Hello, everyone. My name is Tommy. I'm a technical writer at Semaphore, and today I'm going to talk to you about technical writing in the era of AI. If you have any questions, here's my contact information. With that said, let's jump to the talk. I'm going to try to answer the question, is technical writing a worthy career path to pursuit? I think the answer is yes. This is short answer, and the long answer is, it depends. It depends on what type of writing we do and how much we are willing to learn and adapt. The world has had many technological leaps over the centuries. At every stage there have been people who cried they would destroy jobs. And to some extent it's been true. We don't have the typesetter job anymore. But the bottom line is that new technologies have created more jobs and given voice to more people. My question is, is AI just another leap? I don't think so. I think ais and language models are a new class of things. They're a new type of things. What makes AI different is that they can produce text at scale and they can kind of think so. I will use the term AI and language model interchangeably, even though they are not technically the same. So technical writer is not immune. There are a ton of new products dedicated to automating that skill. These are a few examples. They can take code and output documentation. We even have Synthesia that can produce talks with a virtual avatar, so even that can be automated. And the future is kind of grim, at least in the numbers for forecast. This is Forester's 2023 forecast. And in the top right corner, there's technical writer. So we are very much exposed to AI influence. So should we give into fear? I don't think so. Let's pause and think through. I believe there are many reasons why technical writers will be around for a while. I have come up with five of these reasons, and we'll talk about them now. So, reason one is the hype cycle. This is how things appear to be according to the news. There's like an explosion first, then there's a drop in the hype as things began to fail and technology does not fulfill the proposes initial promise. And finally, there's a plateau where things are more balanced and we know the limitations. I think in crypto we already passed the hype cycle. AI is just starting. At least language models are just starting. So we have to be mindful of this because this tends to make news to be very negative. They will print the most extreme cases or the most negative outcomes, because the truth is that years sells. So reason number two is that AI has already failed to replace know. One of the most high stakes businesses, I think, is Hollywood. And this year, there was one of the longest strikes of the Writers Guilds of America. These are the people that write screenplays and films. I don't doubt for a second that Hollywood tried to use AI to replace them, at least during the strike, because they already used AI for other things, like replacing their dead actors and things like that. But the fact that they settled with the guild, I think, says that AI is not as good as a writer, as a human writer. Going back to tech, a few days ago, Google released the state of DevOps 2023. To me, the most surprising fact was on page eight. It says that AI was the lowest contributor to team performance. So either companies are not using AI or they are not being as powerful as promised. So reason number three is that AI has, like any technology, its limitations. They are trained on Internet data, so there's a lot of misinformation there. There's a lot of bad text from the writer's perspective, text that's full of clutter, or very verbals or not to the point, and has a lot of faults. There's another big problem with these engines, that they do not offer the same level of privacy that some companies need. So maybe you cannot use AI at all for compliance or protecting sensitive data. And there's also a few technical limitations, like context size, and how companies make changes to the models without much transparency. But there's one problem that outshines all this. And in my opinion, the biggest problem with AI is the Ankani Valley. So, in a few words, this is how we respond to things that are human, like things that try to appear human but do not achieve it. They create a sense of aversion and negative emotion. So here we have the same robot in a bigger size. This is Hiroshi Ishiwuro, who is a robot inventor. And you can see Dankani Valley quite clearly here. So why I bring this? Because I think text produced by language model suffers from this. It's not as obvious as a picture or as a robot. Maybe if you are casually reading, you might not notice. But if you read enough content, you start to spot things. You have a feeling there's no one at the wheel, that there's no voice, there's no iness in the text. And as a reader, it's very difficult to me to make a connection with AI generated context that's unedited, that's wrong. And I feel this is a bigger problem, the biggest problem, because you need to connect with your readers, even in technical writing. So reason number four is that we can adapt, knowing the limitations of language models, we can work along them. Here's what I call the adoption spectrum. There's full adoptions on the left and zero adoption on the right. So to the left we have the writer. I call these writers the cyborgs, because they use the raw output of the language model. So they're kind of mixture through robots and humans. And these are people that usually need to write a lot of content in a short time. And I think it shows because the output will suffer for the problems language model have. So you can publish an accurate information or text that's too long, does not go to the point, takes too long to reach a conclusion, or makes no sense. On the other side, maybe we have a person that's not able to use AI, or don't want to use AI for some reason, maybe privacy reasons, or there's some compliance that prevents using these tools. I think the optimal place for us is in the middle whenever possible. So we have two more roles. One is the director. The director is a person that guides the AI very carefully, and it's constantly prompting the language model to generate one paragraph. Paragraph. Another paragraph is going very slowly. You're going to rewrite as needed, adjust, go back, go forth, and it will be like an actual director in a film, telling the actors how to show emotion, how to do their lines, and keep things the bigger picture making sense. The other category here is the swiss army writer. I feel I'm more in this camp, but we use the tool for things that we feel the tool is better than we. For example, maybe we use them for outlining, for moving past a writer's block, for research, for summarizing, for translation, for maybe editing or brainstorming. When we need ideas, or when we have some blogs, some difficulties, we rely on the AI to help us. We can think of AI like another writer that's next to us, and we can talk and interchange some ideas. But we rely a little less on the AI because we still write maybe 50% or more of the text by hand. I think we need to be in this middle zone because we don't want to be over dependent on these technologies and we don't want to lose our skills. The final reason, and with this I finish the talk, is that writing is only 20% of the job. I think there's so much more to editing than writing. You have to interview people, you have to talk to SMEs, to developers, to product managers. You have to be in the middle of all this storm. You have to put yourself in the shoes of the users, prioritize information, make all kind of decisions that maybe we're not even aware of. And these are things that the language models cannot do or struggle a lot to do. They're not good at this. So until we have a language model that can have a creating with a developer and ask the right questions, I think we're safe. So I started talk questioning if technical editing is a worthy pad. My answer was, it depends. It depends on the type of writing you do. Do you need the human touch? I think you do in all writing, even on technical writing. Do you need to connect with the audience? I think you always need to connect. So these are things that language models cannot do on its own. I think language models are good for a certain amount of things, and we should be good at the other things, the things that the language models cannot do on its own. So it's up to us to figure out how to fit in this new scheme. And if we do, I think technical editing is a very fulfilling and worthy career. Thank you so much for watching. Since we don't have a chance for a QA, I will leave my contact information. You can dm me or drop me an email. In my homepage you will find my personal blog and links to my work on semaphore and my YouTube videos. So if you want to talk, contact me. I will be happy to discuss and talk to you and hear from you. Thanks again and have a great rest of the conference.