Will GPT-4o Impact Human Language Learning?

Will GPT-4o Impact Human Language Learning

A search on Amazon.com by “wearable AI devices” turns up AI glasses, voice recorders, bodily function monitors, smart this and the other, and a few hand-held translators … but not the USD 699 Humane AI Pin. According to The Verge, total orders in April 2024 for the machine translation-capable device came to about 10,000 units, and now the company is looking for a buyer. 

The main criticisms from regular folks, influencers, and reporters about the device revolve around poor concept execution, problematic UI, unreliable features, limited compatibility, and overall value for money (compared to a smartphone). One of those unreliable features is the translator. 

The Verge’s review from April 2024 recounted a writer’s experience with the pin in Japanese and Korean: “Rather than translate things, it would just say them back to her, in a horrible and occasionally almost mocking accent.”

Can the device be saved? Y Combinator’s Garry Tan commented on X “I wonder if Humane would be better if it literally was known to do just one magical thing very well, like a realtime babelfish with no delay perfect translation.” We asked readers if they also think Humane should pivot to translation for their AI Pin. Most responders (57.9%) said “No.” A little over a quarter (26.3%) said it definitely should, and the rest (15.8%) think it probably should, if Garry Tan says so.

A Job Market in Flux

Every couple of years, the US Bureau of Labor Statistics (BLS) publishes its Occupational Outlook Handbook, a job index that many American organizations rely on to make decisions around things like academic curricula and salary ranges. In its latest edition, the BLS seems to go the opposite way of doomsayer predictions about the jobs of translators and interpreters, projecting a 4% growth in those jobs by 2032.

Content is much easier and faster to create thanks to AI, and will keep growing exponentially. It is only logical to expect that a transformed language services and technology industry will also grow with changing as well as new roles. The jobs of linguists are already being transformed with increasing AI automation, and job growth for these experts will soon be associated with a completely revamped skillset.

We asked readers to tell us how they see the language industry job market in 2024, and a little less than half of responders (43.4%) said that it really depends on the role. Close to a quarter (24.5%) think it is OK but not great, another group (18.9%) said it is tough to find anything, and the rest (13.2%) remain optimistic, saying it is very strong, with lots of jobs across roles.

Reddit Makes a Strategic Plan Out of MT

Reddit, the social home of over 100,000 online communities, began trading on Nasdaq in March 2024. During its first quarter (Q1) earnings call in May, Reddit’s CEO and co-founder Steve Huffman put translation in the spotlight, telling shareholders “I think one of the big unlocks for us in the near to medium-term is machine translation.”

Huffman considers that translation is key to international growth, adding that Reddit’s user geographic distribution is still 50/50 US vs non-US, whereas other social networks are “more 80% to 90% non-US.” Large language models (LLMs), and with them, AI machine translation, appear to be increasingly on the minds of CEOs like Huffman. 

We wanted to know if readers think Reddit’s bet on machine translation will pay off. Over a third of respondents (39.2%) are not too sure, and think it may be the case in 12 months. Two equal-size groups are on opposite ends of the spectrum, one choosing “No, MT alone won’t be enough” (30.4%) and the other group answering “Yes, definitely” (30.4%).

GPT-4o, Oh!

In May 2024, OpenAI released a new model called GPT-4o (“o” as in “omni”), capable of low latency speech-to-speech translation (S2ST), among many other things. The company demoed the feature and posted the video on social media, leaving the uninitiated not speechless.

GPT-4o is impressive, and many of the comments that poured from those who are new to this technology (or know little about language services) conveyed that the model makes it possible to dispense with translators, interpreters, and even language learning.

LLMs continue to evolve, offering increasingly accurate and nuanced translations, so the convenience of effortlessly understanding and expressing something instantly in another language, as with GPT-4o, is very, very appealing. Yet, learning a language and then using it to understand, translate, or interpret extends beyond functional communication. It allows people to connect in a manner unique to each individual, context, and culture.

We asked readers whether they think AI will increase or decrease demand for language learning in the long term. A little over a third of respondents (36.9%) think it will increase. A slightly smaller group (35.4%) thinks it will decrease. The rest (27.7%) think it will have no impact.