Pages in topic:   [1 2] >
AI set to make waves on KudoZ: volunteer testers needed
Thread poster: Lucia Leszinsky
Lucia Leszinsky
Lucia Leszinsky
SITE STAFF
Aug 8, 2023

Hello everyone,

AI technology is being added to KudoZ. Soon, askers will be able to get AI-generated answers to their questions upon posting.

Screenshot_9

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).

Screenshot_11

If you are interested in testing AI in action in KudoZ, please post below. We’d like to work with a group of volunteer testerts on improvements to its usefulness and come up with a resource that supports language professionals when helping each other with the translation of terms.

Looking forward to testing together,

Lucia


Thomas T. Frost
 
Thomas T. Frost
Thomas T. Frost  Identity Verified
Portugal
Local time: 15:10
Danish to English
+ ...
Interesting idea Aug 8, 2023

Good idea. I don't mind having it included in my Kudoz use.

Lucia Leszinsky
 
Samuel Murray
Samuel Murray  Identity Verified
Netherlands
Local time: 16:10
Member (2006)
English to Afrikaans
+ ...
Sure Aug 8, 2023

Sure. And I think it's good that this is not included by default, but that the asker should specifically check a box if they want to get an AI generated answer. Is it clear to the user that they're not getting ONLY an AI generated answer, though?

So, I would suggest this wording:

AI-generated answers
[ ] Include an AI answer (powered by ChatGPT)
[ ] Notify me of the AI answer

...and obviously answerers should be given the option not to be noti
... See more
Sure. And I think it's good that this is not included by default, but that the asker should specifically check a box if they want to get an AI generated answer. Is it clear to the user that they're not getting ONLY an AI generated answer, though?

So, I would suggest this wording:

AI-generated answers
[ ] Include an AI answer (powered by ChatGPT)
[ ] Notify me of the AI answer

...and obviously answerers should be given the option not to be notified of AI answers.

I'm concerned, though, that askers who allow AI answers will look at the AI answer, evaluate it to see if it suits them, and then simply neglect to check back and read the human answers. Perhaps the AI answer should appear only after e.g. 5 or 10 minutes, and not immediately. Or, you should make sure that the asker is notified of the human answers when they occur, so that he doesn't forget about rewarding the helpful humans.
Collapse


Kevin Fulton
Lucia Leszinsky
expressisverbis
Maria Teresa Borges de Almeida
Philip Lees
 
Zea_Mays
Zea_Mays  Identity Verified
Italy
Local time: 16:10
Member (2009)
English to German
+ ...
implications Aug 8, 2023

Will GPT then earn points,
and will it be listed among the translators in the professionals directory?
What's the added value vs asking GPT or Bard etc. directly?
Is the aim of Proz promoting translators or OpenAI?
...

[Bearbeitet am 2023-08-08 12:03 GMT]


Heike Holthaus
Martina Klett
Barbara Carrara
Christopher Schröder
P.L.F. Persio
Maria Teresa Borges de Almeida
Matthias Brombach
 
Heike Holthaus
Heike Holthaus  Identity Verified
United States
Local time: 10:10
Member (2012)
English to German
+ ...
Hello AI – Good-bye ProZ? Aug 8, 2023

If I wanted AI answers, I'd go straight to AI. What's the point of going through Kudoz?

Who ensures AI's answers are correct? If users are not waiting for human comments and/or answers and just choose an AI answer because it sounds right, what will that do to the integrity of the Kudoz termbase?

Quite frankly, I've been contemplating the value of a professional ProZ membership for some time; even considered cancelling altogether. ProZ promoting AI will make that decisio
... See more
If I wanted AI answers, I'd go straight to AI. What's the point of going through Kudoz?

Who ensures AI's answers are correct? If users are not waiting for human comments and/or answers and just choose an AI answer because it sounds right, what will that do to the integrity of the Kudoz termbase?

