| Pages in topic: [1 2] > | ProZ should NOT incorporate AI into KudoZ Thread poster: Christopher Schröder
|
|---|
Anyone who wants to use AI to research terms can do so already. The proposed new feature is not needed.
Asking members to test and then correct AI answers, for free, is asking us to feed the beast that is about to consume most or all of us.
Surely ProZ should be supporting translators, not hastening their demise?
| | | | | Technology doesn't stop because we refuse to interact | Aug 8, 2023 |
Ice Scream wrote:
Anyone who wants to use AI to research terms can do so already. The proposed new feature is not needed.
Asking members to test and then correct AI answers, for free, is asking us to feed the beast that is about to consume most or all of us.
Surely ProZ should be supporting translators, not hastening their demise?
I share the concern of many others in terms of what the consequences of rapidly evolving technology will be.
I don't accept any tasks consisting of checking batches of MT or AI output for the purpose of improving the engines.
However, we cannot stop the evolution.
About Kudoz, I don't see in the presentation that we are asked to correct AI responses. I certainly don't intend to, only give my feedback on whether the feature seems useful or not.
The more you use AI, the more you realise how unreliable it is as a source of facts. You will often get contradictory answers from two different AI engines or even the same, so I don't think it's going to change a lot when it comes to the trickier Kudoz questions.
A few things may happen:
- Translators who didn't think of asking AI about a relatively trivial question may become aware that they could just as well have asked ChatGPT, Bard or another AI engine.
- ChatGPT will solve the trivial Kudoz questions so members don't accumulate points for answering simple questions.
Interacting with AI gives me a better understanding of its strengths and weaknesses and thus better arguments when an agency tries to reduce the rate based on MT or AI.
| | | | | AI in Kudoz is not what Proz should be focused upon | Aug 8, 2023 |
Proz could usefully take a lead from Stack Overflow (SO), which has similar aims to Kudoz but in the field of software development. All use of AI has been banned on SO since the beginning of December 2022:
StackOverflow wrote*, on 05/12/2022:
All use of generative AI (e.g., ChatGPT1 and other LLMs) is banned when posting content on Stack Overflow.
This includes "asking" the question to an AI generator then copy-pasting its output as well as using an AI generator to "reword" your answers.
(...)
Overall, because the average rate of getting correct answers from ChatGPT and other generative AI technologies is too low, the posting of answers created by ChatGPT and other generative AI technologies is substantially harmful to the site and to users who are asking questions and looking for correct answers.
(...)
I believe it would be far more beneficial if, instead of pitting its members skills against AI, Proz were to update the Professional Guidelines https://www.proz.com/professional-guidelines to incorporate rules governing the use of AI by professional translators in their relations with clients.
JL
-------
* The full text of the SO announcement, and links to the ensuring discussion within the developer community, is accessible here: https://meta.stackoverflow.com/questions/421831/temporary-policy-generative-ai-e-g-chatgpt-is-banned
| | | | expressisverbis
Portugal
Local time: 17:42
Member (2015)
English to Portuguese
+ ...
| I have doubts and questions... | Aug 8, 2023 |
I respect some of my colleagues' concerns, and I have doubts and questions...
In my opinion, AI will not get its way with or without some member testing.
How this will be tested?
AI translations cannot respect cultural differences. So how can the answers generated by AI in Kudoz be reliable?
In addition to cultural differences, how to accurately translate humour, puns, idioms, wordplay, irony, sarcasm, etc. Will AI be able to understand these nuances of human language?... See more I respect some of my colleagues' concerns, and I have doubts and questions...
In my opinion, AI will not get its way with or without some member testing.
How this will be tested?
AI translations cannot respect cultural differences. So how can the answers generated by AI in Kudoz be reliable?
In addition to cultural differences, how to accurately translate humour, puns, idioms, wordplay, irony, sarcasm, etc. Will AI be able to understand these nuances of human language?
Also, what about highly technical texts?
This is only gained by real experience, and AI is just a "robot".
Then, what about breaking confidentiality or leak of data?
I'm in favour of using AI to help us to understand certain terms when we don't have any other way, and I support and have always supported the progress made in technology, but I think it's a bad idea to use it as a mean to offer translations in terminology at Proz because of these questions I raise. ▲ Collapse
| | |
|
|
|
Lingua 5B 
Bosnia and Herzegovina
Local time: 18:42
Member (2009)
English to Croatian
+ ...
| Are you aware | Aug 8, 2023 |
There are projects out there to be an AI tester/auditor paid by the hour? Why is ProZ asking people to test it for free? I missed this announcement, never seen it.
| | | | | Just an option, not focus | Aug 8, 2023 |
Jennifer Levey wrote: Proz could usefully take a lead from Stack Overflow (SO), which has similar aims to Kudoz but in the field of software development. All use of AI has been banned on SO since the beginning of December 2022: I believe it would be far more beneficial if, instead of pitting its members skills against AI, Proz were to update the Professional Guidelines https://www.proz.com/professional-guidelines to incorporate rules governing the use of AI by professional translators in their relations with clients.
