Why the language industry with human linguists will thrive in the age of AI

A post on DALL-E 2 by Open AI published last week by one of my local localization industry friends, G?khan Do?ru ?from Translation Technologies Academy made me think of some hot topics in the language industry.?So, I thought I would take the opportunity to express my positive thoughts on the language industry and linguists in the age of artificial intelligence (AI) - I believe the language industry and linguists have very important roles in it.?

DALL·E 2 is a new AI system that can create realistic images and art from a description in natural language, in a plain word, a text description.?DALL·E was created by training a neural network on images and their text description. ?If you are familiar with some AI applications in the language space, you may know GTP-3 and other similar models that “write” a lengthy text with a title and a few short descriptions of it. DALL·E is similar but it creates images instead of a text, and it is called a Text-To-Image (TTI) technology. ?This short video will explain you better and you can see what I am talking about.

?

And DALL·E?is not the only emerging TTI technology, but there are more in the market. ?For example, Google announced last month that their new Imagine AI outperforms DALL-E on TTI generation. You can read about it here in this Google's research project page.?

So why do I believe what I believe – linguists have a bright future in the AI era?? Here are some reasons.?

First of all, to make a TTI system work, the relationship between an image and its text description needs to be accurate. As we all know, there is huge work to clean the noise and correct the wrong annotations and labeling of images in text. I am certain that Open AI or Google, or whoever else that are developing this type of technology, can find a way to do it efficiently in the future, but to get to that point, there may be a huge volume of work by linguists who are culturally intelligent humans who are keen on languages. ?

Furthermore, all of these systems will need an integration of another AI/NLP application which is called machine translation or expand the system with other languages like GTP-3. For example, DALL·E is developed mostly in English. You can see the evaluation by some Spanish speaking AI experts here in this video by DotCSV. They tested it in Spanish as well but they tell us it definitely works better in English and they kept testing it in English.?This will be resolved in near future, and again here may be the job to be done by the language people.

Another aspect where this system requires human involvement is evaluation and rating to capture the performance of the system as you can see on Google’s report.?In addition, when the system renders something wrong, someone must fix it.?To detect the errors and provide that feedback to the system is a human’s job as well whether that is incorrect labeling or lack of appropriate definitions of compound words. I am certain that Open AI or Google, or whoever else that are developing this TTI technology, can find a way to do it efficiently in the future, but to get to that point, again there may be a huge volume of work by trained linguists like us in the short term.

Taken together, these make me feel that the future of the language industry as a whole is bright. Only by looking at this model, we can see some different levels of involvement by humans who are keen on language. There are millions of other models and applications underway.?We have skilled talents in the language industry.?The only thing is that we need to keep shifting what we do and how we work. While our current skills are essential, unless we stay current with the rapidly emerging new technologies and applications in which we can be of service, we won’t be thought of as the go-to talent that we are.

I am honored to work with the Translation and Localization Conference as one of the ambassadors.?I am cooking a workshop for the next edition from September 29 to October 1, 2022.?It will be a workshop to discuss how we can navigate through this volatile, uncertain, complex and ambiguous (VUCA) time. ?Our workshop, "Power Up! Brain Hacks and Work Hacks Workshop" aims to provide you with some tools at the conference on October 1, 2022, in Warsaw.?We’ll be discussing some of the ways to equip you to work efficiently and effectively in digesting the information and staying current with the trends in addition to how to make your day less stressful elevating productivity.??I will be delivering it with Kris Girrell, a seasoned coach and trainer who guided professionals and executives in Corporate America over three decades, and Michal Antczak from Paypal, an experienced localization professional who will be launching as a certified coach in the industry shortly.??It's an in-person workshop where you can see us and work with us with the techniques we teach.?

Happy week, everyone!

PS: I have a discount code for KTLC 2022 as an ambassador.?If anybody is interested, please let me know by DM or just commenting here.?See you in Warsaw soon! (Go to the registration page and enter the discount code: TeamYuka2022KTLC)

#automatictranslation #language #technology #localization #innovation ?#linguists #AI #neuralnetwork #NLP #innovation #MT #machinetranslation

Yuka Nakasone?- Chief Strategist at?Global Bridge. With over 30 years of experience in international business, Yuka helps SMEs in the retail industry including retailers, distributors, and retail technology companies to optimize their global operations and grow internationally with digitization and localization through her consultancy company, Global Bridge. Born and raised in Japan, she started her globalization career in Boston, MA, USA, and now enjoys a borderless life with her own small Franco-Japanese family on the Mediterranean coast near Barcelona, Spain.