Artificial Intelligence in Interpreting
Pro
Videoconferencing in the wake of the Corona pandemic is here to stay. AI already facilitates the intensive preparation phase of every interpreting assignment.
Formal, well-structured text in standard language can already be handled by AI. But … (read the Contra-arguments)
AI can emulate the speaker’s voice. It remains to be seen if this is a positive or negative feature.
AI doesn’t need sleep, food, a hotel room, travel arrangements etc. It may be cheaper than human interpreters. But remember, AI doesn't come for free, either.
2024
State of the art
Unstructured oral utterances:
46 % of AI interpreting is correct. The rest is incomprehensible nonsense. The results are better when the AI was fed in advance with the structured and revised speech texts, as well as the terminology.
Disturbing noises throw the AI off course.
AI interpreting is a literal machine translation meaning you will hear every word which is very tiering for the audience.
Friend or Foe?
The person-to-person approach
What can go wrong
goes wrong
Power outage
Machine interpreting follows three steps, the so-called cascade model.
Step one: speech to text (prone to error)
Step two: machine translation (prone to error)
Step three: text to speech (prone to error)
A little error in step 1
can grow into an avalanche of
errors …
errors …
errors …
The AI is ignorant of these errors and therefore can not correct its output like an experienced human interpreter does.
The arguments presented here are not exhaustive.
For more detailed information, get in touch with me, Michael Miethke, your consultant interpreter.
An informed person avoids costly errors.
Go to
Contra
Semantics and pragmatics are still an unknown dimension to AI. The same goes for tone of voice, facial expressions, body language, emotions, cultural and colloquial nuances, jokes, irony and subtext as well as common sense linguistic decisions.
AI can’t handle spontaneous, casual, unstructured speech, local dialects and less supported languages.
AI can get things perfectly right one moment and completely wrong the next – and you won’t notice the difference.
AI cannot check plausibility of what it says.
AI is hallucinating meaning, it is making things up.
The quality of AI-based translation and interpreting depends on the language pairs and the underlying Large Language Models (LLC).
Speed can be an issue for the machine interpreting, even more than for humans.
Confidentiality is always an issue because machine interpreting depends heavily on the use of powerful cloud services, WiFi connection, and enormous computing power.
AI is not able to select the most relevant speaker during a lively discussion between several participants.
Die deutsche Übersetzung folgt in Kürze.
Künstliche Intelligenz beim Dolmetschereinsatz
Pro
Videoconferencing in the wake of the Corona pandemic is here to stay. AI already facilitates the intensive preparation phase of every interpreting assignment.
Formal, well-structured text in standard language can already be handled by AI. But … (read the Contra-arguments)
AI can emulate the speaker’s voice. It remains to be seen if this is a positive or negative feature.
AI doesn’t need sleep, food, a hotel room, travel arrangements etc. It may be cheaper than human interpreters. But remember, AI doesn't come for free, either.
2024
State of the art
Unstructured oral utterances:
46 % of AI interpreting is correct. The rest is incomprehensible nonsense. The results are better when the AI was fed in advance with the structured and revised speech texts, as well as the terminology.
Disturbing noises throw the AI off course.
AI interpreting is a literal machine translation meaning you will hear every word which is very tiering for the audience.
Friend or Foe?
The person-to-person approach
What can go wrong
goes wrong
Power outage
Machine interpreting follows three steps, the so-called cascade model.
Step one: speech to text (prone to error)
Step two: machine translation (prone to error)
Step three: text to speech (prone to error)
A little error in step 1
can grow into an avalanche of
errors …
errors …
errors …
The AI is ignorant of these errors and therefore can not correct its output like an experienced human interpreter does.
The arguments presented here are not exhaustive.
For more detailed information, get in touch with me, Michael Miethke, your consultant interpreter.
An informed person avoids costly errors.
Go to
Contra
Semantics and pragmatics are still an unknown dimension to AI. The same goes for tone of voice, facial expressions, body language, emotions, cultural and colloquial nuances, jokes, irony and subtext as well as common sense linguistic decisions.
AI can’t handle spontaneous, casual, unstructured speech, local dialects and less supported languages.
AI can get things perfectly right one moment and completely wrong the next – and you won’t notice the difference.
AI cannot check plausibility of what it says.
AI is hallucinating meaning, it is making things up.
The quality of AI-based translation and interpreting depends on the language pairs and the underlying Large Language Models (LLC).
Speed can be an issue for the machine interpreting, even more than for humans.
Confidentiality is always an issue because machine interpreting depends heavily on the use of powerful cloud services, WiFi connection, and enormous computing power.
AI is not able to select the most relevant speaker during a lively discussion between several participants.
