Hacker Newsnew | past | comments | ask | show | jobs | submit | PaulHoule's commentslogin

In the end Iranians who want freedom need to form a different regime but when a group is under attack it rallies around the flag so it is a setback not an opportunity for dissidents.

My wife complains about people complaining about the price of eggs every time the subject comes up because it's her duty as a housewife to know about the price of all the protein sources and they are still a bargain -- who'd have thought that the price of transcribing the receipts would be seen as even more onerous?

Receipt scanning OCR has been around for a long time. Circa 2010 I ran enough HITs on Mechanical Turk [1] that I got my own account representative at AWS and I wondered what other kind of HITs other people were running and thought I would "go native" and try making $100 from Turk.

I am pretty good at making judgements for training sets, I have many times made data sets with 2,000-20,000 judgements; I can sustain the 2000 judgements/day of the median Freebase annotator and manage short burst much higher than that with mild perceptual side effects.

I gave up as a Turk though because the other HITs that were easy to find was the task of accurately transcribing cell phone snaps of mangled, damaged, crumpled, torn, poorly printed, poorly photographed or otherwise defective receipts. I can only imagine that these receipts had been rejected by a rather good classical OCR system. The damage was bad enough I could not honestly say I had done a 100% correct job on any single receipt, as I was being asked to do.

[1] in today's lingo: Multimodal with prompts like "Is this a photograph of an X?" and "Write a headline to describe this image"


I am amused that this in the classic 1955 Asimov story

https://en.wikipedia.org/wiki/Franchise_(short_story)

the protagonist is interviewed as a one-man "focus group" in lieu of a national election and one of the questions he is asked is "What do you think about the price of eggs?" and he said roughly "I have no idea, my wife does the shopping."


Could argue it is the opposite. The more people in a conversation the more you can play social games instead of thinking.

True for humans. AI has the opposite issue — in 1-on-1 there's only one person to impress. Easy to just agree. Add more people and you cant make everyone happy. Thats the whole point

In a n-to-n conversation, it's important to choose which side to be happy. In which side is your boss? :)

Also wait for the correct timing. I still remember 2020 Zoom meetings. A useful tactic was to let everyone discuss the bad options for a while and when the discussion is fading tell the correct one.

I think there was a post of AI plays Mafia https://news.ycombinator.com/item?id=46438599 and perhaps others https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...


"...we are reducing unnecessary Copilot entry points, starting with apps like Snipping Tool, Photos, Widgets and Notepad."

Great!


Great? Maybe! But this doesn't say, "We are removing Copilot from apps."

My personal opinion is "Copilot is pretty good as a chatbot [1] but don't waste your time trying anything multimodal." So I don't mind it at all, in fact I like it enough that I installed the app on my phone. I've got no interest in having it rewrite stuff for me in Word or for LinkedIn though.

On the other hand, Microsoft is famous for killing something good (like OneNote) but spamming the UI with numerous entry points that will make you think "this is some piece of crap that Microsoft is spamming because nobody in their right mind would want it." That they are getting some self-awareness of this is a good sign.

[1] I'd say Google's AI Mode gives consistently better answers (like use "vite-ignore" instead of writing a Vite plugin that doesn't work) than copilot with the reservation that if Google seems to get uncomfortable about a conversation it will end the conversation with a ten pack of search results whereas Copilot tries to simulate a person with healthy boundaries (e.g. "I will help you write a romance story but I won't help you write a sex scene")


My own anecdotal experience is that Copilot doesn't even do a good job as a chatbot. I usually only use it in a few occasions where I don't have access to ChatGPT/Claude.

And I could tell that. In one instance where I asked it to write a script that does a bunch of things, it provided a series of steps to do in the terminal. This is very off my typical experience with other chatbots. I immediately went to Claude which gave me a complete script that does exactly what I need.


Ugh. They horribly borked Notepad. The whole reason I use it is because it's dumb and simple. The moment you change it into a full-featured rich text editor with AI assistants and autocorrect ... you should just make it another app, because it's solving a different problem.

At the very least, don't forget my font setting on the update.


Metrics must have showed disappointing results and they're trying to brand this as a consumer friendly move

"… by making them necessary entry points! Muahahaha!"

> "… by making them necessary entry points! Muahahaha!"

Starting with Windows 11 26H2, the Start Menu will be removed and replaced with Copilot. In order to use a locally hosted app, an externally hosted LLM will need to be instructed to launch it. The reliability is phenomenal: our testing has shown it can launch the right app with 95% accuracy.


What's the accuracy of the current Start Menu interface in real life? There's a certain amount of fat fingering which must be attributed to the software because software decisions affect the fat finger rate. Then there is all the other weirdness, the applications you can search for by name and the ones you don't, search results that appear and sporadically get updated later, etc.

My take though is that Copilot does a better job with bash than it does with CMD.EXE or Powershell so the "AI Natives" will all ask it to install WSL 2 and then tell it to do things there.


