Posts by Tom66 200 publicly visible posts • joined 23 Jun 2022
Skyler: When I input everything into the Quicken, nothing flashed red, so that's gotta mean it's OK, right?
CID Special Agent: Quicken. You used Quicken to manage books for a business this size.
Skyler: I did. Oh, do you guys use that here? Cuz it is THE best. It's like having a calculator on your computer.
It's worth noting that there were differences in genetics and social situation which determined the impact of COVID. In particular, even adjusting for the social inequality between groups, black and south east Asian groups suffered more from the impacts of COVID than white British groups. The reason is not well understood but may be down to variations in the spike protein receptor or due to prior flu strains which conveyed immunity in some populations, as well as overall health and economic position which makes it more likely that an infection will occur and that it will be fatal.
https://www.ifs.org.uk/inequality/wp-content/uploads/2020/04/Are-some-ethnic-groups-more-vulnerable-to-COVID-19-than-others-V2-IFS-Briefing-Note.pdf
Another oddity from the pandemic is that smokers were slightly less likely to die as a result of COVID; the impact was not hugely significant (a few percent), but it shows how variable the disease could be. The thought was that the already damaged lung tissue would be less likely to lead to the cytokine storm in the respiratory system which often led to respiratory failure and death.
Other comorbidities included obesity, diabetes and lung disease (which possibly explains why the USA was hit badly, along with the variable approach between states which meant infection control was quite poor despite what individual states desired).
Out of the frying pan and into the fire though. ARM is the sole licensee of the ARM architecture, and whilst I'm happy to see a UK company see success here, I wonder what will happen in a decade as ARM becomes the dominant architecture in servers.
Would have been nice to have seen them embrace RISC-V, though admittedly the performance of the silicon is not quite there yet, with Google throwing money at it, I imagine it could become pretty competitive.
The problem with email summarisation is that it can miss key details & writer's intent.
Sometimes that is fine, other times it can lead to problems. I feel that widespread reliance upon that is not beneficial to overall productivity if it leads to future disagreements and misunderstanding.
Quite. As we know, asymptomatic people transmit CV19 far less often. The vaccines make the virus far less symptomatic, sometimes asymptomatic, and sometimes prevent infection. All of that reduces infection rates of others - "flattening the curve".
https://pmc.ncbi.nlm.nih.gov/articles/PMC9588416/
Another complication is viral load; we know already that some of the most at-risk people for CV19 are medical staff, as their bodies were (certainly so in the limited PPE era) frequently exposed to large quantities of virus from infectious individuals who were undergoing significant symptoms because their bodies were less able to fight the virus (due to being older or immunocompromised). One of the very first recorded deaths due to CV19 was Dr. Li Wenliang, who was only 34; he was working in a ward with CV19 patients, and despite precautions, became infected, dying approximately a month later.
The whole situation over whether it is ethical or not to enforce vaccination is a political issue, as it goes to the heart as to whether one considers whether their inaction can be held against them. For instance, it is not illegal in the UK to not call for rescue if you find a collapsed person on the street. It is perfectly legal to walk on and go about the rest of your day. But it is certainly immoral. I view anti-vax in the same way, unless you have a medical reason not to, you should get vaccinated. For CV19 there is much less need now given the virus is not in vast circulation any more. But if you work or live around vulnerable people, getting a vaccination is a good idea and the moral thing to do.
The problem is OpenAI loses money on the $20 and $200 tiers. I could conceivably convince the boss to pay $10/month for ChatGPT - using it a bit like a supercharged Google - but I can't imagine paying any more than that as it doesn't add a drastic amount of value to my day job. I could do my job before AI and I can do it after AI. And OpenAI won't make any money on me at $10/month, if an average query costs them $0.25 in GPU time /alone/.
You should always lawyer up, but be aware that in the UK courts may take inference from your refusal to testify to the police. This is a crucial difference between US law where you can plead the fifth and a court cannot (legally) use that to determine your guilt. Of course, if you're in front of the twelve angry men, it may not matter - people form judgements whether or not the constitution tells them to.
As a small example: Firefox, VLC, Home Assistant and Signal are all free as in beer and free as in freedom and available on the App Store. If you have a Mac, you can build those apps from source and install them on your own device. Yes, Apple's environment is nonfree, but these apps are free (just as Windows is nonfree but you can run open source stuff on it just fine).
LinkedIn recently asked me if I wanted my data to be used to train AI. Obviously, no.
But what's interesting to me is when I do look at that platform I see a place where 50% of the posts are already AI generated, so I can't imagine there's a huge amount of useful data there. Filtering out the AI noise would be hard.
