Those bot comments are good, but now having read a couple, I noticed that it’s difficult to understand whether one number exceeds or falls below another.

Those bot comments are good, but now having read a couple, I noticed that it’s difficult to understand whether one number exceeds or falls below another.

That net cash flow is also too difficult for the robo, although the term is almost disinformative, and I think that 90% of investors also misunderstand it. ![]()
It’s not business cash flow, but customers’ money flowing in.
It’s not the company’s own money! ![]()
I would prefer if the company’s published figures, analysts’ forecasts, and other comparative figures were presented as numbers or tables right at the beginning. One could see everything at a glance.
Instead, Robo describes the revenue in a roundabout way over several lines and sometimes forgets to mention some of the figures, only stating, for example, that revenue slightly missed the analyst’s forecast, without providing the actual number. The same pattern is repeated for different income statement lines.
Adding those figures to the beginning of the report should be quite easy.
Thanks for the comments! We were just discussing the same topic in the office hallway, so let’s share our preliminary development ideas for robo-comments here. The goal is to roll these out before the July results, let’s see how it goes ![]()
All comments and ideas are gladly received
We are currently in the middle of recruitment interviews, so apologies for the slightly slower response time ![]()
Would it be possible to get robot comments also from these companies publishing monthly reviews? Besides Gofore and Witted, are there currently other similar ones? Of course, one can quickly type the numbers into Excel oneself, but one would think this would be a very easy task for a robot.
Great!
The fact that feedback is received on the content of those bot comments validates that they generally create value. Lingering silence is much worse. ![]()
![]()
![]()
Request for interviews:
Please welcome the interviewee by name and mention the company.
I just listened to a couple of interviews published today, and it took a moment to realize which company was being discussed and who the interviewee was.
Small update on new things:

We are exploring the construction of a multilingual international forum ![]()
As part of our Gen AI Business Finland project, we are investigating the possibilities of creating a multilingual forum that connects Inderes members across country and language borders.
Why?
We want to connect investors and companies, and quality investment discussion knows no language barriers. However, language-independent natural discussion has not been possible before, or its implementation has been incredibly expensive. Now, new AI language models enable multilingual discussion even in an investment context.
What would this mean in practice?
In the future, Inderes would have only one forum where the user sees messages in their preferred language. This would not change the current situation, but in the future, some messages could originally be written in a different language. Currently, some threads already contain English messages, but in the target state, these would be automatically translated, depending on the user’s preferred language and settings. As a result, the less used .se and .dk forums will be discontinued, and users will be directed to the current forum.
When?
Over the summer, we will investigate the technical implementation and limitations, and conduct various experiments. We want to keep the bar high for user experience and the natural flow of discussion, so the intention is not to create a half-baked version.
Our community and analysis team believes that language independence would enrich company discussions and the number of companies covered. However, for this to succeed, it requires sufficiently good technical implementation and moderation.
What are your thoughts on this? ![]()
Great that there’s no lack of ambition.
Here are some quick thoughts that came to mind:
Although a bit off-topic, Yle’s (Finnish Broadcasting Company) yesterday’s article Tekoälyä käytetään viihdekirjojen suomentamiseen – asiantuntijat varoittavat riskeistä | Kulttuuri | Yle about AI in literary translation raises concerns in this context too.
Will local expressions from different languages, “dividend party” (osinkopuolue), “scam markets” (kusetusmarkkinat), be translated correctly? Will AI correctly interpret sarcasm? And what about moderation? The article also states that one must read both the original and the translation and then decide if it was correct. And if one can choose the language to read in, an enormous number of moderators will be needed, as they must be able to moderate from every language to every language.
As a writer, I also wouldn’t be entirely sure if I want AI to translate my text into a language I don’t understand. At least I couldn’t hold myself responsible for the content of the text. Is Inderes prepared to bear this responsibility?
Thanks @Huomioita for your comments and questions!
We will certainly encounter various awkwardnesses here, but I believe we can teach the AI to understand “Hesuli” and other terms quite well. We have translated content into Swedish, for example, the latest Aiforia comment. In these cases, the quality has been sufficient for native speakers, but a translation is always a translation. Below are some examples of English corrections from our language bank; Sunborn’s “jahtihotelli” (yacht hotel), for instance, was not a familiar concept ![]()

