malshe

About a year and half ago I frequently used ChatGPT for speech to text conversion followed by summarizing the text because I ramble. Although the people receiving the text knew this, I absolutely hated AI's writing style. So I fine-tuned GPT 4o on about 500 short paragraphs and created a desktop app only for my use. It worked fine until about three months ago. It just can't handle the atrocious writing by latest GPT and Claude models. I tried to fine-tune newer models on HF but there is no way I can get rid of the cringy writing style. I even tried converting GPT 5.6 Sol's writing to GPT 4o and then using my app. Nothing works.

I see your point but it's all AI written and I doubt any human actually read it to ascertain accuracy

Gift link: https://www.wsj.com/business/autos/ev-maker-polestar-says-tr...

Link to the open access article: https://link.springer.com/article/10.1007/s11357-024-01307-9

Oh thanks for letting me know. I just used the link that was sent to me as I don’t have a NYT subscription. I hope the archive link works for you though.

Gift link: https://www.nytimes.com/2026/08/10/us/ai-cheating-online-deg...

Archive: https://archive.ph/y08aD

I can't stand it either. I am usually quite optimistic about AI except when it comes to writing. My gloomy prediction is that soon kids will start talking like the way AI writes.

Gift link: https://www.wsj.com/tech/steve-jobs-apple-next-cia-161b65f9?...

> US has the power to just tell any creditor that they won't be repaying

To think about it, we have the perfect president who has been telling creditors to go pound sand for decades: https://www.usatoday.com/story/news/politics/elections/2016/...

Gift link: https://www.wsj.com/finance/regulation/investors-slam-sec-pl...

It's written by AI so it's positive

I use it locally with parquet files

Totally understandable. No human working at BBC will ever write a neutral or positive story about India

A couple of weeks back I was flying out of Chicago O'Hare. The guy in front of me had a carryon suitcase, a laptop bag, and a fanny pack attached to the carryon. The employee checking us in told him that the fanny pack was considered a second personal item. He had to remove it from the bag and wear it on him. Then it became part of his attire and that was A-OK.

Tough to write a title for this post staying within the character limit. On Twitter I see the English translation as follows. This will give you the context if you didn’t already click the link:

Photo of the eclipse taken today with my Xiaomi 17 Ultra and a lens protective filter. Do you notice something weird? Well, yes, the moon was the other one. It detected a bright sphere in the sky, went into "supermoon" mode, and straight up invented the entire surface

Xcancel link: https://xcancel.com/cvillasante/status/2087657106342105228

you are complaining about a problem with this paper without explaining what it is.

The main text is only 23 pages double-spaced. The remaining manuscript comprises images, tables, and references. It's not that long really.

A little OT but I recall Marc Andreessen went on a blocking spree on Twitter when it was still Twitter. He would block people and then tweet about it.

Angry Birds, of course

Gift link: https://www.wsj.com/finance/investing/how-this-100-year-old-...

I am working on a project where we have to classify customer calls into more than 10 categories. As the client wants everything locally I tried a few local LLMs. Gemma turned out to be the best model for this task. The classification accuracy is impressive, and the client is happy that I am using an American model.

This is rich considering LLMs can't even write like an average person let alone famous authors

Same here. I have my 2013 MacBook Pro still going strong. My father uses it when he visits me.

Thanks! I will try this

Knowing Meta anything is possible. It’s not like they never shafted their paid customers. They have been overcharging advertisers by showing wrong metrics for years. I’m sure at some point they will come out with raised hands and admit to this “glitch” they found during an internal review.

There are two separate content moderation issues at play here: user-generated content and Meta & advertisers-generated content. The moderation challenge related to the fomer is much harder. But this article is about the latter. The researchers literally found these ads in Meta's own ad library. Meta sells these ads so they must content moderate them. There is no excuse.

I just don't get their algorithm. I used to show YT videos to my students but I stopped that totally because of this kind of crap.

I think people have been demanding this for ages but lobbying by social media companies against any change in the law is enormous.

As for the cost of content moderation, Meta has zero interest in any type of moderation to begin with. The cost is just a public-facing excuse. In Careless People, Wynn-Williams mentions that Facebook didn't do anything to stop anti rohingya posts in Myanmar because they did not have any fact checkers who could read Burmese. In other words, they did not have a single moderator in the damn country.