Venturing beyond: mapping translation and AI with Danielle Sanchez

On International Translation Day weekend in 2023, I met Danielle Sanchez. Coincidence? I don't think so. Danielle had just upgraded her ProZ.com membership to the new Premium monthly option and something told me she knew exactly what she was doing. The Premium membership option was released to those who would like to enjoy all the benefits available at ProZ.com, but also to those who understand the importance of expanding their online presence (the Premium comes with a website). But that's not all, since Danielle already had a great website that she created herself, I knew she was up to something else: the AI component also included in Premium. (For more information about the Premium monthly option click here).

Danielle is  an English <> Portuguese translator specializing in Engineering, Marketing, Tourism, Mathematics, Statistics, and Computing Science. She has been in the site since 2011 and she is the daughter of Flavio Steffen, one of the first translators to register an account on ProZ.com in 1999, whom we remember with much affection and respect.

I won't explain why Danielle decided to invest in her business through ProZ.com membership, but rather, I went to Danielle and invited her to tell us all a little bit about herself (yes, what you will read is just a little bit of what Danielle has to offer) and her relationship with language, translation and technologies. Meet Danielle Sanchez, or Dani Orgânica...

How did you go from being a language professional to an AI-empowered language professional? What did it take to get there?

I think I already entered the field with a different background and mindset. I am a Statistician, and I had a consulting company for almost two decades in this area, so I have no fear of using any technology at all. For example, for me, CAT tools are tools to build databases, and databases are the bread and butter for statisticians. Mainly, I do not see technology as a limitation for creativity. I think this is a narrow-minded world vision, and it just helps to perpetuate the stereotype of Humanities X STEM areas. I am a statistician and a translator; I refuse to accept this labeling. In fact, CAT tools help to improve my writing, for example, by helping with terminology consistency, so my mind is free to compose the text. And it is the same process with AI.

In what ways do you incorporate AI into your work?

My personal MT, Dani Biônica (Bionic Dani), has been AI-powered since 2023. But I started "playing" with GPT last year, following my husband's advice (a Data Scientist, Mathematician, and Economist). He advised me to talk with it as it is an assistant, and I decided to make a mental avatar for it. Now, I call GPT 4.0 by the name of one of my former interns from when I worked as a Statistician. The person was a 20-year-old IT intern in our DBM department and sometimes arrived still tired from the last night's rave, and always had this attitude of knowing a lot about everything. So, I remembered my interactions with this intern and started to talk with GPT 4.0 the same way, and it worked! And I am "taming" CoPilot now under the same idea. I know it is powered by GPT 4.0, but it still needs more conversation than GPT.

As I said before, there is no prompt per se, but it is indeed a conversation that I have with them. I also request that they narrow down their answers. GPT (CoPilot is based on GPT 4.0, too) is an LLM, and their response is always based on the mode (the statistical mode indeed, what we call "moda" in Portuguese). When I refuse an answer from them, they go after the second mode, and so on. I continue the conversation until they reach a mode that is adequate for me, according to my translation expertise. I ask them:

I really use AI as an intern. Really. An intern with expanded capabilities.

How do you balance the use of AI with your own expertise and intuition during the translation process?

As we say in Brazil, with our unique Portuguese, you have to feed "the flea that lives behind your ears" and keep her healthy and working well. The meaning is to be able to gauge how fishy something is, but in Brazil, we like to outsource this job to a flea (or, better, a colony of fleas!) behind our ears! I feed mine on a daily basis with skepticism and information, and they are particularly hungry for reliable and trustworthy sources.

How do you stay updated with advancements in AI technology and their implications for language professionals?

Trying the tools, reading articles at the source (for example, on the OpenAI website), connecting with people who can talk about this with me, and mainly not being afraid of testing new technologies. Seeing all of this as "stuff" and just tools to help me expand my capabilities: that is my point of view. This part is not what really concerns me. Knowing how to translate is "necessary, but not sufficient," to quote a phrase very common in Mathematical areas. It means that it is tautological, and you must be proficient in translation, period.

Rant warning: My concern is that translators need to learn more about business first. There is no point in going ahead and learning new tools without knowing what to do with them. Learn about their final clients, their market, and be able to define a service portfolio that is not "TRANSLATION" (whatever your definition for translation).

I. Making a wedding cake, or even providing the entire buffet, is not the same as being responsible for the wedding and the meaning this wedding will have for the people in it.
II. Everything has different equilibrium points in the market. Several translators have never read anything near ECO 101, and this is really worrisome.

I have an almost ten-year joke about it (I localized to international audiences, LOL): Dial, Dove, and Lancôme all produce soaps. All of them clean your skin properly. However, Dove's soap has a moisturizer in its composition. And Lancôme has miracle regeneration chemicals. Of course, they are in three very different price tiers for their own reasons. What does your client want? All brands supply "cleanliness" for the client. If your final client needs this, your client is a "Dial" client. However, if your client has dry skin that needs to be taken care of, this client is a "Dove" client, and it will pay more for the added value. Finally, if your final client needs to address, for example, its wrinkles, this client is a "Lâncome" client, and the tier needs to be adjusted for all the added benefits of this service. The idea is the same regarding translation. Market segmentation and offering services/products that are adequate to this specific target. But, as always, I diverge.

What (who?) is Dani Bionica?

Dani Biônica is my bionic version, my adaptative MT, fine-tuned with my translations since middle-2022. Now, it is powered with AI, according to my provider, and it has been offered to my final clients since the beginning of 2023 after some months of training and testing on my side.

She was named after a conversation I had with my husband back in 2020. He said that he sees me, with the advent of AI, as a bionic translator in the likeness of The Six Million Dollars Man (yes, he is a fan of this series).

I can tell you the story of how I felt miserable after trying to create an adaptative neural network in Vertex AI to fine-tune it with my translations. I attended several courses in Cloud Computing to do so. It was a lesson in humility (LOL), so I admitted defeat and looked for what was available in the market, and after some testing on the offers at the time, I opted for ModernMT from Translated, the heart of Dani Biônica.

I can also tell you about all these years that all our hard-crafted HT (human translation) segments were paid once (for us) and profited twice by the agencies (delivering for their final clients and selling the TMs for the MT companies). However, I decided to do something about this: to have my own MT.
I offer her to my clients as a choice, the combo MTPE Dani Biônica + Dani Orgânica (Organic Dani - myself). She is good for MTPE jobs. I also offer my HT combo, as usual.

Currently, Dani Biônica works very well in multiple fields, such as colloquial texts, marketing, life science texts (with a low level of terminology), and engineering, as the MTs from agencies are.

What advice would you give to language professionals who are considering integrating AI into their translation practice?

Don't be afraid. Really. Give AI and related technologies a try, thinking of them as tools. You do not fear a blender, you do not fear a car, you do not fear a pencil, and so on. Nobody will blood-eagle you because you are trying to understand a tool. Ask around, talk with other colleagues, stop with the FOMO. And for the Luddites out there: stop with the technology shaming. Accept the changes.

Thank you, Danielle, for this interview! Your testimonial is yet another proof that, even when we cannot avoid evolution in any industry, including the translation industry, there are always strategies and tools to empower ourselves as professionals and succeed in whatever we set out to do.

Do you want to know more about Danielle? Visit her website. And if you too are ready to empower yourself and embrace new technologies, consider ProZ.com Premium membership that comes with your own website designed by experts, an AI companion tool developed by ProZ.com, an AI community of experts, AI training and events, and more for $25 USD per month.