The increasing amount of data needed to sustain neural machine translation systems may pose a challenge for their long-term viability. Neural machine translation systems require vast amounts of data to train and improve their performance, and as the demand for these systems grows, so too will the need for more data. This could lead to a situation where the cost and complexity of acquiring and managing this data becomes a significant burden for organizations that rely on these systems.
However, it's worth noting that the development of more efficient algorithms and hardware could mitigate this challenge to some extent. Additionally, the use of synthetic data and transfer learning techniques could also help to reduce the amount of real-world data needed to train these systems. Nevertheless, the increasing amount of data needed to sustain neural machine translation systems is likely to remain a significant challenge for their long-term viability, and organizations that rely on these systems will need to carefully consider the cost and complexity of managing this data.