On July 25, 2024, OpenAI announced the prototype version of SearchGPT, available only via waitlist. The move to asking an AI and getting an answer directly instead of a series of links to probe is very attractive. Whether the answer is correct or not is another matter. 

A lot of content showing up in search results is AI-made. OpenAI says SearchGPT’s answers will include sources, so will it be indirectly referencing itself? Odds are, yes, it will.

Back in July 2024, we recounted the times we asked readers to tell us how their use of ChatGPT has evolved, and the latest question was how their use of that or other large language models (LLMs) had developed so far in 2024. In that poll, for a little over a third (33.3%) of readers the use had stayed the same.

A bit later, when asked whether ChatGPT had affected their use of Google Search, most Slator readers chose the option consistent with that latest measure of general ChatGPT use: over two-thirds of respondents (65.7%) are still using mostly the search engine.

Less than a quarter (22.9%) of readers reported using Google Search a bit, and ChatGPT more often. The rest (11.4%) said they mostly use ChatGPT for questions (but aren’t questions also searches?).

An Emoji Might Be More Meaningful

When things are familiar to large sectors of the population, associating a graphic element with a concept is not difficult. For AI, it seems like the entire planet has agreed on the sparkle and the robot emojis to represent it. 

Explaining to your friends and relatives things like “prosody in cascade and direct speech-to-text translation” is… complicated. Even when you fully understand these concepts, the old diagram on a napkin might be far more eloquent than just your words.

Language AI is in the hands and the hand-held devices of consumers. Naturally, as with other technologies, most can probably explain what the AI does, but not how it does it. The many comments on social media referring to AI features as being “like magic” are good indicators that the sparkle represents the concept for most quite well indeed.

Those of us who have jobs in the language industry get the AI “how” questions. And most readers (51.1%) concur that explaining our jobs to family and friends still has the same level of difficulty, compared to three years ago.

A little over a third (36.2%) of readers have found it harder to explain, and a small group (12.7%) thinks it is easier now.

The Tedium of Post-Editing

Machine translation post-editing (MTPE) has been around since, well, since machine translation (MT) has been around. All of this century so far and part of the last one. What is different as we enter the second half of 2024 is that MTPE surpasses editing of human translation. 

More domains once said to be impossible for AI to handle, such as literature and marketing copy, are also now being machine-translated and handed over to an expert-in-the-loop to polish up the result.

The statement published by The Société française des traducteurs (SFT) in early July on AI, since removed from its website, was too little too late. The Society noted that 70% of its members consider post-editing a threat and that post-editing “ultimately creates mind-numbing fatigue,” is a “tedious task,” and is “very poorly remunerated.”

We asked readers if they agreed that post-editing is tedious and mind-numbing, and most (61.2%) said yes. A little under a quarter of respondents (23.5%) said it is so sometimes, and a small group (10.2%) said it is if the wrong tools are used. A tiny group (5.1%) said they actually liked it.

Is It the Weather?

The weather is partly to blame when certain economic indicators fluctuate. In the language services industry, some indirect impact can be expected, for example, if the weather affects international supply chains. That has happened. 

A summer business slowdown, though, can impact any industry, weather and other major causes notwithstanding. What has not stopped is the influx of capital into parts of the industry, including AI dubbing and captioning, M&As, and the stream of announcements of new capabilities, LLM versions, or language additions by the usual players.

We asked readers if they were seeing a summer slowdown in business, and while a quarter (25.8%) of respondents said they are busier than ever, the rest are split between a definite slowdown (40.3%) and business as usual (33.9%).