It's not quite the same scenario when a Kudoz answer is specifically labelled as AI generated. Then it is not AI masquerading as a human answer. It will also make it more difficult for others to post AI generated answers without being spotted, since the Kudoz entry now has an AI benchmark. It seems to be an increasing problem that users who don't speak the language concerned simply post AI-generated answers. That sort of behaviour may become more difficult.
SO can have all the rules they like, but if there is no objective way of enforcing them, it's a bit pointless. At present, there is no reliable method to detect what is AI and what is not. Here is just one URL about it:
https://techcrunch.com/2023/02/16/most-sites-claiming-to-catch-ai-written-text-fail-spectacularly/
More rules or guidelines are not going to make us more competitive and they are not going to stop whatever AI technology will do. I understand the visceral opposition, but remaining in denial may well be like shooting ourselves in our collective feet. To maintain a living, we need to convince clients of our value. I don't see how banning AI replies labelled as AI from Kudoz will help. AI is extremely hyped up right now, but it simply cannot do everything people think it can, and the craze may will fizzle out once reality sets in.
| | | |
Thomas T. Frost wrote:
Technology doesn't stop because we refuse to interact
But why would you want to help it take over and accelerate our demise?
I don't mind sharing some of my expertise with fellow human translators, but I'm not going to give it away willingly to a machine.
Thomas T. Frost also wrote:
About Kudoz, I don't see in the presentation that we are asked to correct AI responses. I certainly don't intend to, only give my feedback on whether the feature seems useful or not.
Translator feedback on the AI is part of the plan presented:
ProZ (or should it be Anti-ProZ?) wrote:
Other KudoZ participants will be invited to peer agree/disagree with answers generated with ChatGPT, and the asker will be able to select them as the most helpful and award points. These points, however, will not be considered for directory ranking (i.e. the profile used to post AI-generated KudoZ answers will not be included in directory search results).
In fact, translator feedback on the AI must be the whole point of the exercise. As translators can already do their own AI search quite easily, the only possible purpose of the feature is to get translators to approve/correct the machine's answers, which will be used to make the AI better.
Otherwise, rather than weeding out stupid answers, it's going to generate even more.
ProZ should be taking a stand on the benefits of human translation over machine translation. It's shooting itself in the foot as well as us. It beggars belief.
| | | |
Ice Scream wrote:
But why would you want to help it take over and accelerate our demise?
I don't, and I don't see how AI on Kudoz or not can make any difference to the eventual outcome, whatever it is. I think the reaction is blown out of proportion.
Ice Scream wrote:
I don't mind sharing some of my expertise with fellow human translators, but I'm not going to give it away willingly to a machine.
I entirely agree and I have no intention of doing that.
Ice Scream wrote: Translator feedback on the AI is part of the plan presented: ProZ (or should it be Anti-ProZ?) wrote:
Other KudoZ participants will be invited to peer agree/disagree with answers generated with ChatGPT, and the asker will be able to select them as the most helpful and award points. These points, however, will not be considered for directory ranking (i.e. the profile used to post AI-generated KudoZ answers will not be included in directory search results). In fact, translator feedback on the AI must be the whole point of the exercise. As translators can already do their own AI search quite easily, the only possible purpose of the feature is to get translators to approve/correct the machine's answers, which will be used to make the AI better.
Proz has not said anything about corrections being fed back into ChatGPT, so unless they do that, none of this will make AI better. And it's just a question of agreeing or disagreeing, not of correcting or improving the AI output.
As for the purposes, I have mentioned other potential outcomes in another post.
What do you prefer?
- Non-speakers posting AI output masquerading as human replies?
- AI output being clearly labelled as such?
Ice Scream wrote:
ProZ should be taking a stand on the benefits of human translation over machine translation. It's shooting itself in the foot as well as us. It beggars belief.
Just last month, I reviewed 8500 words of mediocre Danish, logging 80 errors (since a quality report was requested). Apart from the objective errors, much of it sounded like English written with Danish words. I wished it had been MT instead. That sort of translator has no raison d'être. MT can do it better.
In any case, MT and AI are not going away simply because we choose to look the other way. The day when we can no longer do it better than MT, we're done for. But we're not there yet.
| | |
|
|
|
Michael Beijer
United Kingdom
Local time: 17:42
Member (2009)
Dutch to English
+ ...
Ice Scream wrote:
Anyone who wants to use AI to research terms can do so already. The proposed new feature is not needed.
Asking members to test and then correct AI answers, for free, is asking us to feed the beast that is about to consume most or all of us.
Surely ProZ should be supporting translators, not hastening their demise?
I agree. If users want to use LLMs in their terminology research, that should be their decision. However, I don't like the idea of ChatGPT stealing potential KudoZ points from me, which are useful as they boost me up the rankings and make my profile more visible to prospective clients.
| | | | Mr. Satan (X)
English to Indonesian
| Confidentiality | Aug 9, 2023 |
expressisverbis wrote:
Then, what about breaking confidentiality or leak of data?
I don't think this is a valid counterargument. Whether you ask ChatGPT or the 100% human KudoZers to help you decipher certain terms, you're still leaking parts of your client's material to the public.
| | | | expressisverbis
Portugal
Local time: 17:42
Member (2015)
English to Portuguese
+ ...
