Posting unedited AI output

Forum Feedback
1

Below is what ChatGPT gave me when I asked it about rates for a portfolio in the range of 2-5M CHF with Lombard loan:

Bank/Platform Margin Interest Rate (CHF) Approx. Custody Fees Details on Interest Rates
Interactive Brokers (IBKR Pro) Max 2.28%, decreases for larger amounts ~0.1% of portfolio value Tiered rate structure; lower rates for >CHF 1 million
Pictet 1% - 5% 0.2%-0.6% custody fees Negotiable, starts at ~1% for large, low-risk portfolios
Julius Baer 1% - 5% ~0.3%-0.6% custody Customized rates for client profile, large portfolios at lower end
VZ VermögensZentrum 3.0% - 3.75% 0.1% custody + monthly fees Low rates possible with significant collateral
Raiffeisen Typically 3.25% - 4.5% Varies by regional bank Lower rates for strong client relationships
Cantonal Banks (varies) ~2.5% - 4.5% 0.15% - 0.35% Rates and fees vary by canton
Swissquote Approx. 3.50% - 4.00% 0.20% per annum Flat rate with flexibility on larger loans
Credit Suisse 3.5% - 4.5% Custody 0.4% - 0.6% Often negotiable for portfolios > CHF 3 million

"For private banks like Pictet and Julius Baer, Lombard loan rates can range widely. Although the starting rate for top clients can be as low as 1% to 2% for exceptionally large and low-risk portfolios, the typical range for most high-net-worth clients lies around 3% to 5%. The final rate depends on factors such as the client’s relationship history, risk profile, and collateral specifics. Rates might differ significantly for smaller portfolios or those with more complex asset compositions, thus leading to the variation noted in the table"

Maybe it gives clues which banks to talk to. Perhaps I’ll start a new thread about private banks and loan rates to get all the information in one place.

2 Likes
2

You are not doing yourself any good by spreading fake news instead of doing the “hard work”. How can ChatGPT say that the interest rate for Raiffeisen is “Typically 3.25% - 4.5%”, when no information about this is publicly available? And since when does IBKR have a custody fee of 0.1% of portfolio value? Did you even read the output?

11 Likes
3

I do not know what the rates are and I stated my source is ChatGPT which is for sure not perfect.

If you have other data please share with the community or at least say something more positive than accusing other contributors of spreading fake news. Thank you

4 Likes
4

I welcome your contribution, although I don’t support your attitude. Please calm down, there is no need for such harsh expressions.

6 Likes
5

I acknowledge I maybe could have written my post more clearly but I do not need the emotions these responses have triggered this morning. I do not know what I did to Oswand but they are aimed at me and not the topic. It would be appreciated if we could agree to avoid that

2 Likes
6

Off topic note: can we agree that GenAI output should be avoided unless it was actually fact checked?

That’s the main issue with those, they sound extremely confident but can be total fiction. That really can quickly devalue the confidence in what’s written on this forum.

16 Likes
7

That’s the main issue I see here (AI being used as a source without check), I wouldn’t see anything personal and I don’t perceive heavy emotions either.

I’d always want to see what the sources the AI uses are before trusting its answers. In their current state, publicly available large language model AIs just recombine existing sources with some hallucination thrown in the lot, I’d not consider them as an actual source, just the compilers of it.

6 Likes
8

In a professional context, you would be responsible for anything you deliver, AI-generated or not. But this is not a professional context, we are just letters on a screen. For me it was clear that the response you posted should be treated as a very questionable one. I have also spotted the statement about the custody fees at IB. I have assumed that it would be clear to everyone that the statements from AI should be treated with caution, so personally I saw it as a “meta-analysis” of AI capabilities. Probably it was not straightforward for everyone, but getting overheated about it is also not warranted.

4 Likes
9

Yes, there is. And I had a very similar reaction upon seeing it.

Posting unvetted AI output on things it doesn’t and can’t know, on things it is known to be bad at (variable quantitative facts in niche areas), is essentially spam. It manages to look plausible (its core strength), but is complete bullshit.

What irks me the most is:

  • it is very well packaged half-truths, the worst kind of misinformation (just without being deliberately malicious)
  • it is nearly free to occupy valuble real estate on my screen with it, and waste everybody’s time for no gain or even a loss

Would anybody like me spaming screenshots of my Google searches? “Hey guys, Google thinks those links are relevant.”

Screenshot_20241110-150723_Fennec
Screenshot_20241110-150723_Fennec1080×1920 325 KB

Screenshot_20241110-150737_Fennec
Screenshot_20241110-150737_Fennec1080×1920 335 KB

Screenshot_20241110-150744_Fennec
Screenshot_20241110-150744_Fennec1080×1920 281 KB

5 Likes
10

No, there is not. Maybe you are fed up by fixing AI bugs, but not everyone is in your situation. We all know what spam is because of our long time experience. AI is new, we need time to adjust to it.

