| Pages in topic: [1 2] > | ChatGPT and the ethics Thread poster: Hans Lenting
|
|---|
Dutch newspaper nrc about ChatGPT today (translated by DeepL):
Irresponsible
Not everyone is enthusiastic about it. Tech philosopher Marleen Stikker thinks it is moving too fast. "It is premature that Microsoft is going to integrate OpenAI's technology, which is still in its infancy, into its software on a large scale. If you look at the problems around bias and disinformation and copyright issues with DALLE-2, this is irresponsible."
Professor of Technology and Society Tamar Sharon (Radboud University) is also concerned. "This is AI that is unsafe for now and raises ethical questions. You would want to have a broad public debate and develop regulatory legislation. In the meantime, you should declare a moratorium on the implementation of this kind of AI."
But that's not even possible at her own university, Sharon says. "We recently switched to Microsoft's cloud platform. So you are at the mercy of a tech giant as a public player, and we are now seeing what the dangers of that are."
Stikker believes legislation should prevent an AI like ChatGPT from being used to produce disinformation on a large scale. "If you look at how careful we are with food and medication. Before you can put something on the market, we have to check very carefully whether it is safe. The same should happen with artificial intelligence. It is time for a morgue on this AI, which is not yet ready to be marketed. European Union legislation is too slow and cautious."
Implementing ChatGPT in a program like Word or a search engine like Bing, Sharon sees as dangerous. "There is a big difference with an encyclopedia like WikiPedia. ChatGPT can only produce language, but has no knowledge. You can philosophize about what is knowledge and truth, but WikiPedia is based on the principle of shared knowledge and fact checking. ChatGPT is purely a tool that generates text without knowing what it actually says."
For Sharon, it is still too much about the tool itself and what it can do, and too little about the "political economy" behind it. "By that I mean the billionaires who invest in it," Sharon says, "and who market such technology to make a profit. In the case of ChatGPT, that's an event with a huge social impact, without any democratic control over it. That shouldn't be possible."
| | | | Samuel Murray 
Netherlands
Local time: 01:15
Member (2006)
English to Afrikaans
+ ...
DeepL wrote:
It is time for a morgue on this AI...
Curious to see that not only humans confuse mortuarium and moratorium.
| | | | Hans Lenting
Netherlands
Member (2006)
German to Dutch
TOPIC STARTER
Deleted
[Edited at 2023-01-25 07:49 GMT]
| | | | | Thank you for posting this | Jan 25, 2023 |
It puts the issues concisely and clearly, and it is definitely relevant to translators - it raises once again the issue of form over content. Or the accuracy of a message, whether translated or not.
While we are checking commas and grammar, we must never lose sight of what the text actually says!
| | |
|
|
|
| [OT:] Mortuarium | Jan 25, 2023 |
Samuel Murray wrote: DeepL wrote:
It is time for a morgue on this AI... Curious to see that not only humans confuse mortuarium and moratorium.
Actually, DeepL was not "intelligent" enough to "realize" that a human had made this mistake. Quoting the original Dutch:
"„...In de tussentijd zou je een moratorium moeten afkondigen over de implementatie van dit soort AI.” [...] Het is tijd voor een mortuarium op deze AI, die nog niet klaar is om op de markt gebracht te worden."
(https://www.nrc.nl/nieuws/2023/01/24/microsoft-wil-aan-ai-imperium-bouwen-a4155211)
| | | | | How to use ChatGPT | Jan 26, 2023 |
Just my view on that, anyway.
I have a weakness when it comes to abbreviations and acronyms. Even if the meaning is fairly obvious, I sometimes fail to see it, and even when I've seen one before, I tend to forget it. So this is an area where I often find myself spending time looking stuff up. If you look at the KudoZ questions I've posted over the years, you'll see that a lot of them fall into this category.
This is not specifically a source language problem, by the way... See more Just my view on that, anyway.
I have a weakness when it comes to abbreviations and acronyms. Even if the meaning is fairly obvious, I sometimes fail to see it, and even when I've seen one before, I tend to forget it. So this is an area where I often find myself spending time looking stuff up. If you look at the KudoZ questions I've posted over the years, you'll see that a lot of them fall into this category.
This is not specifically a source language problem, by the way. I also do a lot of substantive editing of medical and technical texts and I sometimes need to check undefined or inconsistent abbreviations in English.