Quite frankly, I've been contemplating the value of a professional ProZ membership for some time; even considered cancelling altogether. ProZ promoting AI will make that decision that much easier.
Collapse


Barbara Carrara
expressisverbis
Christopher Schröder
P.L.F. Persio
Zea_Mays
Maria Teresa Borges de Almeida
Cristina Bufi Poecksteiner, M.A.
 
Lucia Leszinsky
Lucia Leszinsky
SITE STAFF
TOPIC STARTER
Before thread goes off-topic... Aug 8, 2023

This topic is an invitation for testers of the new AI feature in KudoZ. If you would like to discuss AI and its impact on the industry, there is a dedicated forum for that: https://www.proz.com/forum/2219

If you are unsure of how AI will impact KudoZ usage, please volunteer to test it. This integration is not thought of as a replacement, but rather as a supporting feature. If AI technology is
... See more
This topic is an invitation for testers of the new AI feature in KudoZ. If you would like to discuss AI and its impact on the industry, there is a dedicated forum for that: https://www.proz.com/forum/2219

If you are unsure of how AI will impact KudoZ usage, please volunteer to test it. This integration is not thought of as a replacement, but rather as a supporting feature. If AI technology is available out there and language professionals are using it (in many different ways), why not include it in KudoZ as an option? ProZ.com provides tools and opportunities to its members. This is just another one of many and we want to make sure it works to your benefit, hence the volunteer testers invite.

@Thomas T. Frost: thanks for volunteering! I'm contacting you in private shortly.

@Samuel Murray: thanks for highlighting the importance of not sending notifications for AI answers! Checking with developers now to define the best approach to this.

Lucia
Collapse


Thomas T. Frost
expressisverbis
 
Zea_Mays
Zea_Mays  Identity Verified
Italy
Local time: 16:10
Member (2009)
English to German
+ ...
Killing Kudoz Aug 8, 2023

Lucia Leszinsky wrote:

If AI technology is available out there and language professionals are using it (in many different ways), why not include it in KudoZ as an option? ProZ.com provides tools and opportunities to its members.


Why instead including it?
What are _PROZ_'s reasons for jumping on this hype?
And as said, where's the added value here?

You are going to kill Kudoz, plain and simple.
A few weeks ago I wrote "Kudoz is obsolete" making a joke. Now it's becoming reality.

I am seriously considering not renewing my membership if my money is being used in this way.


Heike Holthaus
Maria Teresa Borges de Almeida
Christopher Schröder
Becca Resnik
Peter Shortall
Barbara Carrara
Michele Fauble
 
expressisverbis
expressisverbis
Portugal
Local time: 15:10
Member (2015)
English to Portuguese
+ ...
Will be really necessary and/or useful? Aug 8, 2023

It can bring some benefits, but I cannot understand the reason for including this in Kudoz. Will this option be really necessary and/or useful?
If I want an explanation and/or a definition of a term I have difficulty with, I already use the AI elsewhere, but I don't use it to translate terminology, for that I use other resources and the clients I work with.
If it is a question of testing AI, I appreciate it but I am not interested because I do not have the skills to do so. I'm not th
... See more
It can bring some benefits, but I cannot understand the reason for including this in Kudoz. Will this option be really necessary and/or useful?
If I want an explanation and/or a definition of a term I have difficulty with, I already use the AI elsewhere, but I don't use it to translate terminology, for that I use other resources and the clients I work with.
If it is a question of testing AI, I appreciate it but I am not interested because I do not have the skills to do so. I'm not the right member. I'm sure there are great other members more competent and the right ones for this.
In any case I believe that on Kudoz AI can return better results than certain often meaningless answers from some members.
I totally agree that Kudoz points earned this way won't be considered for directory ranking.
Collapse


Maria Teresa Borges de Almeida
Becca Resnik
Philip Lees
Barbara Carrara
Beatriz Ramírez de Haro
 
Michele Fauble
Michele Fauble  Identity Verified
United States
Local time: 07:10
Member (2006)
Norwegian to English
+ ...
Inevitable Aug 8, 2023

If AI is all that it is cracked up to be, this was inevitable.

