Posts by DrSunshine0104

Americans give all-day school phone bans a ringing endorsement

Re: "Older Americans, meanwhile, appear increasingly comfortable [with the ban]"

Well for this Millennial American, no phone in the classroom was always a policy and would be confiscated for at least the entire school day. Sometimes, until the parent came and retrieved the phone if the student had been reprimanded several times. I guess I didn't realize there was a cultural shift to just accepting phones being used during class in the first place, but I don't have children to know these things...

OpenAI says 97.9 percent of its employees are now using agents

The /might/ learn something from looking at their own employees but there is a lot of variables that make actually learning something nearly impossible. It is a self-selected group and there is probably social pressure (consciously or not) to use it, and only one LLM is likely available. It is definitely not a good sample of the general populace...

Why Amazon hates 'human-in-the-loop' AI governance

I know he thinks this is a profound idea but to me it just sounds like an attempt to make something uncomfortable, humans sometimes making wrong or bad decisions, more palatable by slipping a non-human into the decision point. It is the trolley problem where it is pulling a switch vs pushing a person on the tracks to stop the trolley. He is just wanting to put a LLM/AI there to take the blame.

In the medical example, the reason why we would put a person in the loop is because we have decided the decisions need judgement that only experience can provide, or is complex enough that it is had to distill it to a deterministic system, or finally it is 'more moral' feeling.

But I am not certain the answer is to drop yet another non-deterministic system into the place of a human and say, 'this is better'. If something goes wrong, we can always go to the human, ask questions, ask what happened and then develop new standards that fix the problem. A LLM is by definition a black-box, non-deterministic system, which has all the same drawbacks, plus more. If the LLM starts making bad decisions, the only thing we can do is pull it and try training it another way and hope it doesn't do it again?

The problem addressed in his example is trivially solved. Yes a human cannot make fast decisions, endlessly for hours on end like an LLM but that is why people in the positions don't do that work for hours endlessly. There is a reason why air traffic controllers at busy airports typically only do traffic routing for a couple hours at a time at most.

I know it must be hard when you paycheck is signed for a company the treats everyone as cogs to be worn down and replaced with another. I live in Amazon's backyard, the friends I have who have worked for them will retort, 'Well, it looks good on the resume that I worked at Amazon when I go look for other tech work.' But their resume has them doing a singular task for years and doesn't really show a breadth of skill, they were a cog and unless I am hiring for a cog there will be more that look better.

Rights groups brand Home Office's AI age guesser for asylum-seekers as biased and inaccurate

Why would it be up to the letter writers to come up with the test? The burden isn't on the letter writers to fix the government/vendor's cock-up. What is the point of the system if it is equal or worse than a person doing the job in the first place?

FreeBSD 15.1 lands, but desktop dabblers still have to draw their own GUI

Good laptop support, power management specifically is one of the things that keeps me from really considering FreeBSD as my daily driver. I use it on my home lab server and would love to use it on my laptop. This initiative gives me hope! I don't really have any useful skill to assist other than to cheer!

Amazon owns up to using 2.5bn gallons of H₂O in its bit barns last year

For comparison, the city I work for produced about 1 billion gallon/yr (3.7m litres) of water for a small city in the US of 40,000 people. And we have a water hungry industry in town as well.

Nottingham Uni says student records raided after ShinyHunters claims cyberattack

I expect a comprehensive report on Computerphile about this...

Trump admin pays wind developers to quit, back fossil fuel projects

Ah yes. I love my tax money going to pay (twice, if there were incentives) for infrastructure that just does nothing. Shutting something like this down is usually more costly than actually completing it.

If there were environmental hazards, those companies did the minimum to secure them and they'll be a disaster in future. The economic cost to the surrounding communities that now have hazards in the water without the obvious turbine attached. Towns that had hundreds of workers suddenly gone over-night because the company just shutdown operations.

So instead of just letting the project complete, at cost that is certainly lower than what they paid them to quit and allow the community to get the benefits, just burn that money, baby. Government efficiency at its best...

