Translation and AI: Understanding the Options for Global Training

Artificial intelligence has made translation faster and more accessible than ever, but it isn’t the right solution for every learning project. Interpro Translation Solutions’ Nicholas Strozza, CEO, and Beshar Bahjat, Chief Technology Officer discuss how AI is changing translation, where human expertise remains essential, and what L&D leaders should consider when creating equitable learning experiences for multilingual audiences.

Show Notes:

Nicholas Strozza and Beshar Bahjat of Interpro Translation Solutions discuss the evolving role of AI in translation, the difference between translation and localization, and how organizations can balance speed, cost, quality, and learner experience when delivering training across languages. Key points include:

• Translation and localization are not the same thing. Translation converts content from one language to another, while localization adapts the entire learner experience—including language variations, imagery, audio, cultural references, and regional preferences.
• AI is a powerful starting point, not always a finished product. For many projects, AI can accelerate translation, but human review remains critical when accuracy, brand reputation, compliance, or learner safety are at stake.
• The right approach depends on your goals. Organizations should evaluate audience, content type, language, risk level, and long-term translation needs before deciding how much AI or human involvement is appropriate.
• Learning equity matters. Learners who speak another language deserve the same high-quality experience as English-speaking learners. Poor translations can negatively impact comprehension, engagement, accessibility, and training outcomes.
• Human expertise remains essential in the age of AI. Professional linguists increasingly serve as reviewers, editors, and quality experts who ensure translated content reflects local language, culture, terminology, and organizational standards.

Learn more about Interpro Translation Solutions

Powered by Learning earned Awards of Distinction in the Podcast/Audio and Business Podcast categories from The Communicator Awards and a Gold and Silver Davey Award. The podcast is also named to Feedspot's Top 40 L&D podcasts and Training Industry’s Ultimate L&D Podcast Guide.

Learn more about d'Vinci at www.dvinci.com.
Follow us on LinkedIn
Like us on Facebook

Transcript

Susan Cort: [00:00:00] It can be easy to translate custom eLearning into another language thanks to advances with artificial intelligence.

But how do you know if what you're getting is really best for your learners? 

Nick Strozza: My word of wisdom would be, , what is your ultimate goal? How much translations are you trying to do? Is this like a one-off, right? Or are you looking long term and, you know, maybe this isn't the right long-term fit

Susan Cort: That's Nicholas Strozza, CEO of Interpro Translation Solutions. He and Chief Technology Officer Beshar Bahjat join d'Vinci's Angeline Evans and me to discuss how to select the translation solution that's right for your project. Next on Powered by Learning. 

Announcer: Powered by Learning is brought to you by d'Vinci Interactive. d'Vinci's approach to learning is grounded in 30 years of innovation and expertise. We use proven strategies and leading technology to develop solutions that empower learners to improve quality and boost performance. Learn more at [00:01:00] dvinci.com.

Susan Cort: Me now are d'Vinci Client Solutions Consultant Angeline Evans, and our guests, CEO Nicholas Strozza and Chief Technology Officer Beshar Bahjat from Interpro Translation Solutions.

Thank you both for joining us.

Beshar Bahjat: Thank you for having us

Nick Strozza: Yeah, thanks for having us.

Angeline Evans: We are so excited to have you both on the podcast today.

Susan Cort: And we're really happy to be partnering with you on our translation needs on so many different projects. And for those people who aren't familiar with your company, Nick, start out just explaining what your company does, and maybe if the two of you could just share briefly what your roles are

Nick Strozza: Yeah, absolutely. Um, and, you know, you can think of Interpro as the language people, which is actually our new tagline, which I'm really happy about

Susan Cort: Nice.

Nick Strozza: We’re not a tech company, we're a language services company. So we're focused on, you know, taking content from one language, usually English, to a foreign language and accurately translating it, , leveraging technology and AI and other tools [00:02:00] as needed. Uh, but the people behind what we do is super important. We have a great team here. , Our company has been in business for over 31 years and scalable to help on smaller projects, um, bigger projects, multiple languages at the same time, very much focused on quality and partnership. So a lot of what we do is eLearning, hence our partnership has worked really well since I, I think about 2017 is when I first met, uh, d'Vinci Interactive.