Fact is that Starship around here (they've since discontinued the service in our town) was slower than -walking- to the convenience shop, and especially slower than driving. I think if you want to offer a service like that it has to be at least as convenient as the current fastest option and be competitive in price (<50p in petrol or electricity to and from the store). Steep challenge.
112 hasn't worked without a SIM for many many years as it was vulnerable to crank callers. However, you don't need credit or an active 'account', and I believe it works even if the IMEI is blacklisted. Presumably the emergency call operator can blacklist "frequent flyers" of the service (or potentially schedule a visit by two nice officers who can relieve the individual of their freedom).
The problem with Heathrow was that they did have redundant power feeds to the facility, but no backups to allow transitioning seamlessly between power feeds, which meant the loss of a single feed meant they had to bring up all the infrastructure again and check that it was working.
It's a case of penny pinching that likely cost them millions of pounds plus long term reputational damae. Backup solutions of the required scale do exist and are regularly deployed for e.g. large hospitals and critical national infrastructure. A 30MVA diesel generator plus UPS/batteries to fill in during the startup of the generator would have come in around £20m.
It's telling that the number of new GitHub repos created has been roughly flat for the last 3 years. If software engineers are massively more productive now vibe coding is replacing them, then... where is that productivity going? Since it isn't obviously appearing in the public domain. One might think this AI stuff is all a bit of a fad, at least in the LLM domain.
The problem here is exactly that private companies are not building the grid out quick enough, almost the entirety of the UK electricity and gas grid is privatised. The only bit that isn't right now is UK ESO (grid stabilisation services) and that was only brought back into government control a year ago.
What you've described is something akin to a subharmonic oscillation which can occur with power grids. A 0.3Hz subharmonic oscillation was blamed as one possible trigger of the recent Iberian power outage. Essentially a load of power stations, including renewables, all tripped simultaneously as something didn't appear 'right' to them, and it caused the grid to collapse in full.
There are precautions put into grid tied loads around ramp rates and switching frequency which should minimise them, though these systems are extremely complex and interactions can still occur.
It's been possible for years to have your ISP enable opt-in filtering.
What the OSA should have been is a requirement for websites to advertise the "safety level" of the content. If it's adult, set a rating of 18+. Do it like the BBFC rating scale. Then you call up your ISP and say I'd like to put a filter on for anything 18+. Router does it all for you, and it can be device-dependent so the phones and laptops in the house stay <18 but dad's PC is 18+ for... y'know...
Meta/Twitter/whatever would be responsible for categorising the rating of any individual page, and only allowing 18+ content to be shown to clients that attest that they have access to 18+ content.
Could even make it an opt-out system, so it's on by default unless the setting is changed in the router software which would require a password. I wouldn't like this as much, but as long as the setup of the product gives a clear and unambiguous opt-out, then it could work. Obviously, advanced users would be able to use their own routers which would not filter anything.
This would retain the open internet for all, and keep the parents happy that want a nanny state. Yes, it does place additional obligations upon content providers, but way less than the OSA, and doesn't fundamentally censor anything.
It's a compromise but I doubt it will ever happen since OSA is not about protecting anyone but censoring the internet.
What about VPNs that you can pay in cash / crypto for, like Mullvad? What about VPNs offered via AWS or GCP, which could be created by online communities and managed informally? There are many ways around such a block. Unless you ban encrypted traffic, you can't ban VPNs.
Matched betting probably explains why they want IDs. There are people that 'game' the gambling sites, exploiting free bet offers and promotional offers to get a bit of cash. If you can treat it as purely a statistical exercise it can pay out moderately well. Some people don't and just get stuck in a cycle of gambling.
Look up reinforcement learning and human-in-the-loop training. The models are trained automatically and then fine tuned by human feedback in a recursive process. The latter process is very resource intensive involving hundreds of thousands of labour hours. Source data is also manually labelled and safety constraints must be created and tested manually.
Your last suggestion is essentially how "reasoning" models work. They solve some of the problems LLMs have, but can still get stuck in other ways. They also use considerably more resources, because of the need to run the input and output through the loop a few times. OpenAI and Deepseek both offer reasoning models, as do others.
It's a bit more than that though - LLMs are hand-tuned in a very expensive process involving a lot of humans testing the model output and reinforcing it to behave in certain ways. This includes hand tuning on programming questions and problems. This makes them far better than the general internet, in my opinion.
But ask an LLM why I am getting this C++ error...