Christopher Schröder
Jennifer Levey
Philip Lees
expressisverbis
 
Philippe Locquet
Philippe Locquet  Identity Verified
Portugal
Local time: 15:10
English to French
+ ...
Clever Aug 8, 2023

I think this is a good idea.
Getting the raw Chat GPT output may help make the difference between synthetic and organic answers, it may deter any cheaters and since the process can be automated, why not. I welcome the option.


Thomas T. Frost
expressisverbis
Aurélien ARPAZ
 
Jennifer Levey
Jennifer Levey  Identity Verified
Chile
Local time: 10:10
Spanish to English
+ ...
As NLP/AI experiments go, ... Aug 8, 2023

... this might turn out to be quite entertaining. At least for those whose livelihood (or, indeed, life) doesn't depend on the provision of translations that are fit for purpose.

As a (retired) translator and (amateur) software developer with some (limited...) NLP experience, I can understand the Proz team's enthusiasm as they seek to 'make waves' with ChatGPT in Kudoz. Bandwagons are built to be ridden on, and from a developer's viewpoint the challenges are no doubt considerable.
... See more
... this might turn out to be quite entertaining. At least for those whose livelihood (or, indeed, life) doesn't depend on the provision of translations that are fit for purpose.

As a (retired) translator and (amateur) software developer with some (limited...) NLP experience, I can understand the Proz team's enthusiasm as they seek to 'make waves' with ChatGPT in Kudoz. Bandwagons are built to be ridden on, and from a developer's viewpoint the challenges are no doubt considerable.

Perhaps the biggest challenge is the need to transpose any random Kudoz question into a well-formed AI proompt which will stand a fair chance of triggering a worthwhile response at the very first attempt.

Given the frequency with which Kudoz askers post questions with next-to-zero context, inadequate tagging of the fields, and even with incorrect language pairs, I suspect that ChatGPT will very often return nothing more useful to the asker than a 500-word essay explaining (in well-nigh perfect, albeit utterly boring, prose...) "that I am (only) a 'language model'" and that it cannot be expected to guess what the question is before trying to find the answer.

In 'classic' Kudoz, when questions are poorly presented potential answerers have recourse to the discussion box and any further input from the asker gives us more to work on as we research our answers. And one answerer's answer often contains clues enabling another answerer to post a supposedly better answer - and so on, in a process not entirely dissimilar to that involved in AI proompt engineering. Unless ChatGPT has access - and can contribute - to that progressive accumulation of context and understanding of the asker's question, its chances of 'success' against human answerers will surely be limited.

As regards the presentation of the AI interface in Kudoz, I suggest that it would be of considerable interest to the NLP/AI geeks here if users could view the proompt submitted to ChatGTP, and thus verify that the question as interpreted by the Kudoz proompt generator (and being answered by ChatGPT) does in fact correspond to the human answerers' understanding of the question following eventual clarification through the human discussion channel.

Coming back to my subject line, I welcome this development as an 'experiment', and it may provide useful insight not only to those seeking to impose AI in all walks of life but also to those who feel challenged by AI and see it as a threat to their livelihoods.

I do, nonetheless, feel that much care is needed in the way this and other AI developments are promoted on the Proz platform(s). I have suggested on the 'other' thread ( https://www2.proz.com/forum/kudoz/363307-proz_should_not_incorporate_ai_into_kudoz.html ) that the Professional Guidelines need to be updated to refer specifically to the use of AI in member's professional relations with clients.

I also suggest that a public statement from the site owner is long overdue, to explain the Proz platform's strategy in defense of the users' and members' interests, in the presence of these rapidly developing AI tools.