If someone think that someone else is wrong, express your opinion politely. That what concerns me, not the correctness of AI output.

5 Likes
11

This!

You are otherwise fostering a very toxic atmosphere.

The original poster also literally said in the first sentence where this info is from.
No need for personal attacks. They certainly did not post with malicious intent.

If you would tell someone at my workplace to “stop bullshitting you”, when they merely said something incorrect. You would get a call from HR.
Just because we are kind of anonymous here, doesn‘t mean basic decency is off.

3 Likes
12

I’d still question the value add of that kind of post to a technical topic? I’d be fine having a “GenAI” topic, if we want to do meta analysis in which case it would be clear what the goal is.

Here I feel like it does worsen the experience for both forum regulars who need to learn to ignore those but also casual readers, who may not even be aware of the flaw of that kind of output.

(but yes, no name calling please)

1 Like
13

You are right to question it and you are right to do it with respect to OP. I removed it from the original topic, and I hope OP will agree with my decision.

About an introduction of a policy: well, why don’t we try to use common sense, even if we have to go through some iterations sometimes :grimacing:?

5 Likes
14

The AI output was at least flagged as such, so no foul there. I think using AI output can be useful, but as a courtesy, the output should be reviewed before posting to avoid low quality slop from getting through.

2 Likes
15

I agree, it should be self-understood for the reader that text labelled as AI is not trustworthy. The problem I see more: As a reader, such ai-text is a waste of time. For me it ruins the otherwise interesting time in this forum. If I want AI text, I can talk to an AI myself, then I don’t need a forum.

5 Likes
16

I agree with being polite. It is not difficult to be polite.

However @Barto, I would recommend to be a bit cautious with Gen AI. For things where deduction is needed, the Gen AI output is not very reliable.

1 Like
17

True, maybe some haven’t had enough exposure to this. This kind of low effort spam rapidly burns resources of more knowledgeable participants and missinforms the rest. Such behavior needs very clear words, although not unrelated personal attacks.

This works as follows:

  • Low effort poster post wall of text. It is looks somewhat plausible, but full of errors which look as plausible as the rest.
  • Then there are following options:
    1. Knowledgeable participant gives a big effort to rewrite the whole thing with correct information.
    2. Knowledgeable participant calls it utter bullshit that should be marked as misinformation and deleted.
    3. Knowledgeable participants stop giving a fuck, information quality rapidly deteriorates.

@oswand first tried #2, then upon being challenged by “why do you complain, contribute something yourself” got drained by #1.

This negative value contribution is an evolved form of forum vampirism on steroids.

3 Likes
18

Option 4: Knowledgeable participant points out politely that this is quite useless. OP agrees and modifies their posts. If they insist on keeping misleading information, a mod deletes it completely.

And don’t imply that OP is a freeloader, this is simply false.

2 Likes
19

@oswand first response was very much appropriate and in line with your option 4. Upon being pissed upon by the OP of the low effort comment to

they apparently became more direct in regards to the behavior displayed (can’t quote, since you deleted it).

Edit: I want to add, that I value your work towards keeping the forum friendly.

1 Like
20

A reply in Three Acts:

Act One

Er, what do you mean 
? Common sense?

What is this newfangled thing you are talking about? You seem to imply that sense is common 
?

I think I might have to look up this common sense term – maybe on Chat-GPT 
? – as I am about to chime in here.

Intermezzo Uno: Goofy feels reminded in a non-pleasant way of his earlier days of hearding cats in his gig at the big Hoolie: lots of opinionated (mostly) males, not shy about sharing their views and calling everyone else an imbecile (in less polite terms).
There was no other way really than to shut down hard those bullies. The challenge was not to shut down everyone, but I digress 


Act Two

So, sources for content posted were labeled?
Check.

Maybe’s and Perhaps’s were used already in the original post when potentially drawing conclusions from the generated (and labeled!) AI content?
Check.

The person being accused in a very timely manner explains themselves after being accused. Check.

Intermezzo Due: Goofy also experienced a phase at the big Hoolie when opinions where always 
 hm, checked? Does the expressed view fit the diversity goals, does the opinion potentially hurt someone’s feelings (and is that someone in a minority group?).
Handling this was even more challenging than handling the bullies from act I, as, well, you would not want to encourage anyone to step over the boundaries, but what if they speak the truth 


Act Three

Personal Justice Warriors claim their perceived right that this forum, and posters to this forum, must adhere to their personal interpretation of how posts are supposed to be authored.

In a more moderated form, they remind everyone to be cautious with 
 things.

At best, the Sturm im Wasserglas ends here, but I feel the epilogue hasn’t played out yet.

Epilogue

The suggested course of action for everyone to just chill and use common sense was unexpectedly followed by everyone reading up on the topic.

This almost* resulted in a happy ending.

* There were remaining pundits insisting on Being Rightℱ and so this thread continued until infinite eternity.

I’ll see myself out, thank you, to avoid further sloppy low quality contributions on my behalf. :wink:

3 Likes