Lightnews — Scholar-powered news

猫好きな人 @valoisdubins.bsky.social · 11m

Because that’s a key element of the book. It’s also an experience you cannot truly grasp with a summary.

That’s even if the summary includes it, which LLMs will not, unless you ask them to produce an answer that plausibly resembles a critique of the book, and then you’re basically gambling.

猫好きな人 @valoisdubins.bsky.social · 57m

I’ve replied back with ChatGPT being asked if they’re sure about that and ChatGPT obediently saying “sorry, that was wrong…”

1 1

猫好きな人 @valoisdubins.bsky.social · 58m

A thought experiment.

Imagine a student is assigned to read Jordan Peterson’s Magnum Opus “Maps of Meaning”.

They use an LLM to summarise the text, available online.

Will the AI tell them “this is so full of non sequiturs, it’s essentially verbose pretentious gibberish”?

猫好きな人 @valoisdubins.bsky.social · 1h

Absolutely. It’s stunning how often Google AI links to evidence that simply doesn’t say what it claims. (Even if the claim may be true)

And as many people point out, students using AI to summarise difficult texts means students don’t practice *reading and parsing and analysing difficult texts*. 1/2

1 2

猫好きな人 @valoisdubins.bsky.social · 1h

This comes from the tech prejudice that all education is merely the transference of digitally codifiable data from the brain of the teacher to the brain of the student. 2/2

1

猫好きな人 @valoisdubins.bsky.social · 1h

Absolutely. It’s stunning how often Google AI links to evidence that simply doesn’t say what it claims. (Even if the claim may be true)

And as many people point out, students using AI to summarise difficult texts means students don’t practice *reading and parsing and analysing difficult texts*. 1/2

1 1

猫好きな人 @valoisdubins.bsky.social · 1h

Truly the work of people who don’t understand how the world works aka podcasting grifters

derek guy @dieworkwear.bsky.social · 1h

About a month ago, the Trump administration got rid of the de minimis exemption, whereby packages valued under $800 could slide in without import duties. Now there's a backlog as the government can't process all of this paperwork, leading to UPS just destroying packages

Business Insider headline reads: UPS is telling customers that their packages coming to the US are marked for destruction.

4

猫好きな人 @valoisdubins.bsky.social · 2h

This is a subtweet for the pricks and assholes who harassed Jenn Stubbs off BlueSky *after* she deleted and apologised.

2

猫好きな人 @valoisdubins.bsky.social · 2h

Add Subu Vedam to the list of names to keep alive.

Lynn Becker @cmgiulini.bsky.social · 4h

After spending 43 years in prison for a crime he didn't commit, evidence hidden by the prosecution reversed his conviction. Rather than finally enjoying freedom, ICE abducted him for deportation

Depraved.

www.miamiherald.com/news/local/i...

He was wrongfully imprisoned for 43 years. Moments after being released, ICE took him

Subramanyam “Subu” Vedam now faces deportation.

www.miamiherald.com

1 3

猫好きな人 @valoisdubins.bsky.social · 3h

Absolutely. This is more than about principle. Not a fan of Takaichi for obvious reasons, but she is no worse than Aso in that respect, and probably more competent.

There’s an electoral calculation too.

Hikosaemon @hikosaemon.bsky.social · 3h

It is intriguing to consider after EVERYTHING Komeito put up with from the LDP for 26 years, that it was the appointment of Takaichi that broke the camel's back.

An End to 26 Years of LDP-Kōmeitō Cooperation www.nippon.com/en/japan-dat...

An End to 26 Years of LDP-Kōmeitō Cooperation

On October 10, 2025, junior coalition partner Kōmeitō announced it would leave the minority government and put an end to its cooperation with the Liberal Democratic Party. New LDP President Takaichi S...

www.nippon.com

1

猫好きな人 @valoisdubins.bsky.social · 3h

It’s bullshit. She didn’t double down. She deleted the original post and apologised.

And then cancelled her account because some people are more interested in being abusive than in solving the problem of disinfo and continued harassing her.

1

Reposted by 猫好きな人

Alyssa the Kate @alyssa.codes · 5h

Fresh posts from big accounts said "wow she locked comments" when I could see it was restricted to followers at that moment.

It *is* ironic that, with the goal to correct misinfo, people ignorant of how Bluesky works jumped to conclusions and spread misinfo as a result. It's bad for the platform.

1 2

猫好きな人 @valoisdubins.bsky.social · 3h

She did admit her mistake.

Assholes kept harassing her despite that because they need to express dominance over others

And so a journalist who does campaigning reporting against ICE abuses was forced off the app.

猫好きな人 @valoisdubins.bsky.social · 3h

The apparent u-turn on gutting the CDC disease monitoring might actually be a deliberate ploy to make people think it isn’t happening.

