It seems as though just about everyone in the language industry is trying to “objectively” predict where any of the GPT large language models, including ChatGPT, will lead us. There is much speculation and many developments to keep up.
During the March 2023 SlatorCon, several industry experts discussed the issue of impact at the Intento panel on GPT. While seeking consensus was not an objective, a common theme was that GPT is not yet at the level of accuracy and usability already achieved by machine translation, but it will get there, fast, with proper training. In fact, GPT is already saving a lot of time on many different tasks. MT will be next.
We asked readers what impact they think ChatGPT will have on the language industry in the short term. The majority opted for a cautious neutral impact option (45%), closely followed by about a third choosing a net positive impact (35.0%). To a few, the impact will be net negative (20%) in the short term.
And In Case You Wanted More…
Yes, more GPT stuff, because, why not? Some of the speculation and outright authoritative predictions from thought leaders point to an increase in various services, including different levels of post-editing and model training, as well as new roles.
On potential new roles, examples are beginning to pop up, like the one mentioned by Voiseed CEO and Co-founder Andrea Ballista during SlatorPod #158 (posted on March 29, 2023), “multilingual synthetic audio prompt engineers.” And that is just audio. Hybridization is happening across language tools, and it is only natural that it will occur in human roles.
We wanted to know if this is already happening, i.e., if clients are asking readers about services around ChatGPT. A large majority responded that this is not happening (69.4%). The rest were split into three small groups, with two identical size groups on opposite ends regarding volume: Yes, many (11.3%) and Yes, but very few (11.3%). The smallest group chose Yes, some (8%).
Google Pushes Using MT for App User Interface
Not short on confidence, as being a big tech player gives you license for, Google believes MT is good enough to take on app user interface localization. On March 14, 2023, during the annual Google for Games Developer Summit, the company announced free MT for developers of Android apps.
The service will first be available in seven languages (Simplified Chinese, French, German, Indonesian, Japanese, Brazilian Portuguese, and Castilian Spanish).
Many questions arise, including the pesky issue of context (visual, cultural, and otherwise). Context, as of yet, is a work in progress in MT and LLMs alike. And this is one of those areas in which we can only say that testers, not time, will tell. About localization quality, that is.
We asked readers for their opinion on this development. Two groups of the same size said it was something inevitable (44.7%) or a bad idea (44.7%). And a very small percentage of respondents thought it was a good idea (10.6%).
A Familiar Cycle? Maybe.
If you were around in the dotcom era, you may find the huge spurt of new AI-driven companies akin to what happened during that time. In the early days of eCommerce, it seemed there were no bad ideas. Venture capitalists threw money at anything and everything .com: portal, storefront, online management, SaaS … online learning, streaming via subscription, and self-service brokerage soon followed.
Out of those, the companies with the best business model and leadership, not necessarily the ones with the most money, are still around. Perhaps AI startups will follow a similar trajectory, and some whose valuation is already the size of a small country’s budget (hello DeepL), may be making a new generation of domestic robots in less than a decade.
Slator Machine Translation Expert-in-the-Loop Report
On a more modest purse, we asked readers to choose a category of startup to invest USD 1K. The results show our love (or bias, as AI scientists would call it) for language technology platforms, as most readers chose this as value for their money (25.0%). The general category of Al Agency took second place (22.7%). Data-for-Al, well established as an underlying element to all things AI, was the third choice overall (15.9%).
The Al Writing, synthetic video, and NLP platform categories each had an equal percentage of investors-to-be (6.8%), and the least popular choice was the Voice Al toolkit (2.3%).
Your information will not be shared with third parties. No Spam.