So we've…

Angeline Evans: Yeah.

Nick Strozza: projects, complicated projects together, so we really appreciate that partnership as you said, Susan, earlier. And, , yeah, that, that's us and I'm the CEO currently and really enjoy, um, kind of assisting clients and helping here and ensuring we have a good culture and we maintain our standards of quality and clients continue to come back.

I make espresso in the office, so, you know, kind of a turn- turnkey, uh, turnkey title. But thanks again, and I'll hand it over to Beshar.

Beshar Bahjat: I mean, I'm the chief technology officer at the company. I've been at Interpro since 1999, [00:03:00] I've seen the company grow, but also the industry change. I'm responsible for anything that has to do with technologies. Because even though we don't sell our own technology products, we do utilize and leverage technology, and we have our own internal tools that we use and implement in our work.

So we can always stay upfront with whatever's happening in the, you know, ever-changing technology world. And, you know, we can-- we stay re- relevant to the industry

Angeline Evans: Wonderful.

Well, I am very excited to talk more about AI, which of course has woven its way into all of our lives personally and professionally. But I'm sure there's a lot of myths and misconceptions that we can talk through when it comes to translation. So we know AI has made translation faster, and it's more accessible than ever, but it's also easier to get wrong, and I know I, I'm nervous if the AI translation is accurate sometimes.

So from your perspective, um, as professionals in this field, [00:04:00] what's changing right now in how organizations are approaching translations?

Nick Strozza: Sure. And, and great question, and definitely AI has rocked a lot of industries.

Angeline Evans: Yeah

Nick Strozza: to foreign language translation or localization, which is like ensuring translations are extremely accurate for someone who speaks that language, in theory won't know it was translated from English. Like, that's the ultimate goal of what we do and help our clients with, and where AI tends to miss is that level of quality where…

Angeline Evans: Mm-hmm

Nick Strozza: Kind of like the IKEA effect.

You can kind of know it wasn't written in English, but it still makes sense even though I still break the cabinets.

Angeline Evans: Right.

Nick Strozza: So you know, overall, my, my, my gut take is, you know, it's generated a lot of buzz and industry, um, interest in what we do. So someone who maybe, you know, didn't prioritize language access now has the ability to, hey, AI can help me get there cost-effectively, faster, and then the accuracy is where the mystery is, where, uh, until they've done it wrong or [00:05:00] have understood like what goes into the process of a professional translation, you know, it, it, it seems on the forefront, uh, an easier way to maybe do things.

And, and in some cases, there are good use cases for

Angeline Evans: Right.

Nick Strozza: with AI versus a full, what we call a human translation revision proofreading. On our side it's, it's enhanced, uh, conversations, so we're having more quoting options of, "Hey, here's the full human Mercedes option. Here's the Prius option with AI as a head start."

Angeline Evans: Mm-hmm.

Nick Strozza: are gonna get you, you know, out to the restaurant, but the experience to get there is a little different. So, I would say it's generated a lot of buzz and it it's talking to clients on like what's their end goal? Is it really accurate messaging or is it you know, is it good enough? I've seen that come up a lot where it's like, you know, what is good enough? I don't know. Like, you

Angeline Evans: Yeah

Nick Strozza: do you the what do you think is good enough versus what we think is good enough? Um, so definitely, you know, larger companies are really investing [00:06:00] in AI and some clients are leveraging it internally, like using Copilot.

Angeline Evans: Mm-hmm

Nick Strozza: Clients are developing their own homegrown AI translation tools. So, you know, it depends on the scale of what they're trying to do and if we can help them. So it's a lot of different things are going into it. We actually just obtained our third ISO certification, so there is, from a professional translation company standpoint, there are standards, um, that you should adhere to, to produce quality.