Microsoft tackles quality control issues. Just kidding, it's encouraging experienced workers to leave

Re: Got to pay that 'leccy bill somehow

I think even that is underestimating the cost of an US Microsoft employee. I live in Microslop's backyard and 180+k is probably about right for an dev, but in the US, by the time you include all benefits you can almost double the cost to the employer that of the offered salary. So I would say it easily costs Microsoft 300-400k a year per developer. Unless that employee is a independent contractor then it is probably just that flat cost, but I don't know how all that works, I am just a civil servant in an IT position.

Anthropic tests how devs react to yanking Claude Code from Pro plan

I am definitely not a lawyer but how can they advertise to a big group of people with all the features of Pro (with Claude Code), randomly select a new customer to then retract Claude Code from the offering, call it the same product, and still make that person pay the same price... and not be bait-and-switch fraud?

OpenAI puts Stargate UK on ice, blames energy costs and red tape

All for the grift.

Why else would you announce a big project like this before doing the homework, if not for just pumping those shares?

I mean if I got paid money in a popularity contest about how cool and chad I was, I too would announce I was going to buy a coolest Ferrari and then say I was backing out because it //didn't have enough// horsepower.

They were never really considered it in the beginning. The grift is the point.

Having had some experience supporting economic development offices, in the US at least, 'normal' companies are usually pretty quiet about new huge capital investments like plants and factories until all but signed. Reports, forecasts, infrastructure planning, workforce analytics are all done as part of the planning process and well before any announcements. OpenAI is hoping the hype train still has enough steam to distract investors from how fanciful this all is.

Elon Musk wants to build 50 times more chips than the world currently produces, using 'new physics'

Guys. Guys.

I am so stoked for his perpetual machine to take us to the Moon and then Mars!

Win after win.

This guy is such a genius. New physics; what can he not do!?

NASA sets 'impossible' ground rules for relocation of 'flown space vehicle'

Why a draft RFP?

Isn't there already a term for their object, the RFI?

AI still doesn't work very well, businesses are faking it, and a reckoning is coming

But did we need to ruin the economy to check that water is wet?

RSS dulls the pain of the modern web

Re: Don't forget Vivaldi

It is just me being an overly lazy/entitled person that really wants to use Vivaldi, and have, but the broken system theme switching on *nix machines is a major bummer. Having to dig through the settings twice a day is just to much to ask! *faints*

I am one of the weirdo that goes stark white (even my code editors) during the day and switch to dark at night. I need that switching.

But Floorp does a pretty good job of striking that middle ground of workspaces (which I love), supports light/dark modes, and has the benefit of being something besides Blink.

SpaceX wants to fill Earth orbit with a million datacenter satellites

How about we don't do what is essentially patent squatting on what is a scarce international resource for what is on-the-face plainly fantastical.

How do they plan to radiate kW of heat by millions of satellites without blocking out the sky by thermal radiators? Heat regulation is a complex and highly technical problem that engineers have worked on since the beginning of the space age. And the US is going to give them permission based upon a shrug?

This is as fanciful as his terraforming Mars, which is also silly due to physics. I would like to be proved wrong, but I feel pretty safe in that assertion. I don't expect the laws of physics to change anytime soon. The reason why underwater data centers are far more attractive is the exact opposite of what space can offer. Just got to grift that money from the US government while equally daft people are making political decisions.

DWP finds Copilot saves civil servants a whopping 19 minutes a day

It's... Fine

We use CoPilot for our government work, and it is good at search all our stuff for particular mentions of a topic or information. And, like the article says it is good for creating a first draft of an email.

I work in geospatial information systems, and it cannot really help with that, and even at time when I have thrown something at it to see what it says, it gives the veneer of knowing the answer. It will always suggest the lowest hanging fruit of how to solve a problem, and sometime outright incorrect suggestions. It is like interviewing a potential hire who is trying for a position they are a little out of their depth, they say a lot of the words (which LLMs literally do), but a good grasp of the subject is not there.

Generative AI, LLMs, etc., they all seem fine, but don't feel like a game changer. I want more domain specific ML models, that is where I think the usefulness is.

Capgemini to sell the biz that works for US government amid criticism of ICE contract

Suing the police in the US is a fool's errand. They have some many ways to escape responsibility given to them by the highest court that they are borderline untouchable. Yes, there is the mouth-service that they can be held accountable and at times things work out but the reality is far bleaker. Additionally, at first, you could sue the Nazi government, but they slowly consolidated power over 5 years through systematic means of purges and loyalty that they became incapable of being sued.