I had such a case a few days ago, and it occurred to me to seek the help of ChatGPT. So I asked it:
What do the abbreviations CSR, CTD 2.5 and 2.7 refer to?
Within seconds I had a clear and concise answer. Of course I checked it before proceeding, but it certainly saved me some time.
Another one:
What is the meaning of "Dice" in the following sentence: AI-based performance benchmarks of interest included: accuracy, sensitivity, specificity/recall, AUC (area under curve), precision, Dice, intersection over union (IoU) and F-1 score.
I hadn't seen "Dice" in this kind of context before and I wasn't sure if it was a proper name that needed the capital D, or some special use of the common noun. Again, ChatGPT was able to save me time. It's answer is impressively comprehensive:
In this context, "Dice" is a performance metric used in image segmentation and object detection tasks. It measures the similarity between the predicted segmentation and the ground truth segmentation, and is defined as the size of the intersection divided by the size of the union of the two sets of pixels. A Dice score of 1 indicates a perfect overlap, while a score of 0 indicates no overlap.
Even allowing for checking, it saved me time.
So that's how I'm incorporating ChatGPT into my workflow. I don't really care about what happens in the Microsoft cloud, because I don't use that kind of stuff. YMMD. ▲ Collapse
| | | | Hans Lenting
Netherlands
Member (2006)
German to Dutch
TOPIC STARTER | Sure, when you check the answer: it can be useful | Jan 26, 2023 |
Philip Lees wrote:
Within seconds I had a clear and concise answer. Of course I checked it before proceeding, but it certainly saved me some time.
I agree, but I guess that people who will be using AI integrated in an e-mail client will not interrupt their “flow of typing” to do thorough fact checking. Potentially dangerous.
| | | | Lieven Malaise
Belgium
Local time: 01:15
Member (2020)
French to Dutch
+ ...
| Might be useful | Jan 26, 2023 |
Philip Lees wrote:
Within seconds I had a clear and concise answer. Of course I checked it before proceeding, but it certainly saved me some time.
I gave it a try and I agree. It goes without saying that you have to doublecheck the answers, but it seems clear that this technology can get you way faster on the right track than conventional Google search. I think it might be very useful (in terms of speed) to experienced translators who know to avoid the traps of internet search.
| | |
|
|
|
| Dangerous tool or dangerous user? | Jan 26, 2023 |
Hans Lenting wrote:
I guess that people who will be using AI integrated in an e-mail client will not interrupt their “flow of typing” to do thorough fact checking. Potentially dangerous.
I'm afraid this is one of the hazards of life in the 21st century. Kids should be taught in primary school how to cross-check information they find on line; how to recognise the danger signs of disinformation; how to move outside their own bubbles; how to deal with trolls, etc.
However, as far as I know this isn't happening (possibly because populist politicians who depend on disinformation don't want it to happen).
If somebody who has never had a driving lesson takes a car or a truck and drives it through a busy urban area, injuring or even killing people in the process, is that the fault of the vehicle, or the driver?
Similarly, if people trust the information they get from AIs like ChatGPT without question and bad consequences ensue, is that the fault of the AI, or the user?
In fact, the fault lies in society's failure to keep up with technological developments by providing the necessary education and resources.
But that doesn't mean that we translators, who are aware of the risks, shouldn't make use of the benefits that AI may potentially offer.
| | | | Samuel Murray
Netherlands
Local time: 01:15
Member (2006)
English to Afrikaans
+ ...
| Who is responsible | Jan 26, 2023 |
Philip Lees wrote:
If somebody who has never had a driving lesson takes a car or a truck and drives it through a busy urban area, injuring or even killing people in the process, is that the fault of the vehicle, or the driver?
I used to have the same attitude. We had an e-mail chat group and one rule was that you should trim all content under your signature, but then an Outlook update came out that caused many members' e-mails to start violating that rule without their knowledge or intention, and our (the group administrators') opinion was: it is the responsibility of the user of the software to be familiar with its operation, and so if the e-mail program does something that the user is unaware of, then it's the user's own fault (and he should suffer the consequences).
I have since moderated my view. One simply can't expect of users of software to be aware of all the aspects, limitations, risks etc. of the software. People no longer learn to use software from scratch -- in many cases, they just sort of grow into it and use common sense.