Angeline Evans: Mm-hmm.

Nick Strozza: We have a couple of those already for quality management and translation quality. There is another one called ISO 18587:2017, which is the certification for machine translation or AI with a human post-editor, what we refer to as MTPE. So it depends on the companies, like what are they looking for? If they really wanna get this right but leverage technology in the right way to scale using that kind of process, whether it's with us, a [00:07:00] freelance translator, an in-house person, , is really, really helpful to kind of keep in mind. So , it's definitely things accessible given the perception of getting things done a little bit faster, , and in some levels it does.

So there's, you know, other factors that we'll get into today that I think will be helpful for the audience to know.

Angeline Evans: Well, you definitely have my wheels spinning, and I have a ton of questions, but before I ask more, I f- I-- it came to me as you were talking, I feel like we should level set the audience on the difference between translation and localization at a high level, 'cause I don't s- don't think some folks consider that, and they think it's just always a one-to-one, right?

So can you just briefly touch on that just so, like, maybe some future questions and answers make more sense for them?

Nick Strozza: Yeah, no, 100%. So translation, you know, essentially is making sure, um, the content to a native speaker, you know, speaks to them in their language. It's translated accurately. Um, where localization, you know, in, in my opinion, is taking it a little bit to the next level. So that [00:08:00] could mean, you know, imagery, currencies, making sure that if there aren't like equivalent expressions in that language, there could be what we call transcreation to, you know, make sure that we're aligning the source with the target language.

So think of localization as like the next level,

Angeline Evans: Okay

Nick Strozza: again, someone in that language the Italian won't know it was translated from English. Like that's 

Angeline Evans: Mm-hmm. Yeah 

Nick Strozza: Um, Beshar, did I miss anything there? Would you, would you add anything on that?

Beshar Bahjat: Yeah, I mean, localization is more about like the entire experience. So it's not just translating from English to Spanish, it's translating from English to Spanish that is spoken in Mexico.

Angeline Evans: Mm-hmm.

Nick Strozza: it

Beshar Bahjat: just like, it has like getting it down to the local level as much as possible and to have an experience of, let's say it's an e-learning course. have the full experience for the e-learning course, not just like the text on the screen.

Angeline Evans: Mm-hmm.

Beshar Bahjat: the subtitles, the script on the side, [00:09:00] the, audio that you're listening to, everything is in your language, so that's part of localization

Angeline Evans: Thank you. I think that's really helpful, and I think that also speaks to the quality of translation as we start diving a little deeper. So there's a growing assumption that AI can handle translation end to end. What do you feel some of the biggest misconceptions you're seeing from organizations right now around this?

Beshar Bahjat: Yeah, I mean, uh, AI definitely make things easy and convenient. You can-- anyone can translate now,

Angeline Evans: Right

Beshar Bahjat: But just because you can, that doesn't mean that the final product is what's, what is really desired.

Angeline Evans: Mm-hmm.

Beshar Bahjat: We have like a, uh, the verb that we use right now is gisting. You can use AI to get a gist of what it, what the text means. to be able to get like the final quality, something that has... You have product that you wanna present to your customers, you wanna make sure that, you know, it stands behind, uh, with [00:10:00] everything that you believe in, with the, with your brand. The sim-- you know, it has the right translation. It's not just s-something that can be understood.

It's something that can stand out.

Angeline Evans: Mm-hmm

Beshar Bahjat: Fully automating process doesn't guarantee the, the result.

Angeline Evans: Mm-hmm.

Beshar Bahjat: Translation, again, is just one component of the localization process. Even though AI can handle different steps, but it still cannot handle the whole thing from beginning to end for you. So again, just think of it of a simple process for, uh, translating a course. you'll have the on-screen text, you'll have the script, you'll have the subtitles, you'll have the audio, it... Even if AI can handle all these separate steps, even if it can handle it with, you know, in a way where it's f-workflow, where it's gonna go from one s- from one step to the next and get you the final product, you still have to do other steps.

