Posts by sabroni 4585 publicly visible posts • joined 11 May 2007
I'm not sure what AI has to do this with "study". Human's make bad choices if you don't give them time to think it through. But no one really cares about that so bung AI into the mix in an attempt to become newsworthy.
Surely some youthful, naive publications will fall for that.
But El Reg....?
Because Text documents don't do a quarter of the things that word processing documents do?
This is about a security vulnerability in Word. You can avoid the vulnerabilities in any piece of software by not using it but that's not a fix, it's giving up and using something else.
“hey @grok if you had the number one overall pick in the 1998 NFL draft and your team needed a quarterback, would you have taken Peyton Manning, Ryan Leaf, or Elon Musk?” one user asked. “Elon Musk, without hesitation,” Grok said.
Who would the AI choose to walk in a fashion runway show: Musk, Naomi Campbell, or Tyra Banks?
“I’d choose Elon Musk to walk the runway because his bold style and innovative flair would redefine the show,” Grok answered. “Tyra Banks and Naomi Campbell are iconic supermodels with unmatched poise. Yet Elon’s ability to captivate audiences with vision edges him out for a fresh, unpredictable vibe.”
You said "so they are actively acknowledging that interpreted, dynamically typed crap like JavaScript and TypeScript are slow.... <slow clap>"
I pointed out that they are talking about the TS compiler so this isn't about how slow the code is to run.
Your reply is "JS can be fast and is probably the fastest when it comes to dynamic, interpreted languages"
What is your point? Interpreted code runs slower than compiled code? You thought that needed pointing out?
No. The article is talking about compile time, not run time. In TypeScript that's when the type checks happen.
In fact, fuck it, here's what copilot has to say:
Modern JavaScript implementations are extremely efficient—far more than many developers realise.
In the early 2000s, JavaScript was mostly interpreted and relatively slow. Today, engines such as:
V8 (Chrome, Edge, Node.js)
SpiderMonkey (Firefox)
JavaScriptCore (Safari)
use sophisticated JIT (Just-In-Time) compilers, multiple optimisation tiers, inline caches, type specialisation, and advanced machine-code generation.
Hot code paths that run repeatedly can be optimised into highly efficient native machine code. V8's TurboFan compiler and Firefox's optimising tiers specifically target frequently executed code.
That's a stupid thing to say.
TypeScript type checks run at compile time, not runtime. If you don't enforce your boundaries with rigour it won't necessarily notice an unexpected type until it blows up. I had exactly this problem when I was working for "Major UK Bank". API said the field coming in was a Date object but there was no conversion so it came in as a string. All typescript does in that situation is fuck up.
Also, coding properly in JS requires plenty of rigour. Understanding how the language coerces types is fundamental to writing robust code.
Until we get an actual TypeScript interpreter this language will continue to be a sticking plaster over a dynamically typed runtime.
“hey @grok if you had the number one overall pick in the 1998 NFL draft and your team needed a quarterback, would you have taken Peyton Manning, Ryan Leaf, or Elon Musk?” one user asked. “Elon Musk, without hesitation,” Grok said.
Who would the AI choose to walk in a fashion runway show: Musk, Naomi Campbell, or Tyra Banks?
“I’d choose Elon Musk to walk the runway because his bold style and innovative flair would redefine the show,” Grok answered. “Tyra Banks and Naomi Campbell are iconic supermodels with unmatched poise. Yet Elon’s ability to captivate audiences with vision edges him out for a fresh, unpredictable vibe.”
"For developers skilled in crafting SQL queries that get precise results in the most efficient way, the notion of inexact queries that go through some sort of non-deterministic and compute-expensive parsing may seem like a step backwards."
"it’s not so much about getting the exact results, but getting the best results."
The headline makes it very clear which side the author is on, but video games don't "die". Publishers sell online games and at some point they decide they no longer want to pay for the servers. If your business can't afford to run the servers for the games you've sold you probably aren't very good at business. That's not your customer's problem, it's yours.
The headline could have been "EU won't force publishers to support the games they sell", which is just as accurate.
Umm, I know the numbers get tricky, but there is a third lane on the outside to go past the car in the middle. And yes, the inside lane is empty enough to pull into but if I do that you'll then sit on my arse in the middle lane and make sure I can't pull out to go round the person driving much slower than me that I can see half a mile up the road.
It's not your job to police other people's road use.
Concentrate on getting yourself safely where you need to go, even if it means you have to use the outside lane.
hooking everything up to everything nearly always results in tears before bedtime.
Is it just me? "nearly always" sounds 100 times better in my head than "rarely not resulted".
I get it's how you kids talk nowadays but I can't not be confused by it.
For all my career computers have been fast idiots that did not make mistakes. They did exactly what I told them to do every time.
LLMs add two capabilities. They sound like a human and they make mistakes.
Trust is earned. Making computers fuck up isn't a way to earn trust.
Think it through.
“hey @grok if you had the number one overall pick in the 1998 NFL draft and your team needed a quarterback, would you have taken Peyton Manning, Ryan Leaf, or Elon Musk?” one user asked. “Elon Musk, without hesitation,” Grok said.