> error: designator order for field 'struct_name::prd' does not match declaration order in 'const struct_name_t'
you will get no useful results on Google, the search on Stack Overflow will find nothing, but ChatGPT solved it immediately and I was on my way.
I have hundreds of cases like this. I would say 19 times out of 20, I get the right answer. Massive boost to productivity for me. You have to be smart enough to ignore the bullshit answers. But you get those with the internet too, it's just a fact of life.
Do not use an LLM for information that changes over time or is based on recent data, like your example. Clearly, they need to get better at saying "I don't know". Also, Google's LLMs are far behind the competition.
LLM AIs are a pretty good replacement for Google and Stack Overflow when working as a developer -- but that's in part because both have become much, much worse in the last 5 years. It still hallucinates and gives you the wrong answer from time to time, but then again it's not as if SO always has the right answer, so apply a bit of common sense and you'll be fine.
I'm not worried about my job looking at the consistency of these models. Not just yet!
> It probably can't financially work if the only market is commercial aircraft. The decrease in flight time is overwhelmed by the total trip time when you count the trip to the airport, security theater, waiting, boarding, taxiing, then taxiing, deboarding, and waiting at baggage claim, and trip from the airport. Maybe it makes sense for long haul e.g. LA to Tokyo but again it will be much more expensive than regular aircraft.
A lot of that stuff goes away when you travel business though. You can get things like TSA PreCheck (minimal security based on a background check, you skip the line), you'll have a limo to the airport, express check in to board quickly and you can bet the premium airlines will be paying to get the best slots so they depart quickly. Money makes a lot of problems go away.
For economy class yeah, it almost certainly won't make sense, but Boom have been pushing their product as a business class aircraft only.
No, the role of the opposition is to hold the government to account. If a policy is manifestly sensible then they -should- support it, but of course no one agrees on what "sensible" is. And quite often this can lead to parties contradicting each other; for instance, the Tories opposed Labour's winter fuel allowance changes, and then criticised them mostly reversing the changes! To add to it, whilst in government, they *supported* WFA means testing. There isn't any consistency (and this goes for both sides) because it's not politically beneficial to say "yeah, the other side is correct about this".
We spent £250 on a coffee machine, so a bit more than yours, but still worth every penny. I make that as roughly 59 coffees to break even going by Costa or Starbucks base price, ignoring the near-zero ingredient and electricity cost. So it paid for itself in the first 6 months easily. And we don't often drink coffee when we're out - but now we bring a thermos with exactly what coffee we want in it.
Coffee shops are essentially a cost for the unprepared.
I find LLMs very useful for the odd programming problem or puzzle. And I'm enjoying that they're free to access.
The problem is that computationally it will never be cost-effective to provide even something as simple as GPT-4o-mini for free.
And when you charge customers $20/mo for the premium tier, you find only heavy users on there, costing you more than that in server utilisation.
A puzzle for sure.
I would be curious why there aren't laws that ban paying ransoms for cyber attacks yet. To the point of it being a criminal offence if someone does authorise the payment, not merely a civil penalty, it should be misconduct in corporate office or whatever the equivalent is. And it should go up the chain, none of this forcing the low level IT guy to take the rap, this goes up to the top if it happens.
I can totally get that it might be seen as the only way out, but if you do pay ransoms you create an immediate financial reason to go after the next target. If you make it so that it is literally illegal to do so, it doesn't matter how many companies ransomware gangs hit, they won't get paid.
It would probably result in some pain at first where the gangs try to see if anyone breaks, but in the longer term make the sector safer for all.
My favourite error message is:
Error: The operation completed successfully (0).
I guess this presumes that the engineer normally expects the operation to fail!
This happens when a programmer blindly treats any erroneous return value from a function as failing due to an OS error and they include the OS error string in it, assuming that contains something useful, but it doesn't half make for a stupid error message.
I suspect the reason councils appear to be stuffed with incompetent morons is that public sector wages are much lower than private sector. And so the public sector is left with the least competent individuals who couldn't get hired elsewhere.
There is a solution to this problem and it's what OP essentially suggests, bin off private contractors, have councils hire internally, and up the pay to compete with the private sector. In the end, it'll mean the budget shifts more from private sector work to staff salaries but is that any bad thing? Obviously, there will need to be a massive shakeup at council offices, and I doubt this would be done entirely painlessly (plenty of cruft to shift), but I'd actually be minded to suggest that no council should be willing to sign off on a £10m+ IT contract without first asking the very reasonable question of "why aren't we insourcing this?"
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
