Mark Crowley

So I would advise against using this service, it's very good at what it's doing and it will be very tempting to simply use it instead of reading, will will make you less well informed.

If everyone starts using services like this, instead of reading, even really short articles like this that are actually written by human beings, the acceleration of misunderstanding, apathy, and manipulation of knowledge will only accelerate.

Again, the main details are present from the article in the summary, but the result would be an overconfidence that you've now absorbed the content of the article by listening to the summary without really learning much.

But reading the text as I listened and trying to find any words that anchored what they were saying, I found very little and instead found it that they expanded out from the knowledge that the AI has in its own background.

Even worse the very naturalness of the conversation as if it was highly polished podcast, initially gave him a great confidence. That this was a good summary. I even wondered for a few moments if this was actually a pre-recorded summary of the article created by the journalist who wrote it.

This was fascinating and I'd never heard of this kind of thing before. In this context. All that detail was skipped completely in The voice summary.

In contrast, if you read the article, the first three or four paragraphs are quite specific and detailed talking about this case-study of a person who unknowingly had strokes that led to a change in their behavior.

Now this conversation was very engaging, like listening to an interview on a good NPR or CBC radio show.

But it was also a very short conversation.

The technical details were extremely brief and the bulk of the time was spent explaining standard conclusions hinted at in the last paragraphs.

What it created was a very pleasant conversation where I could listen to two generated voices, a male interviewer and a female expert, who summarized the main points of the article through their conversation.

For this particular article on my phone today, December 2nd, 2025 (always keep track of what version of the AI you are using, kids) I used this feature for the first time.

In short, I think it's the worst example of a very effective summarizer being used to make society worse. It's a little terrifying how easy it is, and how nicely it summarizes the article.

Does ‘laziness’ start in the brain? Understanding the surprising mechanism behind apathy can help unlock scientific ways to boost your motivation

Great article, worth the read if you or someone you know struggle with sudden, growing apathy in life.

But I want to comment briefly on the new AI based "Listen to this article" feature in Chrome.

www.theguardian.com/books/2025/n...

‘Traitor’: US representatives call for Trump envoy Witkoff to be fired after leaked Kremlin call Republicans and Democrats warn Witkoff ‘cannot be trusted’ after reportedly advising officials on peace plan

"That’s what a dealmaker does. You’ve got to say look, they want this, you’ve got to convince them of this,” Trump said.
www.theguardian.com/us-news/2025...

This is a real disconnect when you treat everything like an amoral business deal. This is different, there a right side and wrong side here.

The excessive restraint that has been shown to avoid open conflict with Russia is evidence enough that NATO has absolutely no plans on invading Russia. So why would Ukraine joining NATO make any difference?

Because Russia sees Ukraine as still a part of them.

But they are wrong.

Europe scrambles to join Ukraine talks as EU nations attempt to bolster militaries – Europe live Macron due to unveil proposals for a new voluntary military service to boost the country’s defences without having to return to regular conscription

"Nato membership, western troops in Ukraine unacceptable for Russia, officials say"

Why should say simply: "Tough, suck it up."

Russia's proclaimed red line assumes NATO wants to invade Russia. It's very clear that is the last thing they want.

www.theguardian.com/world/live/2...

Zelenskyy says Ukraine has impossible choice as Trump pushes plan to end war US president demands that Kyiv accepts plan that would mean giving up territory to Russia

No capitulation to invaders. This is a cowardly move by America. Weak. Pathetic.

www.theguardian.com/world/2025/n...

ARC Prize - Play the Game Easy for humans, hard for AI. Try ARC-AGI.

ARC Prize Daily Puzzle
Task: 1b8318e3

⏱️🟥🟥🟥🟥🟥 7:33 sec
🤔🟥🟥🟥🟥🟥 6 attempts

Can you solve it?
arcprize.org/play

I'm going to read this.

I don't know what word to use for this except evil

I feel your sadness. I hope things are going better for you now.

Yes, it's disheartening. In my opinion, the scale is off. Given how extreme the actions are, in many democracies this would lead to ongoing national strikes.

Agreed, that's most shocking thing about this looking from the outside from Canada, and I think Europeans feel the same. That's it's so easy for THE anti-monarchy nation to roll into leader worship. It's truly disappointing how half the country and most of the media are just rolling over.

Doh, I'd have taken it.

UNTIL IT’S DONE, Ep. 5: Vito Marcantonio

There are many who dismiss our vision for New York as impossible. To them, I say we need look only to our past for proof of how we can shape the future.

Tomorrow is Election Day. And this is the final Until It's Done of our campaign.

Oh. Ohhhhh!

Lost but winning, never a good combination.

Agree. To your point, I think that logic and intuitive understanding would be different things.

Another angle, what does "understanding" imply about truth? Do I *understand* that 2 + 2 = 5? Or do I *know* it?
I need to read more cognitive science, basically.

Furthermore, I'd say there's nothing in Transformers or the fundamental approach of semi-supervised or supervised learning learning that can solve this problem of truth in the immediate future.

In other words, more research is required. Until then, you should use these systems with this in mind.

Language models cannot reliably distinguish belief from knowledge and fact - Nature Machine Intelligence Suzgun et al. find that current large language models cannot reliably distinguish between belief, knowledge and fact, raising concerns for their use in healthcare, law and journalism, where such disti...

As I've always been saying, LLMs have no relation to Truth at all. They can answer questions, generate ideas on a theme, and many other amazing things, but they don't know the difference between true and false, opinion and hyperbole. These results support that.

www.nature.com/articles/s42...

I came here only to agree and make sure the top of that list is Watchtower, but I see my work was already done. All is well. Totally agree with the list, though I don't grok #5. Would also throw in Hallelujah as a strong contender.