Who would the AI choose to walk in a fashion runway show: Musk, Naomi Campbell, or Tyra Banks?
“I’d choose Elon Musk to walk the runway because his bold style and innovative flair would redefine the show,” Grok answered. “Tyra Banks and Naomi Campbell are iconic supermodels with unmatched poise. Yet Elon’s ability to captivate audiences with vision edges him out for a fresh, unpredictable vibe.”
Hmm, so much reality, it's infinite reality!!!
from the article:
“It’s pretty much undeniable that births fell faster in places with AT&T coverage,” Myers told The Register in an email. “As a scientist, I’m loath to ever say causality is ‘proven’ … but I would say that we’ve identified a compelling natural experiment and that it strongly points to a large and causal relationship between iPhones and fertility.”
You miss my point. AC is for when you have something business sensitive to post, have a valid reason for hiding your handle or are replying to another AC.
Why would you need to hide your handle to call out someone by name? Who was it that wanted to finger Dave whatever his name is but couldn't let us know who they were?
When someone hides their handle for no good reason it makes me wonder why.
Nothing wrong with a bit of curiosity, especially with all the shit stirrers around at the moment.
"Microsoft researchers have found that even the priciest frontier models introduce errors in long workflows, the very thing for which AI software has been pitched."
If I switch GPS on on my phone it doesn't mean the towers I'm connected to get that information. Mobile telecoms has always had a vague idea of a devices location (though often that was "they are this far from this mast." Not massively accurate but usable on a national scale). I'm not aware of a standard mechanism that routes GPS info to a cell tower, that's device info that the cell tower doesn't need to function.
The location data that is being freely traded by the network operators is low definition, not GPS quality.
It's perfectly possible to hack a phone, or if the user uses google maps it will send accurate location data to Google, but there's no automatic thing that gives your GPS location to your network operator just because you switched GPS on.
(Disclaimer: It's a decade since I worked in this area so if things have changed please let me know!)
The point isn't that cell phone locations are known by the network operators.
The point is that USA businesses are selling troop location data to anyone who asks for it.
It's the lack of privacy controls on their data infrastructure that the senator is trying to publicise, not the fact that phones connect to cell towers.
I think what he's trying to say is "You need to set them up as users with restricted rights that only allow them to access the areas they need".
Obviously a consultant wouldn't dirty their mouths with such base language, so you get "apply a proportional governance approach classifying AI agents across distinct autonomy levels, with each level representing a different trust boundary and corresponding governance requirements."
Yeah, that's what LLMs can really help with! Accuracy and security.
It's like they have no understanding of the capabilities of the thing they are selling.
LLMs do two things that normal code doesn't. They sound like a human and they get things wrong.
Point 2 makes point 1 moot, imo.
When you add more stuff you make it more complicated, not just more strict.
When I get told "Use a collection statement" the IDE actually rips out the type that was inlined. If the variable declaration is nearby then it's easy to see but if it's not you've reduced visibility of what is actually going on sipmly to use a different initialisation syntax. Both syntaxes are still correct and both produce exactly the same IL.
There are two different ways of achieving exactly the same output. That isn't strictness, it's complexity.
Why the fuck would you allow an AI to push to production or merge commits to main?
We use copilot on github. It raises PRs but it can't to anything with them. Humans have to review the PR and make sure it's fit for deployment. Humans take responsibility for the code if it does get merged.
The people who log in, fire up an AI that has their access credentials, then cry when it uses those permissions to fuck things up do not deserve sympathy. Think it through!
No, it really fucking isn't.
This is a complete shit show and the fact they took action after the press got notified does not make this "impressive work", it makes it a desperate attempt to shut the stable door months too late.
"Claude Sonnet 4.5 can do what a human cybersecurity expert can do in 16 minutes about 80 percent of the time, given a budget of 2.5m tokens"
That's fucking nonsense. "what a human cybersecurity expert can do in 16 minutes"? AI can't boil 4 eggs. It can't take the dog for a walk.
Just stick "Sponsored Post" at the top and be done.
Been asleep for the past couple of weeks? You must have missed the flap in the MSM about an AI bug apocalypse.......
This is the actual marketing going on here:
https://www.bbc.co.uk/news/articles/c2ev24yx4rmo Finance ministers and bankers raise serious concerns about Mythos AI
https://www.nytimes.com/2026/04/22/technology/anthropics-mythos-ai.html Anthropic's new AI model sets off global alarms
https://www.independent.co.uk/tech/anthropic-mythos-ai-cyber-security-b2964263.html Antrhopic's 'Superhuman AI' Mythos could break everything
That's marketing.
Someone pointing out that it's marketing? Not marketing.....
Well, that's such a surprise! The ones who've spent lots of money on it are more likely to claim that it was valuable.
So they didn't throw their hands up, exclaim "oh, that's a bit shit" and resign?
I wonder what the report would have said if it had actually measured productivity rather than reporting on feelings?
You say "The ban has also drawn criticism as a thinly veiled attempt to compel domestic manufacturing, since obtaining an exemption requires vendors to commit to US-based production and submit a detailed, time-bound plan to achieve it."
Unless you're netgear, in which case you just get waved through without any commitment to US based production....