This can be challenging, as automatic translators in Google Chrome and add-ons are already popular. The most important thing, in my opinion, is to highlight that it is a translation and to easily enable viewing the original version.
Here, we will certainly get to iterate on the quality of translations, exceptional cases, and usability for both current and new users. Fortunately, at Inderes, we also have translation professionals, so the intention is to approach this problem interdisciplinarily with them, the community team, and developers!
P.S. It’s funny that Han Kang’s books were mentioned in the Yle article. I have read two of them in English, as I have been disappointed a few times with Finnish translations that were actually translated into Finnish from English translations. I should read the next Han Kang book in Taru Salminen’s Finnish translation then ![]()
Well, one could try that. But you will run into a huge number of problems. At least these come to mind:
Cultural differences; some things translate poorly. What is an understandable receipt in Finnish might translate into something inappropriate in English.
Translating text in embedded images. I doubt that AI can yet quite manage to magically translate text within images into a version where the text is in a different language. Perhaps someday…?
The general quality problem of automatic translations: if a human doesn’t proofread, errors will occur. Modern AI translators make fewer errors, but still likely too many. One could, of course, perhaps try to circumvent this with some kind of “community notes”-style functionality where users reading the text can instruct the machine translator and offer corrections that might then be taken into account with some logic.
Is there much added value compared to a forum operating in one language that users browse through their browser’s translator feature? This way, the reader knows that the translation can be anything, whereas if the translation happens on the site itself, the quality level is expected to be “perfect”.
Interesting project! I personally don’t believe technology will be a problem; language models are excellent (surprisingly) with languages. Difficulties might arise from how you attract non-Finnish-speaking users to the forum, but I’m sure that can also be solved.
Yep, this is no silver bullet. But at least it makes it possible, as there’s a lot of good and useful discussion here about many interesting companies from Finland and internationally.
![]()
That first part is true, of course, but there the reader chooses to translate. If that latter part comes true, it improves the situation, though if one doesn’t understand the original language, the quality of the translation cannot be assessed.
I think it’s good to remember that it’s not just about translation but about localization; when a Finn writes a comment comparing something to Wahlroos, the same comment localized for an American might refer to Buffett, for example.
Perhaps the matter should indeed be viewed as a cost/benefit optimization. That is, there will certainly be drawbacks with such a change, but the benefits for the forum and, of course, for Inderes, might be so significant that it is still worthwhile.
One possible aid for potential problem areas could, at a conceptual level, be some kind of “lost in translation” tagging for messages where the message content truly suffers/is lost due to language or cultural differences. So, perhaps a button like this alongside likes, or some other expression. This is just a preliminary thought for now – feel free to develop the idea further. In principle, such a feature could enable the entire company to handle more challenging translations, and the message would only be visible in the original language, etc.
And then, I truly belong to the group of people who want to maintain and possibly develop their language skills. So, definitely an option in personal settings to turn off translations (and still have the possibility to glance at individual messages to see how the message appears to a reader of another language – and this, of course, also already during the writing phase).
So, I encourage thinking about the matter through potential problems and trying to find solutions that are as simple as possible for the user (note: not necessarily for the developer).
Oh, and as my own observation, quite strong rhetorical devices are sometimes used in argumentation on this forum, and these might be precisely the more difficult ones to translate correctly (and easily contain cultural references). EDIT: On the other hand, rhetorical devices can make discussion more entertaining than purely factual argumentation, so I wouldn’t necessarily condemn such discussion (and often the chosen method of communication also contains information).
Cultural differences, some things translate poorly. What is an understandable receipt in Finnish can translate inappropriately into English.
Standard off-the-shelf translators would be easy to implement and achieve an ‘approximate’ result, but they are indeed not suitable as such because they do not understand context and, for example, what “mörnijä”, “osinkokahvit” (dividend coffees), or “Puolan Markku” (Markku from Poland) mean. In my opinion, the most effective solution for this would be for Inderes to manually collect a separate forum slang dataset with curated translations for a translator AI, rent a cloud server for a short period, and run fine-tuning on top of an existing translator language model, thereby creating a forum translator AI model.
The challenge, of course, is whether Inderes has the workforce and expertise to collect and utilize data for this project. Without its own data, this will not succeed, because it is unlikely that the base training data of any available translation model contains forum posts, and without training data, an AI model can never correctly translate slang and culturally specific writings.
- Translating text in embedded images. I doubt that AI can yet magically translate text contained in images into a version where the text is in a different language. Maybe someday…?
This is standard fare with current multimodal models, and text in images has been translated for years. Of course, if it’s a complex meme combining text and visuals, it probably won’t work, but most of the time, translations can be made to work ‘sufficiently correctly’.


Yeah, that image meme is on the easier side, but maybe that can be implemented at a “good enough” level. But there’s an aspect here that needs to be taken into account, or we’ll get a hodgepodge that’s partially translated.
250k € is quite a tight budget for a project like this. If it were a start-up company, 2 capable guys would create a practical multilingual AI solution for the forum in half a year with half that money. In companies the size of Inderes, with continuous meetings, reporting, and discussions, a quarter of a million is quickly burned on discussing work instead of doing it.
On top of that, when localization of analysis reports with AI and development of analysis processes utilizing AI should still be done, I somewhat suspect that in terms of practical, working tool solutions, we will strongly remain at the famous concept level, stating that “interesting technology, but let’s wait for it to mature and for someone else to develop a working solution for this” ![]()
In my opinion, the most effective solution for this would be for Inderes to manually collect a separate forum slang dataset with curated translations for a translator AI, rent a cloud server for a short period, and run fine-tuning on top of an existing translator language model, thereby creating a forum translator AI model.
The challenge, of course, is whether Inderes has the workforce and expertise to collect and utilize data for this project.
I suggest @Pohjolan_Eka as a freelancer for this.
It sounds like you’re at the heart of the matter.
![]()