When they do this sort of thing several times, (and have) it starts to look like a conscious tactic.

criticalpressure.bsky.social @criticalpressure.bsky.social · 3h

According to the last couple of paragraphs, they're not all being asked back. It's another "it's 100% tariffs, no now it's only 20%, doesn't that make us seem more reasonable since it's not 100%?" move. Classic abuser.

The agency’s entire Washington office, which was laid off on Friday, will not be rehired. Nor will employees of the office of the director of the center for injury prevention, or those at the division of violence prevention policy.

“This is going to be devastating to Americans and to the global community,” said Dr. Debra Houry, who served as the agency’s chief medical officer before she resigned in August in protest against the administration’s policies.

“They are dismantling public health,” she added.

1

猫好きな人 @valoisdubins.bsky.social · 3h

You get British people believing that parts of the UK are governed by Sharia law, so yeah, I think there’s a swathe of Americans who believe this.

1

猫好きな人 @valoisdubins.bsky.social · 3h

In general people have a bad habit of assessing AI according to what they want to prove. A lot of the time it’s that genAI is *good* so they isolate gains thus turn losses into externalities.

But it works the other way too. Dismissing genAI wholesale as never useful doesn’t stand up either.

3

猫好きな人 @valoisdubins.bsky.social · 3h

🔖 list of problems of AI in education

猫好きな人 @valoisdubins.bsky.social · 3h

“I produced a report so quickly with AI”

Well done, go report a massive LLM time saving.

And you pass on the flaws and errors of the LLM-written report (which likely exist because of how LLMs inherently work) to someone else to lose time on fixing.

Research needs to be holistically focussed.

1 3

猫好きな人 @valoisdubins.bsky.social · 3h

And you may say, well, we don’t train our experienced programmers - we don’t rely on a feeder system like that. People move from company to company all the time.

So look holistically: the industry is losing its training programmes.

1 4

猫好きな人 @valoisdubins.bsky.social · 3h

While I’m at it, another error is the determination not to view tasks holistically, but to isolate supposed gains.

eg if you replace entry-level coding jobs with AI, you may save money immediately on those tasks, but you have also just lost your training programme for experience programmers.

猫好きな人 @valoisdubins.bsky.social · 4h

One error in thinking efficiency gains will be huge is forgetting that a lot of the tech we already have already does a fair bit of the actually useful stuff that Gen AI can do. GenAI offers a marginal improvement on that tech.

But some compare to a world where none of the previous tech exists.

Dare Obasanjo @carnage4life.bsky.social · 5h

The corporate myth is that embracing AI will get you results that are cheap, fast and good with fewer employees.

The reality is a lot of companies will find out they got things done for a lot cheaper and a lot faster but if definitely wasn’t good.

2 2 9

猫好きな人 @valoisdubins.bsky.social · 3h

Putting one’s feelings about Silver aside:
This appears to be representative of many people’s experiences.

LLMs offer some improvements in some specific tasks but not good work wholesale.

Otoh, they often produce bad work but with the signifiers of good work, and that is an *inefficiency*.

猫好きな人 @valoisdubins.bsky.social · 4h

And also:

Genetic [edit]
Research into genetic associations in antisocial personality disorder suggests that ASPD has some or even a strong genetic basis. The prevalence of ASPD is higher in people related to someone with the disorder. Twin studies, which are designed to discern between genetic and environmental effects, have reported significant genetic influences on antisocial behavior and conduct

1

Reposted by 猫好きな人

Dare Obasanjo @carnage4life.bsky.social · 5h

What in the nepo baby is this?

Barron Trump tipped for top TikTok job

US president’s 19-year-old son could be appointed to the app’s board

www.telegraph.co.uk

8 10 83

猫好きな人 @valoisdubins.bsky.social · 4h

I think one reason the writer’s strike ended when it did was when executives realised AI simply couldn’t produce scripts as good as humans could.

1

猫好きな人 @valoisdubins.bsky.social · 4h

One error in thinking efficiency gains will be huge is forgetting that a lot of the tech we already have already does a fair bit of the actually useful stuff that Gen AI can do. GenAI offers a marginal improvement on that tech.

But some compare to a world where none of the previous tech exists.

Dare Obasanjo @carnage4life.bsky.social · 5h

The corporate myth is that embracing AI will get you results that are cheap, fast and good with fewer employees.

The reality is a lot of companies will find out they got things done for a lot cheaper and a lot faster but if definitely wasn’t good.

1 1 5

猫好きな人 @valoisdubins.bsky.social · 4h

Anyways, he’s blocked me, which is a shame, as I was about to correct his Ukrainian, for shits and giggles.

1