You still have to make sure that sounds good. For example, if you can do AI voiceover, [00:11:00] it's not just about, you know, the words spoken in the target language. Sometimes you have translation guidelines and things where AI by default is not gonna pick up on its own. You need to make sure that you have something in the process that catch all these things and guarantee you, like, the final output is something that you really want to present to your customers.

The way that, uh, large corporations now look at translation using AI is this way. They try to break it down by language and by importance. So what does that mean? If it's something that in a, done in a language that AI can handle well, then it will definitely, they will definitely use AI for it. If it's something where a language, that really doesn't have any resources available for it, so it's a language, a low-resource language where AI is not gonna do a good job for you. But your option is like, you know, spending a lot of ti- [00:12:00] resi- time, money, resources on this language or just giving the user something that they can see and understand, they might opt for AI too, because they can tell you it's not high-end, but you get something out of it. The other side of it is the importance. So if it's something like, you know, this is your brand, this is your marketing, this is your, uh, communication, you wanna make sure that it's not just AI, it's something that's reviewed by s- an expert. If it's something where internal communication and you just wanna make sure that people understand one another, you can just run it through AI. So it's, there are different, uh, dis- uh, components that define which, what level of human interaction is involved in the process and what is left for AI.

Angeline Evans: Mm-hmm.

Beshar Bahjat: All these things are factored in before a decision is made what to use AI for completely, what to use AI for as a starting point, and what not to use AI for

Susan Cort: so helpful as people are trying to [00:13:00] s- to decide, you know, what do we use AI for? When do we need a professional

Angeline Evans: Mm-hmm.

Susan Cort: your company to step in? And it's not just budget that's part of the question, it's really that whole experience. And Nick, you, you know, we've talked about this before, what experience do you want people to have with the eLearning?

You know, and, and at the end of the day, if they know you've put effort into making sure that it's accurate, it speaks to them literally, uh, they're-- you're gonna get better results with the learning than if you take a shorter path.

Nick Strozza: Yeah, and equity is important, like where if, if someone is looking at the English and that's a professionally written, well-developed course, and then the Spanish, you know, just has captions or there's errors or, you know, it doesn't look good on that company, right? So, um, I understand that budgets are part of things, right?

And we can come up with creative solutions, but, you know, at the end of the day, what is the end goal? What is the, you know-- And then work backwards from there and, and provide different options and pros and cons, right? So, you know, the--

Angeline Evans: Yeah. [00:14:00] And so

Nick Strozza: Good points

Angeline Evans: To sum it up, you know, AI-driven translation performs well in just the down and dirty, like word-to-word translation, and where it might break down is in that overall, you know, holistic experience. Is that, would that be accurate to say? Am I getting that correct?

Nick Strozza: I think so, and, and it's, it's a starting point, right? So it, it depends on how good-- So essentially, yes, there's some languages like Haitian Creole, Somali, Rohingya, like there's some languages, Amharic, that it's just not useful at all as that starting point. Um, it's better to just do it from scratch or it's not even, you know, a mature enough language.

It's not-- might not be available at all. But let's say Spanish or French, where maybe it's a decent starting point depending on the subject matter, right? So it's, it's, it-- there's a lot of factors that go into selecting the right tool that the right engine that would be used for a project. Uh, we've seen the most success with, um, [00:15:00] those more mature language models and content that is like highly technical, repetitive, or clients that have built a terminology database, a glossary of terms that are translated up front that are approved for that subject matter. Anything like marketing related, anything, you know, safety training, high visibility, you're putting some stuff at risk. Like we're gonna make sure on our side as a human in the loop, as the language people, to make sure that it's accurate, right? Or if it's not gonna work, it's not gonna work. Um, but again, it, it really kind of depends on the end goal and what is the quality expectation that the client is okay with, right? Um, but it's a starting point for translations. Where it goes from there is depending on so many different factors, but, you know, it could be a human in the loop on the client's end too. They might have a SME who speaks Spanish, which we, we would love to work with those resources, but having those resources, you know, could be needed, a volunteer to just make sure things are accurate. [00:16:00]

Nick Strozza: the starting point. Um, but there's other tools with AI, there's other things coming up that do enhance the languages. Um, so I think it's, it's basically starting point for a linguist to do editing work.

