My 7 key takeaways from GALA 2023
22. March 2023
4 days in Dublin with so many old friends and business contacts: It was wonderful to see that GALA with its spirit of inclusivity and openness has not changed a bit. On the contrary, GALA 2023 in Dublin only barely missed to reach the attendance of the largest-ever GALA in Munich, with only a dozen attendees less at about 485.
And right from the beginning, in the Business Meeting, it became clear that GALA has spent the last years thinking heavily about the future of our industry and of the association itself. Alison and the board laid out a strategic framework with four focus areas that are right on: Technology, business models, talent, and industry impact. And of course, the conference programming was totally aligned with these.
Recent advances in AI have brought about the greatest opportunity for our industry to contribute to the development of business and society at large. If we do it right, we can not only revolutionize the way content is being created and translated, but we can finally achieve the visibility and attention we have so long been talking about. It is clear from many presentations that large language models and NMT will continue improving but will only be useful in production with a human in the loop. And as LSPs we know how to build exactly this process: We learned it with NMT. Time to turn Language Service Provider into Language Success Partner.
2) Business models
As always, there was a lot of discussion about getting away from words and the downward spiral of pricing on words. I am not a believer in getting away from words as a unit of measurement per se. But I think that particularly combined with the human in the loop AI-supported production processes, a lot more focus should be on what actually "is" a word. What is the value of a word, what is its impact on the business? Not all words are equally valuable and therefore the discussion around quality should change and focus on "fit for purpose" rather than "one size fits all". This I think is the critical message to be sending, not to get away from words. And we are already seeing this happen. The great inputs from the Quality Management SIG, plus the newly released standard ASTM 2575 both have this at their core: What purpose does a word have, how should it be produced, and what are its value and cost?
The question of how to attract young people into our profession is a very difficult one. While I don't think the new generation is fundamentally different than previous ones, they do have a lot more possibilities and freedom of choice. Picturing translation as a complex and challenging game of being a language detective may not be the right or at least not the only approach to portraying our industry in an interesting way. Young people live in a time where a digital nomad is more the norm than the exception. And what profession could support this better than ours! Being at the forefront of a changing world of AI, work when you want and where you want, this is interesting for young people. Let's portrait our industry like that.
On a different note: Yes, we need to work together with education and parents to make them understand this is a viable career path. And of course, with universities, which make up 20% of the GALA membership already, they are easily accessible in theory. But I think we should not concentrate on translation or language universities. I think in the future we will not only need trained translators, but we will also need subject matter experts who happen to be multilingual. And we find these people not only in "our" universities but everywhere. I think we need to open our minds to this new requirement and launch the respective visibility campaigns more broadly.
4) Industry impact
Coming back to the first topic of technology, it is clear that it is here to stay and will have an impact on our own business and on our customer´s business. So, what we need to learn is to master the skillful use of machine translation and large language models. The interaction between these two and the differences might still be in the making, but it is clear that we are "moving from a human-driven process to a machine-driven process with human oversight", as Wayne Bourland from DELL put it.
5) Neural MT or Large Language Models (LLM)?
The current state of technology seems to indicate that large language models are more expensive, slower, harder to embed into our process, and less predictable than our already advanced NMT engines. Marco Trombetti, in his wonderful closing keynote, defined ChatGPT as a „very accurate predictor of the next word“. This is a wonderful way to explain what it actually does. And it is getting rapidly better at it, of course. While in 2000 these models could look at the last 3 words and in 2010 at the last 5 words to predict the next one, this skyrocketed to 1.500 words by 2020 and 25.000 words in 2023.
LLMs are more fluent than machine translation, but less accurate, or put another way: GPT is very fluent and quite accurate, while NMT is very accurate and quite fluent. And of course, this is because NMT is more constrained. After all, it takes the source text into account. Given this, LLMs will affect our business by dealing with things like copywriting, content restructuring, summarizing, etc. But since more and more text will be created, this also means more and more translation is needed. And coming back to points 1 and 4, this means we need good production methods to make sure we use the right process for the words depending on their value, business impact, and "fit for purpose" quality requirements.
6) The keynote
Every great conference has to have a great keynote, which presents a view that is normally outside of our business focus. This year's keynote completely fulfilled this expectation, with David Siegel, the CEO of Meetup and author of „Decide and Conquer“, talking about decision-making. From the top 10 of his list, I found 4 particularly interesting. When taking a decision, consider which one of them increases the number of opportunities or closes doors and restricts opportunities. Also, think about whether you are a more long-term or short-term decision-maker and deliberately find someone to help you balance this. Always be honest with people surrounding you in order to avoid surprises. And finally, be aware of your biases (this seems to be David´s favorite one) like recency bias, confirmation bias, status quo, or the sunk cost fallacy, because they influence your decision in negative ways.
7) The future
For our own strategic orientation as a company, GALA clearly showed that technology will continue to be the main driver of change, business models, and business impact. Since our technology focuses particularly on "global content quality", I think we are very well positioned to continue our successful journey, keeping some of the above in mind, particularly when making decisions 😉.
As a side note, it was also interesting to see how many have jumped on the global SEO bandwagon, but also that there really is no tool out there that supports the kind of work "global SEOers" do, such as moving away from the "string-based" approach, considering search intentions, URL localization, keyword ranking, etc. This and the future dominance of AI in content production and the "human in the loop" supervisory role should be in our vision when we start the redevelopment of our quality evaluation platform Globalreview. I think the time is ripe for a new approach to "reviewing" translation which should take into account all the above items 1 through 6.
These are my 7 key takeaways from GALA 2023. What are yours?
Kaleidoscope: Taking your content global
We combine our expertise and software solutions as well as those of carefully selected technology partners to create the right solutions to enable you to achieve success on the global market with your content. Thanks to our innovations and further developments, we continuously make it easier for you to manage terminology, quality, reviews, queries, and automation.Contact us!