical of the output. The human behind the machine is indispensable and must always ensure the quality, reliability, and safety of the data collected.

Foxiz and AI

At Foxiz we implement AI in a safe, smart, and effective way to aid our specialists with their work on technical documentation and translation. This way we create high quality technical content, combining the best of both worlds!

Are you in need of proper technical documentation for your product or system? Want to pick our minds on how to improve your documentation and its translation? Or are you interested in a job as technical documentalist? Please contact us! We are happy to help you.

TIP: read our last blog on writing smart prompts for AI . In our next blog we’ll answer the questions if and how AI can help you write an entire new manual. Until then!