Angeline Evans: Mm-hmm.

Nick Strozza: quality of that starting point depends.

Angeline Evans: Right.

Nick Strozza: hope that, that helps.

Angeline Evans: Absolutely, yeah. And I, I do think human in the loop is important in all aspects of AI at this point, not even just translation, right? Everyone's, you know, moving cautiously and you can't necessarily accept everything you receive from AI. Um, but especially-- so when I think about translation, and you mentioned this, Nick, in high-risk industries, so if we think of pl- you know, industries like healthcare or compliance or safety, the margin for error is so small because it could be lives at risk, it could be like, there could be a lot at stake.

Susan Cort: Yeah.

Angeline Evans: not training folks correctly.

Um, and so you, it would really make me nervous to rely heavily on automated translation. Can you just talk a little bit [00:17:00] about your experience with that and, and why it's so important?

Nick Strozza: Well, it makes me nervous

Angeline Evans: Yeah,

Nick Strozza: it's, it's just talking to the... It, you know, it's again, it goes back to like talking to the client. So there was like a waste management company, um, that we, we've worked with, and when I met the, the person who's involved with like training there, uh, he was saying like, "Yeah, we do onboarding training in Spanish and, you know, we use different tools."

And the audience says they understand, like they're gonna check the box because they want the job or whatnot, right? But then they get hurt and

Angeline Evans: Yeah.

Nick Strozza: really understand, and it's scary 'cause it might be too late at that point, and it might not be like a big enough lift that people think to do it accurately.

Angeline Evans: Mm-hmm. 

Nick Strozza: um, we haven't heard any like true horror stories yet, but we have seen clients that, you know, pushed a button for automated translations and, you know, it didn't get through their internal QA process, like something broke or something went wrong. Um, but you know, again, it's, high-risk industries are ones where [00:18:00] need to have that conversation.

Um, we've also seen some situations where, um, on paper upfront, it, we didn't think AI or machine translation would be the best fit.

Angeline Evans: Mm-hmm.

Nick Strozza: but it ended up being pretty good. So sometimes you don't know until you do the actual work.

Angeline Evans: Mm-hmm.

Nick Strozza: Sure, any client that invests in building a glossary of terms, has some kind of translation memory database as a starting point, wants to be involved in some capacity with their reviewers, you're setting yourself up for some success, right?

But it does go back to the source English,

Angeline Evans: Mm-hmm.

Nick Strozza: Then Susan, you know, what is the end goal in mind, right? What, what are they trying to do? So that answered

Angeline Evans: It-- No, it definitely did, and I just even think about it from an accessibility standpoint, right? So as trainers, we want every learner to have the same experience and get the same value out of everything. And if it's not, if someone does speak another language and it's not translated in good quality, you're taking away from that experience and from that [00:19:00] accessibility component, right?

And you're not giving them an equal opportunity to learn in the same way. So it is, it is so critical.

Beshar Bahjat: Yeah, I mean, the AI keeps improving, but, I mean, there are - like examples of things where it's just didn't get it right.

Angeline Evans: Yeah

Beshar Bahjat: Know like if you're talking about, uh, high compliance of healthcare, I think like there's a, a example of a pharmaceutical company that had a paper about, uh,

Angeline Evans: Mm-hmm.

Beshar Bahjat: so the AI understood that the topic is hypertension, so it inserted the name of a drug that is not part of the

Angeline Evans: Oh, no.

Beshar Bahjat: itself.

Susan Cort: Oh, wow

Beshar Bahjat: just understood that, oh, this is hypertension, this is the drug that you use for hypertension, just put it in here.