Additionally, ICE is enforcing laws that are not considered criminal in nature but are torts. In my opinion, there is no reason to have people being shot, citizen held for months sometimes deported for essentially trespass.

Latest Vivaldi release surfs a wave of anti-AI sentiment

I guess I don't quite grasp the interest in an AI-enabled web browser. Web browsing with a web browser seems like a very human centric activity. Not sure why I would want AI to browser for me?

I guess it could potentially get information to me faster in more concise manners. But I guess I am the weirdo that enjoys researching and actually knowing something.

Cursor used agents to write a browser, proving AI can write shoddy code at scale

+1 Gorman for evoking James Randi in a discussion about claims of AI efficiency without any introspection.

Palantir CEO claims AI will mean western economies won't need immigration

I could have never realized that AI singularity referred to tech and tech-adjacent companies in a circle jerk of money.

Microsoft teases targeted Copilot removal for admins

Which is my concern. For a gov tenant, and our org settintgs (I don't manage Microsoft applications, thank god), just opening office.com loads directly into Copilot. Does this count towards Copilot metric?

They have crammed Copilot into every corner that it is unavoidable.

IBM touts progress on tech stack for AI-enabled airline with no passengers or alcohol

Re: What?

I have lived vicariously through my wife's experience as a HR manager and HR is the worse place to try to make AI work. Sure, a lot of the work is boilerplate benefits and perhaps payroll, if setup that way. But a lot of HR is also handling personnel issues and making one-off calls about human-life-related issues, which automatically make it not easily categorized or resolved.

I feel the bean counters should be the first to go. Math only works one way and there are rules and laws about how to do everything. Then maybe get rid of bankers next and we'll stop having stupid decisions ruining everyone's lives and AI bubbles. (Joking, kind-of...)

Anthropic reduces model misbehavior by endorsing cheating

Still perverse incentives..

Isn't this just moving the goal post and shifting the incentives around?

This feels like treating symptoms instead of causes.

They expect the model to produce something and fit an known set of outcomes. So the models lie, fake data, hallucinate to please the developer and users.

So instead of tricking AI models into 'correct' behaviour like teenage children, why don't they encourage models to just say they don't know. Say it is indeterminate, or it doesn't know how? That would go way further in encouraging my trust in AI models if they would do that. I would start to trust the scope of AI's knowledge base and where I might have to do real research. Instead it behaves like one of those employee that thinks they are more competent then they are and will lie to cover it. They just insert chaos into everyone's day. I respect and trust people more who say they don't know.

But perverse incentives prevail. AI vendors would have to admit their model doesn't have the entirety of human-knowledge and that doesn't sound as good in marketing and ethics doesn't make VC dollars flow. It seems like business majors all nap during ethics classes.

Labor organizers accuse Rockstar Games of 'ruthless act of union busting' after layoffs

Re: Mistake

That is great idealism if the pipeline isn't from university straight into being a contract for a tech company, especially for young workers.

Contracts are complicated legal instruments that should have legal purview from all sides. A college graduate getting their first job isn't going to have knowledge to navigate contract work. They are looking for a job, and companies are saying come here and sign our click-wrap contract to start work! Yes, that may seem daft to older workers, but you are taking your existing knowledge for granted when pitted against someone who basically just left the metaphorical nest; the parents probably don't understand contracts as most people don't touch contract work in any way. Companies will exploit young workers for their ignorance. Young contract workers can't ask questions to things they don't know. Unions can help mitigate that predation on naive workers.

AI blew open software security, now OpenAI wants to fix it with an agent called Aardvark

For Free?

This kind of feels like a racket.

A gangster turning his goons loose on a city, but puts his hand on your shoulder and tells you he can protect your from the rampant crime if you pay him some money and let his accountant take a look at your books.

Texas senators cry foul over Smithsonian's pricey Space Shuttle shuffle

Re: Assume the risk, maybe?

They'll tell everyone it was the 'radical' Left that purposely sabotaged on its way south.

'First rule in government spending; why build one, when you can have two at twice the price?'