Mr. Satan wrote: expressisverbis wrote:
Then, what about breaking confidentiality or leak of data? I don't think this is a valid counterargument. Whether you ask ChatGPT or the 100% human KudoZers to help you decipher certain terms, you're still leaking parts of your client's material to the public.
I'm not stupid. I know how to formulate my own research questions without compromising my projects and my clients 
On Kudoz you see many members (perhaps due to lack of experience) disclosing a lot of information, leading to the asker's client. I have seen this before and I still see it happening.
Just by way of example, I have asked ChatGPT the following (of course I know the meaning of this "regional" word and it gives me a wrong answer; this slang term is not used in Brazil):
"What does it mean "porreiro"?
"Porreiro" is a Portuguese slang term that is commonly used in Portugal and some other Portuguese-speaking countries, such as Brazil. It is an informal expression that is used to convey a sense of approval, agreement, or satisfaction. In English, it can be roughly translated to "cool," "nice," "awesome," or "good." It's often used in casual conversations to express positivity or to indicate that something is going well. The word reflects a laid-back and friendly attitude in conversation. Keep in mind that the exact interpretation and usage might vary depending on the context and the region where it is being used."
Do you know what project I am working on? Do you know the field? Can you guess to what client I work with?
We can safeguard our information, but there are still many people who don't know how to do it.
[Edited at 2023-08-09 14:05 GMT]
[Edited at 2023-08-09 14:10 GMT]
[Edited at 2023-08-09 14:13 GMT]
| | | | Tom in London
United Kingdom
Local time: 17:42
Member (2008)
Italian to English
Ice Scream wrote:
Anyone who wants to use AI to research terms can do so already. The proposed new feature is not needed.
Asking members to test and then correct AI answers, for free, is asking us to feed the beast that is about to consume most or all of us.
Surely ProZ should be supporting translators, not hastening their demise?
As of now I will probably stay away from Kudoz.
| | |
|
|
|
Zea_Mays
Italy
Local time: 18:42
Member (2009)
English to German
+ ...
| @ Proz: Make a difference! | Aug 9, 2023 |
While all are jumping on the AI bandwaggon, Proz as a platform for HUMAN language service providers
could - and imho should - make a difference.
An agency I collaborate with for example has sent out an issue of their newsletter clearly stating
that they have chosen the human path.
I believe many of us are waiting for a clear statement from Proz (someone alredy mentioned this these days)
on how they will support and foster the people at the core of their bus... See more While all are jumping on the AI bandwaggon, Proz as a platform for HUMAN language service providers
could - and imho should - make a difference.
An agency I collaborate with for example has sent out an issue of their newsletter clearly stating
that they have chosen the human path.
I believe many of us are waiting for a clear statement from Proz (someone alredy mentioned this these days)
on how they will support and foster the people at the core of their business.
Implementing AI in Kudoz is _not_ the kind of support translators need. It is something that excites developers.
What about implementing AI to automatically send invoices (and reminders) to clients instead? 
Proz.com statement on this website:
Serving the world's largest community of translators,
ProZ.com delivers a comprehensive network of essential services,
resources and experiences that enhance the lives of its members.
▲ Collapse
| | | | Mr. Satan (X)
English to Indonesian
| Counterargument | Aug 10, 2023 |
expressisverbis wrote: I'm not stupid. I know how to formulate my own research questions without compromising my projects and my clients
I’m not doubting your competence, in case you think that.
Do you know what project I am working on? Do you know the field? Can you guess to what client I work with?
No.
We can safeguard our information, but there are still many people who don't know how to do it.
True. But you can also scrub confidential information and reconstruct your prompts to ChatGPT. Ergo, the counterargument still has loopholes.
| | | | expressisverbis
Portugal
Local time: 17:42
Member (2015)
English to Portuguese
+ ...
| No, not at all. | Aug 10, 2023 |
Mr. Satan wrote:
I’m not doubting your competence, in case you think that.
I didn't think of that.
Just an aside: Proz should be aware of several things that don't work properly.
Although I find Kudoz a useful tool and still have the pleasure of communicating and exchanging views with good professionals, it still remains a jungle-like place (I apologise for the choice of word, it was the one I came up with for want of a better one...)
So, with AI or without it, I believe Kudoz will remain like this and maybe even worse....
But what is worse? Artificial Intelligence or Human Ignorance that we sometimes see?
| | | | | Pages in topic: [1 2] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » ProZ should NOT incorporate AI into KudoZ | Protemos translation business management system | Create your account in minutes, and start working! 3-month trial for agencies, and free for freelancers!
The system lets you keep client/vendor database, with contacts and rates, manage projects and assign jobs to vendors, issue invoices, track payments, store and manage project files, generate business reports on turnover profit per client/manager etc.
More info » |
| | CafeTran Espresso | You've never met a CAT tool this clever!
Translate faster & easier, using a sophisticated CAT tool built by a translator / developer.
Accept jobs from clients who use Trados, MemoQ, Wordfast & major CAT tools.
Download and start using CafeTran Espresso -- for free
Buy now! » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