Angeline Evans: Right

Beshar Bahjat: can be scary sometimes. I mean, you have,

Nick Strozza: stories, you

Angeline Evans: Yeah.

Beshar Bahjat: you have, uh, like German language has like some really long long terms.

Susan Cort: Mm-hmm

Beshar Bahjat: AI decided for one long t- word, instead of actually [00:20:00] translating the word, just putting the definition for it.

Nick Strozza: Hmm.

Angeline Evans: Interesting.

Nick Strozza: So it took-- AI took a shortcut, it sounds

Angeline Evans: Yeah.

Nick Strozza: Yeah, there's a funny story about, um...

story about a client that came to us and was like, "Hey, can you just take a look at how our quality is?"

Angeline Evans: Yeah

Nick Strozza: like a manufacturing pump company, and the German, going back to German, ironically, uh, it translated it as like women's pumps, not like

Angeline Evans: Oh, no

Susan Cort: Oh, funny

Nick Strozza: So, you know, again,

Angeline Evans: Totally different context there.

Nick Strozza: might not be good. A-and that would be like, hey, I thought manufacturing technical could be good for AI, and it wasn't in that case. 

Angeline Evans: Right 

Nick Strozza: again, these are funny stories, and it's all fun until

Susan Cort: Till it's not funny. Yeah.

Angeline Evans: Yeah.

Nick Strozza: Right.

Susan Cort: Ugh

Angeline Evans: So talk to me about some words of wisdom that we can share with organizations who are looking to embed translation tools. Um, you know, what should they be paying closer attention to? What should they be looking out for?

Nick Strozza: Yeah, [00:21:00] and again, these tools are being more and more enhanced, , with like a, what we call a TAF, Translation as a Feature, like Canva, Articulate. You push a button, right? And it translates it and, you know, it's all marketed as great, you know, it's quick and easy. Uh, what, what we're noticing is and I think the best thing to keep in mind, you know, these are not translation tools. These are tools for design, authoring, something else. They were not built professional translations in mind. They might be using like a DeepL or Google Translate or some kind of engine, AWS, but still, the purpose of what that tool is for is not for translations.

Angeline Evans: Mm-hmm

Nick Strozza: one of the biggest disadvantages right now to those, uh, tools, um, is it's not building a translation memory.

So I refer to it as kind of like the Wild Wild West. They push a button, Spanish, it's okay. have a human in the loop, whether it's a professional company, a freelancer, an internal SME, [00:22:00] edit it, right? So someone edits it, it goes live, whatnot. They update the course or something similar comes around in Spanish. They have to start all over again.

Angeline Evans: Oh, I did not consider that

Nick Strozza: And

Susan Cort: Interesting.

Nick Strozza: thing 'cause like we've had clients now that came to us where they were working with a previous resource and they have internal reviewers that review the translations, which is really good to have,, to partner with that. they were getting upset 'cause they're changing stuff they already changed, like they, they changed sneaker to shoe. Like, so it's, it's time-saving in the beginning, but my word of wisdom would be, , what is your ultimate goal? How much translations are you trying to do? Is this like a one-off, right? Or are you looking long term and, you know, maybe this isn't the right long-term fit because you're not getting those benefits for why you would want to work with a professional service of some sort, right?

Susan Cort: Mm-hmm.

Nick Strozza: don't know that, and it's just having that [00:23:00] conversation and level setting the expectation.

You know, Vyond has Wellsaid Labs or embedded of some sort, you know, and it's good, and that might be a good use case. So 

Angeline Evans: Mm-hmm. 

Nick Strozza: AI voiceover, one of my talking points today is, uh, it's getting a lot better. AI voiceover with a human in the loop is a really good use case for an AI process for e-learning.

So again, consider it the Wild Wild West. It looks

Angeline Evans: Yeah

Nick Strozza: but you're redoing stuff every time and you're not getting the benefit of working with a professional solution for what you need long-- short

Susan Cort: Yeah.

Nick Strozza: long term.

