News for hackers. Stuff that matters.
Re:Markov chains are funnier than LLMs by mdp2021 on Monday August 19, 2024 @04:13AM (#41288941)
> modern LLMs are wrong in a "My uncle doesn't know basic geography" way

Oh, certainly not. I have been using LLM chatbots extensively this week. If your uncle were not just confident while stating falsities, but uncapable of self-assessment, he would have been murdered.

Re:Markov chains are funnier than LLMs by TeMPOraL on Monday August 19, 2024 @04:22AM (#41289009)
Or made a stellar career as a salesman or politician.
Re:Markov chains are funnier than LLMs by mdp2021 on Monday August 19, 2024 @04:29AM (#41289045)
Until you are caught (in good societies).
Re:Markov chains are funnier than LLMs by TeMPOraL on Monday August 19, 2024 @04:33AM (#41289064)
Or fresh ones. Established democracies end up with political structures that self-select for being able to sound right and cooperate right, entirely independent of truth and merit.
Re:Markov chains are funnier than LLMs by mdp2021 on Monday August 19, 2024 @05:00AM (#41289217)
I coincidentally just stumbled on a recent implicit comment by Angél Boligan Corbo:

Title: "The Talker" - https://image.cagle.com/287573/980/-the-talker.png

Re:Markov chains are funnier than LLMs by sitkack on Tuesday August 20, 2024 @01:36PM (#41302101)
That sounds like all bureaucracies.
Re:Markov chains are funnier than LLMs by Kiro on Monday August 19, 2024 @04:54AM (#41289187)
> I have been using LLM chatbots extensively this week.

This reads as "I've actually used LLMs" like it's a rare occurrence giving you a unique insight, when most of us have been using LLMs extensively for almost two years now.

Re:Markov chains are funnier than LLMs by mdp2021 on Monday August 19, 2024 @04:58AM (#41289205)
> This reads

Oh please. Guidelines: «respond to the strongest plausible interpretation».

Does the form "An occasion of extensive use of the main current chatbots for circumstantial reasons in the past few days confirmed and reinforced the assessments raised years ago already" work better for you?

Re:Markov chains are funnier than LLMs by wizzwizz4 on Monday August 19, 2024 @05:17AM (#41289306)
I think most of us haven't, but those who do are vocal about it. (Those who haven't aren't all that vocal, excluding weirdos like me.) Only 62% of respondents to the 2024 Stack Overflow Developer Survey are “currently [using] AI tools in [their] development process” (up from 44% last year), and non-programmers seem to use them far less.