I really liked Primal, but part of that was the fact that they had the balls to kill the main character. I'm assuming season 3 is going to start with Zombie Spear jumping over a megalodon.
- Posts
- 5
- Comments
- 683
- Joined
- 2 yr. ago
- Posts
- 5
- Comments
- 683
- Joined
- 2 yr. ago
Short answer, no.
Long answer: We are a long way off from having anything close to the movie villain level of AI. Maybe we're getting close to the paperclip manufacturing AI problem, but I'd argue that even that is often way overblown. The reason I say this is that such arguments are quite hand-wavy about leaps in capability which would be required for those things to become a problem. The most obvious of which is making the leap from controlling the devices an AI is intentionally hooked up to, to devices it's not. And it also needs to make that jump without anyone noticing and asking, "hey, what's all this then?" As someone who works in cybersecurity for a company which does physical manufacturing, I can see how it would get missed for a while (companies love to under-spend on cybersecurity). But eventually enough odd behavior gets picked up. And the routers and firewalls between manufacturing and anything else do tend to be the one place companies actually spend on cybersecurity. When your manufacturing downtime losses are measured in millions per hour, getting a few million a year for NDR tends to go over much better. And no, I don't expect the AI to hack the cybersecurity, it first needs to develop that capability. AI training processes require a lot of time failing at doing something, that training is going to get noticed. AI isn't magically good at anything, and while the learning process can be much faster, that speed is going to lead to a shit-ton of noise on the network. And guess what, we have AI and automation running on our behalf as well. And those are trained to shutdown rogue devices attacking the cybersecurity infrastructure.
"Oh wait, but the AI would be sneaky, slow and stealty!" Why would it? What would it have in it's currently existing model which would say "be slow and sneaky"? It wouldn't, you don't train AI models to do things which you don't need them to do. A paperclip optimizing AI wouldn't be trained on using network penetration tools. That's so far outside the need of the model that the only thing it could introduce is more hallucinations and problems. And given all the Frankenstein's Monster stories we have built and are going to build around AI, as soon as we see anything resembling an AI reaching out for abilities we consider dangerous, it's going to get turned off. And that will happen long before it has a chance to learn about alternative power sources. It's much like zombie outbreaks in movies, for them to move much beyond patient zero requires either something really, really special about the "disease" or comically bad management of the outbreak. Sure, we're going to have problems as we learn what guardrails to put around AI, but the doom and gloom version of only needing one mistake is way overblown. There are so many stopping points along the way from single function AI to world dominating AI that it's kinda funny. And many of those stopping points are the same, "the attacker (humans) only need to get lucky once" situation. So no, I don't believe that the paperclip optimizer AI problem is all that real.
That does take us to the question of a real general purpose AI being let loose on the internet to consume all human knowledge and become good at everything, which then decides to control everything. And maybe this might be a problem, if we ever get there. Right now, that sort of thing is so firmly in the realm of sci-fi that I don't think we can meaningfully analyze it. What we have today, fancy neural networks, LLMs and classifiers, puts us in the same ballpark as Jules Verne writing about space travel. Sure, he might have nailed one or two of the details; but, the whole this was so much more fantastically complex and difficult than he had any ability to conceive. Once we are closer to it, I expect we're going to see that it's not anything like we currently expect it to be. The computing power requirements may also limit it's early deployment to only large universities and government projects, keeping it's processing power well centralized. General purpose AI may well have the same decapitation problems humans do. They can have fantastical abilities, but they need really powerful data centers to run it. And those bring all the power, cooling and not getting blown the fuck up with a JDAM problems of current AI data centers. Again, we could go back and forth making up ways for AI to techno-magic it's way around those problems, but it's all just baseless speculation at this point. And that speculation will also inform the guardrails we build in at the time. It would boil down to the same game children play where they shoot each other with imaginary guns, and have imaginary shields. And they each keep re-imagining their guns and shields to defeat the other's. So ya, it might be fun for a while, but it's ultimately pointless.
For someone who spends a lot of time alone and on a computer this will seem anathema, but go find some sort of physical activity (sport) and start engaging in it a few times a week. Not only does this get you out of the house, it creates opportunities to engage with people socially and it is good for your health.
I am very much a stay at home, be in front of my computer type hermit. I was this way most of my life and even being married didn't help much as my wife is the same. A good Friday night for us currently involves playing Baldur's Gate 3 until much too late. We have a very small circle of friends and don't get out much at all. However, now in my late 40's I am having some health issues and that finally gave me the push to get out of my gaming chair and get my body moving. I took up climbing at an indoor rock climbing gym and I really enjoy it. The regularly changing routes on the walls mean that I get to engage the puzzle solving part of my brain, and I am pushed physically as I try to get better. In between climbs I'm near other people with an obvious shared interest and can practice talking to other people by discussing the routes (social skills are like all skills, they take practice). And the exercise has made my doctor visits a lot less "you're going to die horribly" and more "we've got things pretty well controlled". I also just feel better.
So ya, go out and find some sort of physical activity you enjoy. Don't be afraid to try new things, you'll suck at them but that's to be expected. The first step in being good at anything is sucking at it. Use that suckage to engage with other people and learn how to suck less. This will help you suck less at socializing. I won't say that any of this is easy, it's not. I know there is the hermit piece if me which always wants to fall back into just hiding out in my basement (literally, my office is in my basement). But, I've also made a habit of climbing 2-3 times a week and 3 years into doing that I am now looking forward to that time. I get excited when I walk into the gym and see one of the walls changed and now get to solve a new set of climbing routes. I still kinda suck, but not anywhere near as much as I did on my first day.
They actually did replace the CEO which seemed like the least they could do. Whether or not the culture of "eh, who needs all these bolts?" has changed is still an open question. I for one still feel a bit better when I see an Airbus or Embraer jet roll up to the gate I'm boarding at.
I'll admit, I hadn't tracked it that closely and didn't know the details. I was just being glib. Thanks for the info.
And nothing of value was lost. Sure EA has published a few gems in recent years, but as a developer it's all sports games and Battlefield. The talent isn't at EA, it's at the developers they have been supporting. If we're lucky, the leveraged buyout will result in anything good owned by EA being sold off for parts and the worthless husk of EA saddled with the debt and left to go bankrupt.
Who know, maybe the license to make Star Wars games will go somewhere that isn't dead set on fucking it up as hard as possible to meet the Christmas season deadline.
With intermittent errors like that, I'd take the following test plan:
- Check for disk errors - You already did this with the SMART tools.
- Check for memory errors - Boot a USB drive to memtest86 and test.
- Check for overheating issues - Thermal paste does wear out, check your logs for overheating warnings.
- Power issues - Is the system powered straight from the wall or a surge protector? While it's less of an issue these days, AC power coming from the wall should have a consistent sine wave. If that wave isn't consistent, it can cause a voltage ripple on the DC side of the power supply. This can lead to all kinds of weird fuckery. A good surge protector (or UPS) will usually filter out most of the AC inconsistencies.
- Power Supply - Similar to above, if the power supply is having a marginal failure it can cause issues. If you have a spare one, try swapping it out and seeing if the errors continue.
- Processor failure - If you have a space processor which will fit the motherboard, you could try swapping that and looking for errors to continue.
- Motherboard failure - Same type of thing. If you have a spare, swap and look for errors.
At this point, you'll have tested basically everything and likely found the error. For most errors like this, I've rarely seen it go past the first two tests (drive/RAM failure), with the third (heat) picking up the majority of the rest. Power issues I've only ever seen in old buildings with electrical systems which probably wouldn't pass an inspection. Though, bad power can cause other hardware failures. It's one reason to have a surge protector in line at all times anyway.
I started self hosting in the days well before containers (early 2000's). Having been though that hell, I'm very happy to have containers.I like to tinker with new things and with bare metal installs this has a way of adding cruft to servers and slowly causing the system to get into an unstable state. That's my own fault, but I'm a simple person who likes simple solutions. There are also the classic issues with dependency hell and just flat out incompatible software. While these issues have gotten much better over the years, isolating applications avoids this problem completely. It also makes OS and hardware upgrades less likely to break stuff.