Angeline Evans: Yeah

Susan Cort: quick and easy isn't always the best option

Beshar Bahjat: Exactly. It's ease of-- ease and convenience. The problem with ease and convenience leads to, like, complacency.

Angeline Evans: Hmm. Yeah

Beshar Bahjat: things start to slip without you even noticing.

Angeline Evans: Yep

Nick Strozza: there's, there's pros to some of these tools. It's not like all negative. Like there's review access that makes some changes, um, right to left like Arabic

Angeline Evans: Yes

Nick Strozza: a little more streamlined. So, you know, it's, it's-- there's pros and cons. You know, you have to kind of [00:24:00] weigh what

Angeline Evans: Absolutely.

Nick Strozza: What is

Susan Cort: Yeah. What do you, what do you wanna get out of it?

Angeline Evans: and with these tools, there is more translation happening than before, which may not have occurred if they had to find, you know, so we are making it more available.

Beshar Bahjat: Yes

Angeline Evans: So as AI continues to evolve, how do you see the relationship between technology and human expertise changing in the translation space that you're working in?

Beshar Bahjat: Yeah, I mean, translation is actually one of the earliest usages of, uh, usage of AI. Uh, that goes back all the way to the '50s.

Angeline Evans: Mm-hmm.

Beshar Bahjat: the current large language models were actually made possible because of advancement in technology that was used first in translation. Technology is there, keep changing. It's gonna change the role that, uh, the human would be doing, the professional linguist will be doing, but that doesn't mean it's gonna [00:25:00] take away the job itself.

Angeline Evans: Mm-hmm

Beshar Bahjat: It just like instead of your job now of translating the 2,000 words a day, your job is gonna be more of a, as a post-editor or maybe something like, you know, you're monitoring the glossary, you're training the engine. wanna make sure that you have someone in the process who's, who knows what he's doing or what the, what she's doing that will be able to take the output from the AI and make it better,

Angeline Evans: Mm-hmm

Beshar Bahjat: the same time train the AI to make sure that the next time it does better. It's gonna keep on evolving, it's gonna keep getting better, that doesn't mean it's gonna away ch...

the, uh, use of the actual translator, but it will just change their role.

Angeline Evans: Mm-hmm

Beshar Bahjat: that's like a standard thing with technology. Change is always gonna happen. There are certain jobs that might go away might just change the way you're doing them

Angeline Evans: Mm-hmm.

Nick Strozza: Yeah, and I think, um, this kind of goes [00:26:00] back to a book I read which was really good called "Same as Ever." I believe the author is Martin Housel, and it kind of-- the, the main theme of the book is over the course of history, something is always gonna happen that you can't expect, you have to adapt and pivot, right?

And it's, it's important to know that, hey, change is gonna happen. Many, many years ago, professional translators were nervous when what is now com- called computer-assisted translation or translation memory, um, tools were in, in, you know, out there. And man, this is, is this gonna take over our job? But what ended up happening is now this is the standard and it's enhanced the quality of work that they were doing.

So,

Angeline Evans: Yeah

Nick Strozza: is just another example of, you know, let's, you know, something that rocked many industries, you know. And with translations, you know, what are the pros and cons? When is it a good use case? are things that are changing gonna help our clients? And just having those conversations and understanding, communicating the quality [00:27:00] expectation pros and cons and what you're getting so that, you know, people are happy with what the end result is, right?

That the worst thing you could do is offend someone with a

Angeline Evans: Yeah

Nick Strozza: You might only have one shot or a life or death situation, 

Angeline Evans: Right

 So where else are you seeing AI break down along the process? Like, for example, I know something that we've been investigating at d'Vinci is just the use of inclusive language with AI and just making sure when we're using AI models, we're, uh, being mindful of any biases that might come across in the outputs we're getting.

Is that something you guys are seeing too?