When I first encountered a bicycle staircase (I was an adult at the time), I didn't look for instructions and warnings etc. on the internet etc. I saw how other people use it, had a brief look at the operation and used my common sense to figure out how it works, and then went ahead and tried to use it. Of course, I was unaware of the risks of incorrect usage, until I suffered abrasion injuries a few times. Fortunately, no-one else was injured, but what if they were... shouldn't society (and the staircase designer) have some kind of duty to prevent suffering?
I'm happy that Chat GPT already does this: it has warned me several times about its limitations during our conversations. It does not warn me each time it gives an answer, though. It told me (without blinking) that the first South African in space went to space in 2002 and that the second South African in space went to space in 1993. If I hadn't asked the one question, I would not have known to be suspicious of the other question's answer.
[Edited at 2023-01-26 08:55 GMT]
| | | | | Regulated versus laissez faire | Jan 26, 2023 |
Samuel Murray wrote:
I used to have the same attitude. We had an e-mail chat group and one rule was that you should trim all content under your signature, but then an Outlook update came out that caused many members' e-mails to start violating that rule without their knowledge or intention ...
MS Outaluck quickly broke most of the quoting rules that kept the newsgroup and email group community sane, and was to a large extent responsible for the outbreak of so-called TOFU posting (text over; full quote under). I still trim emails and post my replies beneath the relevant quoted parts, as I believe it makes things easier for the recipient, but it seems that most people nowadays don't know that this used to be the standard at one time.
... and our (the group administrators') opinion was: it is the responsibility of the user of the software to be familiar with its operation, and so if the e-mail program does something that the user is unaware of, then it's the user's own fault (and he should suffer the consequences). I have since moderated my view. One simply can't expect of users of software to be aware of all the aspects, limitations, risks etc. of the software. People no longer learn to use software from scratch -- in many cases, they just sort of grow into it and use common sense.
I accept your points, Samuel, but not your conclusion. No, we cannot expect all email users to be fully aware of how it works, but in the case you mention, if somebody is told, "Your email program is breaking the rules. This is how you fix it." then I think it's reasonable to expect them to do that. That's just part of being a member of a community.
Otherwise you're in the position of someone whose neighbour's dog repeatedly craps on their lawn, and when they complain the neighbour says, "Well, it's a dog. What do you expect?"
| | | | | Scientific author | Jan 27, 2023 |
Here's an article about how ChatGPT is influencing the scientific publishing world.
"... the program can also produce fake scientific abstracts that are convincing enough to fool human reviewers."
It fooled me at first - until I checked the references. This is the kind of area where I expect it to cause the most disr... See more Here's an article about how ChatGPT is influencing the scientific publishing world.
"... the program can also produce fake scientific abstracts that are convincing enough to fool human reviewers."
It fooled me at first - until I checked the references. This is the kind of area where I expect it to cause the most disruption, basically because it's so damn good.
"On Thursday, Holden Thorp, the editor-in-chief of the leading US journal Science, announced an updated editorial policy, banning the use of text from ChatGPT and clarifying that the program could not be listed as an author."
I wonder how they plan to enforce that ban. ▲ Collapse
| | |
|
|
|
Lingua 5B
Bosnia and Herzegovina
Local time: 01:15
Member (2009)
English to Croatian
+ ...
When it starts producing mistakes. If it requires heavy human checks and editing, then it’s technically not an author. LoL It should be more independent than that to be called an author.
I can see more and more instances where GPT was banned, for various reasons. Will this amazing and useful tool end up being allowed anywhere?
| | | | Samuel Murray
Netherlands
Local time: 01:15
Member (2006)
English to Afrikaans
+ ...
| More Chat GPT fun | Jan 28, 2023 |
There's an apocryphal story from the 1970s about a problem solving computer that was given this problem: most accidents on staircases happen at the top or bottom step. The computer replied: simply remove the top and bottom step. Chat GPT hasn't learnt from this.
On the other hand, Chat GPT was smart enough to know that while you can throw a brick through a window, you can't throw a window through a brick:
[Edited at 2023-01-28 11:42 GMT]
| | | | Tom in London
United Kingdom
Local time: 00:15
Member (2008)
Italian to English
| I refuse to sign up for chat gpt | Jan 28, 2023 |
...because the first thing they ask you for is your phone number. They don't say why they want it. So - I'm out.
| | | | | 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 » ChatGPT and the ethics | 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 » |
| | Trados Business Manager Lite | Create customer quotes and invoices from within Trados Studio
Trados Business Manager Lite helps to simplify and speed up some of the daily tasks, such as invoicing and reporting, associated with running your freelance translation business.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
