In 2018, IBM Watson Health made headlines for all the wrong reasons. Its flagship AI system, Watson for Oncology, was found to have recommended unsafe and incorrect cancer treatments, according to internal documents and customer feedback. Fast forward to June 2025, and Microsoft has just unveiled an AI system that reportedly outperforms doctors in diagnosing complex health conditions, marking a major leap toward what it calls “medical superintelligence.”

So, what changed in just seven years? The answer lies in one word: data.

IBM Watson: a cautionary tale

Watson for Oncology was trained on a limited set of synthetic cancer cases and heavily influenced by the opinions of a small group of specialists. This narrow training scope led to a system that often failed to generalise to real-world clinical scenarios. Hospitals reported that Watson’s recommendations were not only inaccurate but sometimes dangerously wrong. The root cause? Poor-quality and insufficient training data.

Microsoft’s leap forward

In contrast, Microsoft’s new AI system was trained on millions of real-world patient records, encompassing a wide range of demographics, conditions, and clinical outcomes. This breadth and depth of data enabled the system to learn nuanced patterns and make evidence-based decisions that rival — and in some cases surpass — those of human doctors.

According to Microsoft, the system demonstrated superior diagnostic accuracy in complex cases, such as rare diseases and overlapping symptoms, where even experienced clinicians can struggle. This achievement underscores the transformative power of high-quality, diverse, and representative datasets.

The difference in data

GAI Translate: precision through human-verified data

At GAI Translate, we’ve long understood that AI excellence begins with data integrity. That’s why our translation precision engine is trained on high-quality datasets that are meticulously labelled and verified by human experts. It is the only AI translation solution with a secure 1-click expert review option to verify AI generated results, and in turn we guarantee that training datasets are always the best quality. This unique combination of expert and machine learning ensures that our AI systems not only learn from accurate and context-rich data but also continuously improve through expert oversight.

Unlike generic translation engines, GAI Translate™ is built to handle complex multilingual content with precision, cultural sensitivity, and domain-specific accuracy. Whether translating legal documents, technical manuals, or global marketing campaigns, our AI delivers results that meet the highest standards – because it’s trained on data that does too.

Professionals who understand the importance of verified data, know why GAI Translate™ performs so well.

“GAI’s ability to create domain specific language vocabularies that produce precise results makes it stand out from other solutions. It’s not just an AI tool – it’s a trusted partner in our mission to improve safety and efficiency.” said Paul Evans, CTO Gammon Construction H.K.

Why it works

• Human-verified datasets eliminate ambiguity and error.
• Expert-in-the-loop training ensures continuous refinement.
• Context-aware models adapt to industry-specific language.
• Ethical data sourcing builds trust and transparency.

As AI continues to reshape industries, GAI Translate remains committed to the principle that quality data equals quality outcomes. In healthcare, translation, and beyond, the future belongs to those who train their AI with care, precision, and human insight.

GAI Translate, a secure precision translation engine, delivers these benefits directly to the legal industry. Working with legal and compliance partners like Tenet Law and Les Ambassadeurs Club, it offers high-volume text and document translations, with an expert review feature to cross-check for validity, all within one click.  

Contact us for more information, or start a 14-day free trial with us today.