Users will also need to drink a Monster™ verification can every time they launch the start menu if they do not have a Premium AI PRO Ultra MAX account. Users may chose to skip verification process if they agree to the new EULA where it is stipulated that they must meet a weekly quota of Big Macs™ stamps. Failing that your Copilot™ Account will enter lock-down mode where a full document, body and facial scan must be "performed" to recover it.

There was a really amusing article in Bloomberg Businessweek a few years ago which pointed out that most of the really big donors just sprayed money at a unicause indiscriminately and that Michael Bloomberg was the only one that showed any sign of investing rationally.

I mentioned that to my wife and she of course rolled her eyes because it seemed so self-serving to her. (Last night we were sitting around the kitchen table and talking about how much better The Economist was than Bloomberg Businessweek and how I finally canceled my subscription to the latter when they hired genius financial writer Matt Levine [1] to write a whole issue boosting crypto in a 200% cringe writing style just before the FTX scandal broke)

[1] ... sent him an email about how sorry I was for him!


If you start looking at "candidate spend" vs "results" you get metrics that .... people don't want to talk about.

Of course the media tending toward "every election is super close, impossible to call, tune in tomorrow" before the election and "it was so obvious he'd win" afterwards doesn't help.


> If you start looking at "candidate spend" vs "results" you get metrics that .... people don't want to talk about

Not if you correct for incumbency. The thing people want to talk about is that money buys elections.


The whole point is that “money buys elections” is what’s under discussion - is it true? Does it, or does the money spent, even if it correlates, not cause?

> whole point is that “money buys elections” is what’s under discussion - is it true?

No. What is true is candidates and NGOs push this angle, and folks who lose elections like it as an explanation for why they aren’t represented.


See

https://journals.sagepub.com/doi/10.1177/21582440241279659

it's not necessarily straightforward that "more fundraising => win" because "better candidate => more fund raising". Like definitely if a candidate gets people excited they are going to raise more small money donations and some big donors are sensible, though of course one senseless whale can blow out the numbers. [1]

Note Clinton and Harris outraised Trump by large margins in 2016 and 2024

https://en.wikipedia.org/wiki/Fundraising_in_the_2024_United...

[1] as someone who has run third party candidates for office I am going to push back on some of the discourse around access because in most places the restrictions aren't that bad and if you find it hard to get enough signatures on the ballot and find it hard to get at least some money from donors you are going to find it hard to get votes


3-electrode EEG devices don't really work and that's one of the reasons why biofeedback was a fad of the 1970s. There are some very slick devices out there now like

https://choosemuse.com/

but at that price it is not going to replace the Polar H10 in my biosignals kit, the respiration radar or the GSR and EMG sensors -- and any of those hidden under coat can tell my phone to tell me that I tilted before I realize it on my own.


It seems to be little known you can not file your tax returns for years and usually nothing happens but in the current situation I suspect that some years could get added to those years unless like, they found out I wrote some post here where I said something like "Trump wasn't the best president of all time" and then they decide to come rip up all my floorboards looking for gold bars.

This is what makes RISC-V so much fun.

You might never be able to get a RISC-V laptop that can compete with an ARM laptop, but you sure can take a RISC-V core and modify it in whatever way you can imagine.


>You might never be able to get a RISC-V laptop that can compete with an ARM laptop

If we are to believe Tenstorrent's roadmaps, the expectation for parity is 2028.


When I was involved it was an x86 machine in a rack in Rhodes Hall.

I had a copy of the whole thing under my desk though in Olin Library on a Pentium 3 machine from IBM that was built like a piece of military hardware. In April the sun would shine in the windows of my office, the HVAC system was unable to cool my office, and temperatures would soar above 100F and I'd be sitting there in a tank top and drinking a lot of water and sports drinks and visitors would ask me how I could stand it.


Thanks for confirming. We need to stop marketing for AWS by talking about the ability to use the internet in AWS branded product terms.

The S3 API/UX/cost model is so seductively simple for static hosting though. I kind of think they deserve their ubiquity. Not on 90% of their products though.

It's great for some applications, like to serve up the QR codes for this system

https://mastodon.social/@UP8/116086491667959840

I could even make those cards tradeable like NFTs, use DynamoDB as the ledger, and not worry about the cost at all.

On the other hand if you are talking about something bandwidth heavy forget about AWS. Video hosting with Cloudfront doesn't seem that difficult, even developing a YouTube clone where anybody could upload a video and it gets hosted seems like a moderate sized project. But with the bandwidth meter always running that kind of system could put you into the poorhouse pretty quickly if it caught on. Much of why YouTube doesn't have competition is exactly that: Google's costs are very low and they have an established system of monetization.

I am keeping my photo albums on Behance rather than self-hosting because I lost enough money on a big photo site in AWS that it drove my wife furious and it took me a few years to pay off the debt.


> I lost enough money on a big photo site in AWS

I’m sorry what. This is supposed to persuade me?


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: