Posts by Ken Hagan 8486 publicly visible posts • joined 14 Jun 2007
In evolutionary terms, the love is proportionate. A slab of air will absorb a fair amount of UV and the further you go into the IR the larger your optics need to be (and there are other absorption bands there too) so Nature's choice of waveband has a lot going for it. Also, the Sun is quite bright at visible wavelengths but not so much at shorter ones. Various non-human eyes can see bits of spectrum that we can't, but they don't differ much. Stepping outside of biology and into engineering doesn't help much either since any passive sensor is bound by the same constraints.
"There’s plenty of room to grow, because the IPv6 number space covers 340 undecillion addresses – 2128 compared to IPv4’s 232. "
Not nit-picking, since I'm not actually sure what the correct answer is, but the above quote is something of a mis-representation.
In IPv4, the addresses are so densely allocated that (ignoring the multicast and experimental blocks) we probably are using more or less every one.
In IPv6, each individual client of an ISP gets a /48 or a /56 prefix and most people probably have only one LAN hanging off that (so they're only using a /64, despite being given 250 or 65000 of them). It's great for routing, but we're never going to fill the space densely like we've done with IPv4. In addition, by no means all of the possible prefixes are assigned for unicast addressing.
I've no idea even how to estimate the number of devices we'd need to have to "pack out" the IPv6 address space, but I suspect its logarithm in base 2 is nearer 56 than 128.
When a chip is designed in the US but fabricated elsewhere, how hard is it to verify (by inspecting a random sample) that the samples match the design?
If this is done, and the malicious fabricator expects it to be done, how hard/costly is it to manufacture a small number of trojan horses and hide them in a large batch?
This approach obviously makes it less likely that a bad chip will be used in a sensitive system. To what extent does that make the technique effectively useless?
What sort of chip (CPU, GPU, other) is the best target for attack?
I'd say the second was more likely, but I'd echo the "please don't" comment above. It won't end well.
When it was added, it might have seemed like a clever addition (especially in the light of the Revolutionary War) but it is already true that a violdnt revolution can overthrow a tyrant and already true that people will find the weapons they if they need them. It didn't have to be cast in quasi-legal form.
Compared to other countries (that don't have a similar right) the US appears to be no more resistant to bad government and significantly less resistant to violent crime.
"Harris had already lost the vote because of the Biden administration's support for the Israeli attacks"
Possibly true, but breathtakingly depressing given The Orange One's known crawling subservience to Netanyahu.
I suppose it is like all those Green voters, foolishly thinking that a FPTP election still allows them to "vote with their hearts" without the consequence of letting in the guy who most stands for all that their hearts hate.
I think that's just a matter of cloning 5000 SD cards for your Raspberry Pis and then supergluing them in place. (Yes, glue it in. The pi, plus case and power supply, is well under 100 of your favoured unit of currency, so it is not the end of the world if you have to replace a few every now and then.) Your physical attack surface is then pretty minimal. It only has to connect to one particular service and run one particular RDP-like application so all 5000 can basically be confined to your internal network. Your software attack surface is then also pretty minimal.
In short, the risk of someone coming in via the thin client here is miniscule compared to the risk of someone phishing or otherwise forcing their way into the full-fat Windows desktop that it is talking to. Yes, that can be maintained with all your Active Directory tools, but it is an ongoing task and not without effort. I would be quite surprised if the Windows VMs didn't take up the vast majority of your time. (By the way, just how many organisations do you know with 5000 seats and no IT staff?)
Embed your message in a picture using some steganography. It doesn't need to be clever because you choose a picture that is likely to be reposted on social media without attribution and by a foreigner. Wait for that to happen. Then "innocently" tell your intended correspondent about the amusing post you saw yesterday. You can do that in cleartext. They can tell (from a particular phrase, or time of sending, or your choice of account, or ...) that it is worth pointing their tools at the picture.
They then have your message, which you did not send to them and which they did not receive from you.
What did you think all those cat videos were for?
Three: all internet commerce becomes unsafe. The politicians, realising this, introduce an exemption for commercial sites. Everyone starts using commercial sites to send their secret messages.the politicians amend the exemption to apply only to "major" commercial sites. All others are forced to use (say) Amazon Marketplace. Everyone now notices that, by EU law, most EU commerce is now paying a percentage to a US company.
I expect I cluld go on with this insanity but ... the only winning move is not to play.
In fairness to the AIs, we actually use the same method to train our own offspring. We don't leave them to drown in a social media swamp. We send them to school and expose them to a controlled curriculum, under professional guidance, and hope that enough of the quality rubs off on them before the exam.
"better than average quality comes at a cost"
Whisper it, but it's almost as though traditional publishing houses performed a useful function. One that they have sometimes over-priced perhaps, but a useful function nevertheless.
Amusing also to realise that training an AI on humanity's normal output doesn't work. To have any hope of seeming truly intelligent, you have to train it on our "best behaviour"!
I doubt anyone knows, or expects to know. A neural net isn't an algorithm. Although the different stages may often be inspired by the kinds of categorisation that humans apply when doing the same task "by hand", there's no hard and fast rule requiring this. You could just start with a layer containing all the raw measurements, end with a layer delivering the time-series of the forecasts for different places, and stick a few layers of "mix it all up" in between and see how you get on.
If it works, it's useful. Eventually, one might study the best of the working models and see if the layers are doing something specific that no-one had thought to try before. It could give you clues that eventually lead to genuine scientific progress. In the meantime, not needing a super-computer to create a forecast might even save you money. (Perhaps not, mind, since you probably do still need something quite chunky for the training.)
If these are old PST files with historical mail, the thing to do is set up an IMAP server (Dovecot on an RPi will do, visible purely over the LAN) and run that old Outlook one last time to copy the entire archive over to that server. It's then accessible to any email client on any OS.
As others have pointed out, the key point is whether or not you were foolish enough to be related to anyone with said 'idiot gene'. Given the number of relatives that the average person has, those aren't good odds
DNA ought to covered by rules on personal info, but since this is a US company I'm not sure that would really help in this case. In short, I think the horse has already bolted for most people in Europe and North America.
If it makes you feel better, there are nothing like enough nukes to do the job. There are 8 billion of us and while there are concentrations in cities, almost every country has significant numbers in rural or small town settings. In addition, any sane targetting strategy would direct many warheads to military targets, not population centres, because you definitely don't want your enemy's military to survive after you just murdered their friends and families.
I think "the certificate" in this case is the same one that signs all other third party drivers and it hasn't actually been compromised.
The correct solution is to teach Windows to recognise this particular driver, which apparently MS have already done in the case of Win11. (Kinda odd that Win10 has missed out on that. Are they trying to blackmail us into upgradjng or something?)
It would be mildly impressive to see an AI that could even attempt answers to the average IQ test, let alone score 100. In my experience, these tests have all sorts of question types: verbal, numerical, and visual.
Can you scan images into ChatGPT and ask it to draw the next in the sequence?
I keep reading this claim, but the link takes me to a paywall so I can't see the evidence.
Meanwhile, the evidence that I can see suggests that ChatGPT and its ilk cease to be convincing if you actually bother to ask related questions or demand some sane levels of consistency from one question to the next.
Obviously there are some people who also struggle with this, but I prefer to summarise that by ssying that they fail the Turing Test. I'm not prepard to lower the bar just because a few of my fellow Homo sapiens are ... a bit thick.
It sounds like avoiding big companies is a good idea. Management there seems to be mostly by professional managers and that's always a risk. With SMEs, you're much more likely to have a boss who either understands your job or understands the /need/ for your job (and that's why they hired you).
"one wonders if the mangerial ranks have selection biases that attract sociopaths more than mere competence."
If I were an incompotent sociopath, I'd probably be reluctant to hire or promote anyone who might bring compotence or social graces to the management team.
So, yeah, possibly a selection bias there.
That, as I understand it, is a rewrite starting from a blank sheet of paper. Writing a kernel in Rust would also be OK and efforts are underway. What is problematical is adding Rust to an existing GPL project and then "taking away" the GPL at the end when none of the origonal code remains. The trouble is that all of the intermediate phases had your Rust code under a GPL licence.
I'm not enough of a lawyer to say how hard it would be to develop under two licences. I think VLC's x264 code does that, but it isn't FOSS.
Added: I am enough of a Reg reader to know that the kernel has not migrated to GPL3 because it would be Hard. That sounds to me like it might be a relevant observation.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