De Nederlandse vertaling zal binnenkort gereed zijn.
Kunstmatige intelligentie tijdens tolkwerkzaamheden
Pro
Videoconferencing in the wake of the Corona pandemic is here to stay. AI already facilitates the intensive preparation phase of every interpreting assignment.
Formal, well-structured text in standard language can already be handled by AI. But … (read the Contra-arguments)
AI can emulate the speaker’s voice. It remains to be seen if this is a positive or negative feature.
AI doesn’t need sleep, food, a hotel room, travel arrangements etc. It may be cheaper than human interpreters. But remember, AI doesn't come for free, either.
2024
State of the art
Unstructured oral utterances:
46 % of AI interpreting is correct. The rest is incomprehensible nonsense. The results are better when the AI was fed in advance with the structured and revised speech texts, as well as the terminology.
Disturbing noises throw the AI off course.
AI interpreting is a literal machine translation meaning you will hear every word which is very tiering for the audience.
Friend or Foe?
The person-to-person approach
What can go wrong
goes wrong
Power outage
Machine interpreting follows three steps, the so-called cascade model.
Step one: speech to text (prone to error)
Step two: machine translation (prone to error)
Step three: text to speech (prone to error)
A little error in step 1
can grow into an avalanche of
errors …
errors …
errors …
The AI is ignorant of these errors and therefore can not correct its output like an experienced human interpreter does.
The arguments presented here are not exhaustive.
For more detailed information, get in touch with me, Michael Miethke, your consultant interpreter.
An informed person avoids costly errors.
Go to
Contra
Semantics and pragmatics are still an unknown dimension to AI. The same goes for tone of voice, facial expressions, body language, emotions, cultural and colloquial nuances, jokes, irony and subtext as well as common sense linguistic decisions.
AI can’t handle spontaneous, casual, unstructured speech, local dialects and less supported languages.
AI can get things perfectly right one moment and completely wrong the next – and you won’t notice the difference.
AI cannot check plausibility of what it says.
AI is hallucinating meaning, it is making things up.
The quality of AI-based translation and interpreting depends on the language pairs and the underlying Large Language Models (LLC).
Speed can be an issue for the machine interpreting, even more than for humans.
Confidentiality is always an issue because machine interpreting depends heavily on the use of powerful cloud services, WiFi connection, and enormous computing power.
AI is not able to select the most relevant speaker during a lively discussion between several participants.
La traduction française sera bientôt disponible.
L'intelligence artificielle au service de l'interprétation
Pro
Videoconferencing in the wake of the Corona pandemic is here to stay. AI already facilitates the intensive preparation phase of every interpreting assignment.
Formal, well-structured text in standard language can already be handled by AI. But … (read the Contra-arguments)
AI can emulate the speaker’s voice. It remains to be seen if this is a positive or negative feature.
AI doesn’t need sleep, food, a hotel room, travel arrangements etc. It may be cheaper than human interpreters. But remember, AI doesn't come for free, either.
2024
State of the art
Unstructured oral utterances:
46 % of AI interpreting is correct. The rest is incomprehensible nonsense. The results are better when the AI was fed in advance with the structured and revised speech texts, as well as the terminology.
Disturbing noises throw the AI off course.
AI interpreting is a literal machine translation meaning you will hear every word which is very tiering for the audience.
Friend or Foe?
The person-to-person approach
What can go wrong
goes wrong
Power outage
Machine interpreting follows three steps, the so-called cascade model.
Step one: speech to text (prone to error)
Step two: machine translation (prone to error)
Step three: text to speech (prone to error)
A little error in step 1
can grow into an avalanche of
errors …
errors …
errors …
The AI is ignorant of these errors and therefore can not correct its output like an experienced human interpreter does.
The arguments presented here are not exhaustive.
For more detailed information, get in touch with me, Michael Miethke, your consultant interpreter.
An informed person avoids costly errors.
Go to
Contra
Semantics and pragmatics are still an unknown dimension to AI. The same goes for tone of voice, facial expressions, body language, emotions, cultural and colloquial nuances, jokes, irony and subtext as well as common sense linguistic decisions.
AI can’t handle spontaneous, casual, unstructured speech, local dialects and less supported languages.
AI can get things perfectly right one moment and completely wrong the next – and you won’t notice the difference.
AI cannot check plausibility of what it says.
AI is hallucinating meaning, it is making things up.
The quality of AI-based translation and interpreting depends on the language pairs and the underlying Large Language Models (LLC).
Speed can be an issue for the machine interpreting, even more than for humans.
Confidentiality is always an issue because machine interpreting depends heavily on the use of powerful cloud services, WiFi connection, and enormous computing power.
AI is not able to select the most relevant speaker during a lively discussion between several participants.