These days, I run everything in containers. My wife and I play games like Valheim together and I have a Dockerfile template I use to build self-hosted serves in a container. The Dockerfile usually just requires a few tweaks for AppId, exposed ports and mount points for save data. That paired with a docker-compose.yaml (also built off a template) means I usually have a container up and running in fairly short order. The update process could probably be better, I currently just rebuild the image, but it gets the job done.
But, but, docker, kubernetes, hyper-scale convergence and other buzzwords from the 2010's! These fancy words can't just mean resource and namespace isolation!
In all seriousness, the isolation provided by containers is significant enough that administration of containers is different from running everything in the same OS. That's different in a good way though, I don't miss the bad old days of everything on a single server in the same space. Anyone else remember the joys of Windows Small Business Server? Let's run Active Directory, Exchange and MSSQL on the same box. No way that will lead to prob... oh shit, the RAM is on fire.
It's caused by the speaker not liking what another person is doing, but being unable to articulate a real argument against the behavior. So, they call it "childish" as a way to express that dislike in a socially acceptable fashion.
If you're relying on someone else's computer to keep your data safe, don't be surprised when they use that data as a hostage to demand more money. Sure, using other peoples' computers to host your infrastructure can make a lot of sense. Just be sure you have a backup plan for when they send Guido around to demand more money.
Harm was going to happen no matter what you do in the trolley problem. There is no situation where harm does not happen, but there is a situation where you directly are causing harm.
Yes, exactly. By taking no action some amount of harm occurs, had you taken action that harm would not have occurred but other harm would have. Ultimately, this is analyzing the extent to which a person is willing to allow harm via inaction versus cause harm through direct action.
Almost none of them actually having a real world application...
Like many thought experiments, the Trolley Problem is an artificial situation intended to isolate certain decision making points so that they can be analyzed. Yes, reality is messy and we often have more than two options. But having this sort of analysis ahead of time can make the real problems less complex to consider. It is also useful for looking at our philosophical frameworks and where they break down.
Personally, if I could go the rest of my life without hearing about the trolley problem that’d be great actually.
The Trolley Problem is a tool for examining our beliefs. Throwing it away because it is imperfect and uncomfortable only leads to a blindness of self.
One of the issues the Trolley Problem explores is people's differing willingness to allow harm versus cause it. And that can hold even when the level of harm caused by inaction is significantly higher than what is caused by taking action. E.g. If your personal philosophy dictates that killing someone is always wrong, does it hold if your inaction causes 5 deaths, 10, 50? What if we start tinkering with the people dying? Would you kill a 90 year old man to save a train full of children? The Trolley Problem is really just a starting point to examine that dichotomy between causing harm and allowing harm and just how permeable the line between them can be when you start changing the conditions. Attaching other moral choices to the problem is one way to use the problem to explore a set of beliefs.
Dead is dead, parts is parts, dead guys is parts.
Azure Linux, because why wouldn't you want Microsoft in your Linux? /s
Choosing not to act is still making a choice and may still result in a negative outcome. It's the classic trolley problem. While you may not cause harm through an active choice, your inaction can still lead directly to a negative outcome.
DC really could use more public places to piss.
Mexican, because life is just better with spices.American, given that American "culture" is really just cultural appropriation with added sugar and calories, I'll get to try something close to every other cuisine.
As a species, homo sapiens have managed to adapt to every environment on Earth. We are the first species to have any measure of control over the natural forces which have wiped out countless other species. Diseases which once ravaged our populations are now gone or minor inconveniences and we continue to find new ways to mitigate the worst effect of many diseases. Should a large asteroid be heading our way, we are the only species which may stand any chance of diverting it or mitigating the long term impacts when it does hit us. While it was certainly not a "choice", the evolution of higher cognition, problem solving and intra-species communications has put our species in a unique position of having a high degree of control over out fate. Sure, it has its downsides (we are the only species which might be able to end all life on Earth), but it's been a pretty amazing run for us. On the balance, I think we're in a much better position to keep going as a species than our ancestors or cousins (homo erectus, homo hablis, neanderthal, great apes, chimpanzees, etc).
So, was it a "mistake", I think the current state of evidence is against that. While it may result in a really shit deal for individuals of the species from time to time, as a species I think it would be silly to consider it a mistake.