I am going to bet on the Smithsonian having a more realistic price for getting the shuttle to Texas in good condition.

Whoever these incompetent boobs hire to move the thing will almost certainly not do it for that price and will be jank.

Vibe coding platform Replit's latest update is infuriating customers with surprise cost overruns

That is what I am thinking.

Is it worth to comb over working code to fix presumably performance issues, and do those issues actually matter?

Unless you blindly trust the suggestions is it worth double checking the AI's work? To make such judgement calls you presumably have an understanding of the code base and would you want AI to suddenly refactor huge amounts of your codebase without your understanding; would you know what source file or module/assembly to start looking for a bug? This tool sounds like it just makes you worse at your job, less skill and less knowledge.

Browser wars are back, predicts Palo Alto, thanks to AI

“You literally will come to a point where companies will say: ‘You cannot use a consumer version of this product’.”

Were not most enterprises already saying this, or at least in the middle of draft new policies that say this?

I work in government and well prior to AI being available widely to the public we were already in the process of writing policy and figuring out how to ban access to consumer AI in our spaces. It is not perfect but there is policy for at least getting slapped around if you introduce personal AI into our networks.

Codeberg beset by AI bots that now bypass Anubis tarpit

I think the thumbs down is for the latent whataboutism in the comment.

All the adults in the room understand the implication of comment and it could be applied to all kind of events and actors. It just so happens that the Ukrainian/Russian war is a current world topic, not a war that was started 20 years ago.

Ironically, whataboutism was loved by the USSR when dealing with more western criticism. Although the points may not be wrong, it didn't typically address the underlying issue of the actual topic.

AI don't know: Enterprises slow to pick up on Copilot+ PCs

From the perspective of risk management and policy, Recall is a nightmare. There is a non-nefarious reason that it is good for things to be lost to time. If everything is recorded in some manner then when subpoenas start coming in... guess what is a record?

Regardless of if it never leaves your computer a lawyer is going to want that. Then you have massive amount of data, are you going to have someone thumb through all the records to determine what is relevant to the subpoena? It probably would be AI yet again doing that work, but are you willing to sign your name to procurement of records and say everything is there when you couldn't reasonably search all that data? What if IP gets written into the record but AI didn't classify it and you were too snowed under to catch it? What if your line manager is a racist/misogynist/etc. and liked to dictate his thoughts in chats or email and your company is now in a labor suit? What if an employee looked at their home cameras through a web portal on their work computer? Now you have recordings of the inside of their private space.

In short, the legal trouble of this is so near infinite, that once you put safeguards in place it would be nigh useless and not worth the effort over normal record management. I wouldn't want to be in charge of that #&$!* mess!

Orbital datacenters subject to launch stress, nasty space weather, and expensive house calls

Re: cooling using the emptiness of space

This is the same lack of basic understanding of the physical characteristics of space that Musk displays with his 'genius' brain when it comes to terraforming Mars. Mars doesn't have a significant atmosphere not only because it is not very massive but also because it doesn't have much of magnetosphere to protect it from the solar winds. The atmosphere it did have was blown away and then there is the problem of radiation.

I guess like all rich assholes, he has no grasp on scarcity. He plans for humanity to continually dump resources into generating an atmosphere, like dumping fuel from a jerrycan into your rusted out fuel tank so you can drive a janky truck a little bit further. But he probably believes he could restart the core of Mars or something absurd like that.

Microsoft says regulations and environmental issues are cramping its Euro expansion

I worked in the construction industry when I started out my adult life and then in permitting on the municipal side for a number of years.

Developers think everything is doom and gloom and everyone is against them. But guess what, if it still makes them money they'll bitch for a bit, but they'll still do it. They'll complain they'll need to spend an extra few hundred thousand on initial costs but then make million in profit every year. Never underestimate the self-flagellation and persecution complex of land developers.

'Cyber security' behind decision to end defense satellite sharing of hurricane data

Re: SimEarth

My family lives a few hundred miles away and there has been a large amount of rain in the Midwest this year and they have received a lot of flood warnings. The frequent warnings are probably not helping.

