Mojeek summariser telling me to kill myself

I have tried to bring this up with your mastodon people, but I keep being ignored or get PR answers. (Mojeek: "@sotolf@alpha.polymaths.social @thedoctor@alpha.p…" - Mastodon)

I went to your search and used the summariser on my nick “who is sotolf”

Are you really okay with your site telling people to kill themselves?

It’s a bit less bad now, but it is still linking to the same quote saying it’s about me.

Now What if a person that is suicidal, or have mental issues does an ego search and get something like this served to them. Are you really okay with this?

Take responsibility please.

After reaching out continuously there has been no explanation on how they will make sure that this doesn’t happen to other people who may be in a worse place than I am currently and can take serious damage from things like exactly this.

To just let an AI post stuff on your page is so irresponsible. To not make sure things like this doesn’t happen is irresponsible.

If you let an AI post something in your name, YOU are responsible for what it confabulates.

Hello @sotolf and welcome to the Mojeek community.

Thanks for bring this to our attention. I retried your search and didn’t get the same (as below) but that can happen, so not denying your evidence (posted summary).

Yesterday was a bank holiday here. I apologise we did not get back earlier. We are a very small team.

I’ve looked at our response and some the details myself and from what I understand @Josh has responded in a timely maner, after taking time to look at everything

As @Josh also replied: “We’re aware of problems inherent in summary and are working to improve it. That is why it is turned off by default.” These summaries are only generated upon request after a Web search, and with cited sources.

None of the above are excuses and we will give the matter further consideration.

1 Like

As mentioned in context, then click the link and you will be directed to the quote “a person that would be enormously improved by death” again, sure it’s more hidden this time, still it’s not feeling great.

I’ve looked at our response and some the details myself and from what I understand @Josh has responded in a timely maner, after taking time to look at everything

The only response I got was first “we’ll be looking into it tomorrow” then

We’re aware of problems inherent in summary and are working to improve it. That is why it is turned off by default. Evidence shows that some people appreciate it, and some people do not. Some opinions are strong, notably at this time. On the deletion from the other thread, summaries are only stored for 24h.

This does not have anything to do with my problem even though it’s somehow directed at me, I see nothing else on the mojeek mastodon account.

The problem is still that you are publishing things like “a person that would be enormously improved by death” directed directly at a person, and evade answering to any of it in any substantive way other than a copy paste canned email answer of “we understand that there is problems” no you don’t, you publish on a webpage that I as a person would be better off dead, that has nothing to do with problems with the summary, it’s about you publishing text saying I’d be better off dead.

Please stop the canned PR message, and talk like an actual human being, don’t you understand that this can be immensly harmful to people?

Hiding behind a disclaimer banner does not make you any less responsible about what you choose to publish on your site.

Yes, sorry that wasn’t swift enough for you but bank holidays are days which are conventionally not work days. Also it didn’t seem like something to respond to unilaterally so I put that holding and then waited for today to take it elsewhere.

It’s not about speed, it’s about the content in the message that I got today, as you may have noticed I waited after getting the one mention yesterday, expecting there to be at least some answer that was substantive.

I got another adressed to me today

We’re aware of problems inherent in summary and are working to improve it. That is why it is turned off by default. Evidence shows that some people appreciate it, and some people do not. Some opinions are strong, notably at this time. On the deletion from the other thread, summaries are only stored for 24h.

But it doesn’t seem to be related to any of my issues, but the things other people were talking about in the thread. So I guess you don’t care what you are publishing on your pages, and this is a non-issue to you guys, at least that is the impression I get from every answer.


And here is the current state of the query, we’re back to it being threatening agin…

@sotolf please share your link that produces this summary

Here’s the one I have for the screenshot above: who is sotolf? - Mojeek Search

https:// www mojeek com/search?q=who+is+sotolf%3F&mal=1&kalid=3bdf4ffb59ce7347&fmt=summary

https://www.mojeek.com/search?q=who+is+sotolf&fmt=summary&kalid=ad3396b562ff6c17

Thank you. The AI produces different summary outputs, and in your case it is bad. These LLMs are not fully deterministic which might explain that.

I was using defaults (English, Snippets length of 160) and location is GB. Anything different or other Preference changes for you?

We are looking into this as a higher priority.

1 Like

I just clicked the summary button and got this thing.

Yeah, I get that, an LLM is just fancy autocorrect, but it’s weird that it’s ignoring all the other citations that it finds more important, only focuses on number 6 which is very negative, somehow sees that my old goodreads account liked this quote 7 years ago, and confabulates that this quote is about me.

Now, keep in mind this is not about me, I don’t really care that much about the search result for me personally, I’m in a state where I can treat it as a joke.

But a person that is already in a mentally bad state, if by some twisted chance they get a similar thing, it can have a really bad outcome, even me being in a decently good place got kind of a shock seeing this being written about me. The only thing I really am out for is that you will try to do better and not endangering people just to follow the whole AI fad thing, I get that it’s something that is “needed” to be “relevant” now a days, but it’s a loose cannon, and remember the case with Air Canada or what it was, what an LLM is saying that you publish is something you are responsible for.

2 Likes

I read twice and haven’t seen where it tells you to kill yourself? Could someone point to it?