Nick Strozza: Absolutely. So we're more and more as clients are interested in pursuing, you know, we, we wanna do our due diligence. Um, and we have found out that AI does not handle inclusive language, and we have a couple clients where this is really, really important to them, right? Accessibility. And it just-- it would be more expensive to try to edit it in language than it is to [00:28:00] just do it from scratch.

So super big limitation. Um, you know, we're, we're looking at other situations where, you know, the AI biases just cause issues, right? Uh, Bashar, you said you had an example of 

Beshar Bahjat: I

Nick Strozza: some job professions

Beshar Bahjat: Yeah, I mean, it's like if you ask, you know, any AI engine to generate you an image of a nurse, it will be a woman, or an engineer, it will be a man. That's just by bias. Those are the... I mean, they're trying now to a-address it, but the problem is, like, they're not addressing it in a real way where it's like

Angeline Evans: Right?

Beshar Bahjat: data. They're trying to say, "Oh, just to avoid these make it the other way around," which doesn't even solve the problem. It just, like, creates a new different way of, you know, covering the issue.

Nick Strozza: And, and some less challenging things, but also important to know, it kind of goes back to like the localization example.

Angeline Evans: Mm-hmm.

Nick Strozza: we had a client that developed an e-learning, it was healthcare, and they wanted, you know, MTPE, machine translation AI with a post editor, um, [00:29:00] into French Canada. So nor- anytime we do translations, it's with someone on the ground immersed in the culture, same thing with post-editing. And when they received the French starting point, French Canada, it was mostly French, like Europe. And that's because these engines are trained for that language.

Angeline Evans: Right

Nick Strozza: problematic, you know, it caused more headaches and problems, and if they didn't use a human in the loop, would... the Canadians that speak French Canada would not be happy with that.

Angeline Evans: Right

Nick Strozza: that also has happened, Bashar, maybe with Spanish, like Spain versus

Beshar Bahjat: Yeah.

Nick Strozza: America. Maybe not as, uh, you know,

Beshar Bahjat: Yeah, so

Nick Strozza: of, of negativity, but yeah.

Beshar Bahjat: generally speaking, a lot of the translation models built with, you know, the data available. So there's a da- available data from the English, of course, 'cause, you know, US market is , the major market, but also there is the European Union. So all the standard European Union languages are like the main source for these train these [00:30:00] engines. So when you're-- it's obviously if your European market is, your Portuguese is for Portugal, your, uh, Spanish is for Spain. So if you're, you're using some model that was trained on EU data and for Spanish, and then you wanna take that and apply it to Mexico or Latin America, it's not gonna give you the desired output

Nick Strozza: Lot of factors and, and a lot of it's case by case, but, you know, for clients that are looking at certain types of, you know, end messaging in mind, we need to evaluate is it gonna work or is it gonna be more problematic than a full, you know, human version?

Angeline Evans: Absolutely. You definitely want to be mindful of that a-ahead of time

Nick Strozza: 100%. But again, at the face of it, it looks, hey, it's faster, it's, you know, it's

Angeline Evans: Yeah

Nick Strozza: to do, it's accurate, and it might not be the case every time

Susan Cort: Yeah, this is good advice. I think as people are decisions about how to pull AI into their translation, , projects, they're gonna really be giving [00:31:00] this some new thought. So thank you both…

Angeline Evans: Yeah, this was super insightful. Thank you

Nick Strozza: Yeah. Thank you

Susan Cort: good advice.

Nick Strozza: And, you know, this conversation in six months might be a little different than the one we have

Angeline Evans: We might be bringing you back on then.

Nick Strozza: yeah, if, if you'll have us, we'll be there

Susan Cort: I think we should plan for that.

Angeline Evans: Yeah.

Angeline Evans: Thank you

Susan Cort: My thanks to Nick Strozza and Beshar Bahjat from Interpro Translation Solutions and d'Vinci's Angeline Evans for joining me today. If you have an idea for a topic or guest, please reach out to us at poweredbylearning@dvinci.com. And don't forget that you can subscribe to Powered by Learning wherever you listen to your podcasts.