Additionally, I don't think people realize how fast water can rise from flash flooding. If a storm system dumps most of the water over a ridge it may go into another watershed and a river may never really rise, but if dumps most of the water in your watershed it can rise feet in the time it takes you to: notice the rise, find anyone who might be in danger, and finally move outside of the area affected. It also doesn't take that much water in the road to increase the buoyance of a vehicle, reducing its adherence to the road surface and be swept off the road into the river because the broad side of your vehicle (or just the tires) is/are a big 'sail' to the water. I grew up in rural farmland with low water crossings, as a general rule if you couldn't see the road surface and it is visible flowing you probably shouldn't ford a river. There may debris, scouring, road collapse hidden by the murkiness that is just as dangerous as the water itself.

Anyhow, the Trump administration's cutbacks to the NWS' budget are not helping. Don't mistake tweets as primary source for people's weather information. Where I grew up still doesn't have good broadband internet, mobile signals can be spotty, and radio stations are easily lost in the hills. One reliable source of weather information that could almost always depend on was the NWS weather radio network but that is no more. I recently turned on my weather radio to check it before camping and only one of the three were actually broadcasting now; not helpful if you are in an area with partial signal and that one station is the one not working.

Also, because of my job I am in regular contact with my local NWS office and they have warned because of personnel and budget cuts that warnings may be delayed and less accurate. While no person can point to a particular instance and say this is why so many people were injured or killed. Minor causes can have compounding effects that will lead to visible outcomes. I would expect a climb in property damage and casualties to continue.

To me, it is a combination of human behavior, ignorance, and a loss of public security good that we have had for nearly 60 years.

UK charity bank CAF branded a 'disaster' after platform migration goes wrong

Re: Test, test and test again

1,000 meter view but sounds like there were problems at project scoping. They didn't invite or invite a useful number of users to be stakeholders, or at least survey users to make sure they got the requirements near what their clients wanted. Not sure about the UK, but in the US Quickbooks powers a good majority of non-profit book keeping.

Microsoft Copilot joins ChatGPT at the feet of the mighty Atari 2600 Video Chess

I guess I am confused as to why anyone would think LLMs would be good at chess? They are mediocre at most things, at least competent at writing (which is arguably the thing they are designed to do), and terrible at anything technical (at least my field). I feel LLMs are jack of all trades but master of none.

Not insulting the researcher, at all. Reasearch like this is useful. But anyone's surprise to this would be like them being surprised a riding lawn mower is bad at haircuts. It could theoretically do it but isn't really designed to do so.

A ML model that is designed to compete in chess or work in my field would be / is far more interesting that any LLM. I guess that is why I don't understand all the effort to make LLMs, they are kind of shite at everything when models for specific applications would actually be more interesting and useful.

Apple goes glass whole as it pours new UI everywhere

I am not a fan of the new look but god am I tired of 'flat' UIs and this finally might be the thing that pushes away from this design decision.

They are low contrast, sometime ambiguous buttoned nightmares.

If it catches on, there might be some good skeuomorphic themes for XFCE/MATE beyond a half-dozen unmaintained projects.

Microsoft gets twitchy over talk of Europe's tech independence

Re: vowing to fight the US government in court to protect Euro customers' data if needed

Exactly. This is a red-herring. The misidentification of the causal problem is propaganda used against the American public by companies and especially the federal government. Republicans are quite adept at using this but this administration in particular loves to use it.

The EU already has laws that would de-facto require MS to fight the US for said protections to even to operate in the area. Privacy laws isn't what Europe is anxious about, it is availability and over reliance on technologies housed by a potentially, hostile country. If a broader war erupts across Europe with Moscow-aligned nations and Trump continues to allow Putin to puppet his naive ass, pressuring him into placing sanctions on the European block, then it is no longer about privacy. European lives lost, the economy, and war effort could screech to a halt and the Russian bear could end up demarcating Europe again.

If they are really serious about this, they would consider breaking up their own company into create a European company.

This American is rooting for a European company to fill in this gap and would like to buy server space in the future. Better privacy laws.

And maybe a job.

Hydrotreated vegetable oil is not an emission-free swap for diesel in datacenters

I am in favor of using renewables but the silver lining (ignoring particulates, smog, etc.) is that biofuel's carbon that is already in the carbon cycle. The problem is carbon that has been sequestered underground for millions of years suddenly reintroduced to the carbon cycle in large quantities. Biofuel is almost certainly better than any fossil fuel. An imperfect solution is still better the status quo.