JL
Collapse


Zea_Mays
Philip Lees
Barbara Carrara
Endre Both
Heike Holthaus
Maria Teresa Borges de Almeida
Anton Konashenok
 
Michael Beijer
Michael Beijer  Identity Verified
United Kingdom
Local time: 15:10
Member (2009)
Dutch to English
+ ...
count me in Aug 8, 2023

Although I suspect this might be a bad idea, I would like to see first-hand what's going on. I'd therefore like to test it out!

Michael


Thomas T. Frost
Aurélien ARPAZ
 
Philip Lees
Philip Lees  Identity Verified
Greece
Local time: 17:10
Greek to English
Already covered Aug 9, 2023

Before I post a question on KudoZ, I will already have tried to find the answer using a variety of other resources. Currently those resources include ChatGPT.

So for me this adds nothing new to the KudoZ service, and may even detract from it, by including material I have already inspected and dismissed.

I also share the reservations of other posters about how well ChatGPT will fare in the face of unclear and poorly-defined questions. More specifically, I think that ther
... See more
Before I post a question on KudoZ, I will already have tried to find the answer using a variety of other resources. Currently those resources include ChatGPT.

So for me this adds nothing new to the KudoZ service, and may even detract from it, by including material I have already inspected and dismissed.

I also share the reservations of other posters about how well ChatGPT will fare in the face of unclear and poorly-defined questions. More specifically, I think that there is a danger that, rather than asking for specific terms, some people will just post whole sentences and expect to get a full translation back from the AI. (There are people who already do that with the existing KudoZ.) It should be made clear from the start that anyone who does that, does so at their own risk.

I find ChatGPT to be a useful aide in my translation work (and in other areas), but I question its value as an addition to KudoZ. On the other hand, this experiment may prove useful in exposing the weaknesses of ChatGPT and other large language model AIs.
Collapse


Barbara Carrara
Zea_Mays
Tom in London
Christopher Schröder
ibz
expressisverbis
Heike Holthaus
 
Tom in London
Tom in London
United Kingdom
Local time: 15:10
Member (2008)
Italian to English
Non-pro Aug 9, 2023

Can the AI be trained to identify questions that should be classified as non-pro? That would be useful.

Christopher Schröder
Thomas T. Frost
Tony Keily
 
Philip Lees
Philip Lees  Identity Verified
Greece
Local time: 17:10
Greek to English
Good versus bad testers Aug 9, 2023

Lucia Leszinsky wrote:

If you are interested in testing AI in action in KudoZ, please post below. We’d like to work with a group of volunteer testers on improvements to its usefulness and come up with a resource that supports language professionals when helping each other with the translation of terms.

I just had another thought. The people who are most likely to volunteer as testers are those who have a good idea how to phrase a question for ChatGPT in a way that is likely to get a productive answer. They may even already be using ChatGPT in this way.

But those aren't the testers you want, are they? The testers you want are those that don't know how to use ChatGPT properly, those who are apprehensive or even a little afraid of it, because those are the users who theoretically might benefit the most, if the implementation was successful.

For this reason, I'm ruling myself out as a volunteer, as it would be too easy for me to make it work.

But how do you attract the "bad" users as testers? That's a difficult question. Maybe just start throwing the panel in at random, offered as an "option" for KudoZ users that could easily be dismissed (i.e. with a single click) by those who can't be bothered.

There would still be some degree of selection bias, but not so much. It might be a better way to give this feature a fair trial, assuming that's what you want.

[Edited to correct syntax error]

[Edited at 2023-08-09 06:58 GMT]


expressisverbis
Maria Teresa Borges de Almeida
 
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 »

AI set to make waves on KudoZ: volunteer testers needed






TM-Town
Manage your TMs and Terms ... and boost your translation business

Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.

More info »
Wordfast Pro
Translation Memory Software for Any Platform

Exclusive discount for ProZ.com users! Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value

Buy now! »