Who are the humans in the loop?

Last week, in the midst of a project kick-off call, a regulatory client of ours confided something…

They had been approached by a new player in the translation market – an AI translation platform making big claims.

Our client was interested to see if the platform actually performed, so they decided to run a test using some of their own regulatory content. The results, they told us, were interesting.

The translations generated by the platform were generally not bad, and they were most certainly quick. So the benefits in terms of time saved seemed clear…

On the flipside, the results for some languages were significantly better than others. And, given the strict requirements for regulatory content, it was clear all languages would require detailed human review and editing before they could be considered submissible.

Highly structured content returned good results for a number of high-resource languages[1], while the results for less structured content, across all languages, were categorically not so good.

So, did the new platform win them over?

Ultimately, our client concluded that this platform was not for them. And so ended their experimentation with AI for the foreseeable. You might think this was our cue to sit back and wipe our brow; relieved to fight another day. Well… actually no.

As life sciences translation experts, we at Conversis are constantly evaluating new applications for AI. The backbone of our offering has always been to create meaningful efficiencies for our life sciences clients. So, as far as we’re concerned, AI is another welcome tool to enable this. Currently, we use AI across some of our multimedia services; we use it for a number of translation tasks; and to create efficiencies across many languages (not all, though the number is growing).

Telling truth from fiction

So, to bring us back to where we started – our conversation with our client – what we found most interesting was not the use of AI itself; it was that the person selling this particular platform had been convincing enough to persuade our very busy client to take the time to test it, especially when the results would fail to convince for his specific use case.

It got us thinking – there have been so many new entrants to the translation industry in the last year or so – offering everything from “human-quality content” to “full workflow automation”. And of course there are! With all the hype around AI and its many potential applications, the market’s been flooded by people looking to make their mark (or at least make some money).

But if everyone with access to ChatGPT now thinks they’re a translation expert, and if they can all use an LLM to convince you they’re life sciences experts too, how do you go about sorting the hype from the real deal? How do you know if the tool they offer will actually work for you?

AI’s meaningful contribution

As we’ve mentioned, in life sciences and in translation, most of us are now using AI to one degree or another. It would be silly not to given the many repetitive tasks it can perform, freeing experts up for where they’re actually needed.

However, in a lot of cases, with these new entrants – regardless of what they claim – the repetitive tasks is where their contribution begins and ends. Their businesses have been built in a day and they don’t have the background or the expertise to add any more meaningful value.

But adding meaningful value is often what makes the difference between a genuine improvement and a vanity exercise. Any AI offering is only as good as the humans who ideate, create, train, iterate, and implement it – the so-called humans in the loop.

Who are the humans in the loop?

If use of AI can generally be assumed from most potential partners, to separate the wheat from the chaff, we suggest you not get blinded by the tech, and instead look to the humans around it.

In other words, and to unashamedly toot our own horn (something we don’t do often enough!), look to us! We have the people, the processes, the experience – over 20 years of it! Everything these new entrants to the market lack.

We know things unprompted (pun fully intended), like when AI will help expedite a given result, and how and when to change things up based on our clients’ needs.

We have a vast linguist network in both high-resource and low-resource languages, living in their countries and communities, experts in their specific fields, and fully up-to-date on relevant (e.g. regulatory) developments.

And we have people with decades of experience in-house who are continually testing, training, and approving AI applications, who let us know when we should integrate new tools. As well as the PMs who feed them specific client requirements and areas for process improvement.

For all the value they bring, LLMs don’t have this grounding in terms of expertise and context (and, frankly, neither do our clients – that’s why they come to us!)

Changing perspective

So, let’s stop looking at this as a battle between traditional translation suppliers and new AI players. The AI part is now a foregone conclusion for pretty much everyone. It’s the quality of the humans in the loop that matters – who they are; what they bring to the table (in addition to AI); and whether you can trust in their expertise, quality processes, and real understanding of what you need from a translation partner.

In a world where AI options are a dime a dozen, it’s up to you to do the due diligence for each new tool or supplier before you spend. Alternatively, you could pick one partner you know you can trust, and let them handle it all for you…

If you would like to look into partnering with Conversis, please get in touch. We would be delighted to discuss your challenges and priorities, and look at where we can help.

[1] i.e. languages that have large volumes of data available to train AI (like English). As opposed to “low-resource languages”, which have a lot less (like Thai).