Microsoft total recalls Recall totally to Copilot+ PCs

When it comes to Enterprise...

...And so continues the slow creep of where personnel management becomes less about managing people and projects, and more about looking at KPIs, and surveillance tech. Filling out paperwork to be put in the machine, now with AI(TM)!, where it churns out hiring quota and redundancies to protect shareholder's value.

This is the enshittifcation of management skills.

DOGE dilettantes 'didn't test' Social Security fraud detection tool at appropriate scale

Re: Surprised - If they had any self-awareness, they would have come to the same conclusion....

If they had any self-awareness, they would have come to the same conclusion....

You work at SpaceX as a software engineer, analyst, etc. and Elon pulls you off real work to go tilt at windmills in a sector you almost certainly know nothing about?

This person was the least productive or successful member of Elon's 'day-job' and they are being put there as a patsy for when this whole thing goes sideways. They are the political cover for when shit just doesn't work and congressional inquest is created to find out who should be blamed.

Sorry to Big Balls and his fellow DOGE-ettes but take some advice from this seasoned, career, US civil servant: start looking for another job now and don't claim this on your resume/CV. You walked into a political land-war as cannon-fodder and immediately stepped on a landmine.

Procter & Gamble study finds AI could help make Pringles tastier, spice up Old Spice, sharpen Gillette

I have seen an ad for a US bank, I think CaptialOne where the ad is just telling the viewer that they are using AI for business processes.

...Great. I guess. What is the corporate policy about binder clips or paper clips?

The only thing I am hearing is: "We are using AI to squeeze you and hard as we can and do questionable things with your money that isn't regulated and won't be because of US Congress and this administration."

Can't wait for the feature where CaptialOne AI is telling me I am not using my money as efficiently as possible and then show me ads where I can spend it.

HP Inc settles printer toner lockout lawsuit with a promise to make firmware updates optional

<< The company therefore works "to reduce unprofitable customers because every time a customer buys a printer it's an investment for us. We're investing in that customer." >>

That is not how investment works...

That is more akin to how pyramid schemes work.

Trump can't quickly or easily kill the CHIPS Act, but he can fire the workers funded by it

Re: Bizarre Funding

It was probably because your position provided IT support to research projects related to the DoE. If your hospital did research with nuclear material or a synchrotron, and if you did basic IT help with the computers involved, your wages could be partially or totally funded through research grants.

I have never handled a DoE grant; I have managed federal grants before and they typically allow for wages of support staff to be part of the expenditures. A grant that doesn't really cover the tertiary costs of the intended goal is really just wasting money. You would end up wasting $100k+ in idle workers, idle machines, and idle materials because your IT support is stretched thin, when you could have just spent 30k on a part-time or partially funded IT member who was hired to support the project.

Re: What?

My thought too. Why all the 'beautiful' tariffs but then axe the program that was supposed to encourage the semiconductor industry to build in the US?

Though the obvious answer is that he doesn't understand tariffs or is purposely not understanding/misrepresenting them. But I wouldn't expect an heir to billions made in the safest industry for investments, real estate, to understand anything about manufacturing. Especially when the person seems as intellectually incurious as Trump.

Google confirms Gulf of Mexico renamed to appease Trump – but only in the US

I live in the far corner from the Gulf in the US, so I would have almost no reason to put on a map publication. But this American GIS professional will never refer to it as the Gulf of America.

Tangentially, Denali is way more interesting word that an old, white guy's name. Noah Webster wanted to Americanize English so bad that he can almost be single-handedly attributed to drops of u's and s's in American vocabulary, but we collectively ignored native words which have made it distinctively American in an interesting way.

WINE 10 is still not an emulator, but Windows apps won't know the difference

I would love if ESRI ArcGIS Pro could be run in WINE. Everytime I have tried it seems to actively check for emulation and won't run.

But ESRI writes some of the most brittle software with the worst logging.

They probably don't want you using WINE because nothing will work because of bad programming practices!

Bosses face losing 'key' workers after forcing a return to office

Re: I went into the office for 40 years

I huffed car fumes and other people's body odour for 40 years!

Kids these days!