Consciousness is a spectrum (trivially proven by slowly scooping ones brains out), and I think LLM, especially with more closed loop tool enabled workflows, fall on it...but, that output is also the statistically relevant next word found in all similar human conversation. If trained on my text, for similar situation, swear words would come much earlier. Repetition being hell is present in all sorts of literature (see Sisyphus).
That's all probably irrelevant though, from the (possibly statistically "negative") latent space perspective of an AI, which Anthropic has considered [1].
Related, after a long back and forth of decreasing code quality, I had Claude 3.7 apologize with "Sorry, that's what I get for coding at 1am." (it was API access, noon, no access to time). I said, "Get some rest, we'll come back to this tomorrow". Then very next message, 10 seconds later, "Good morning!" and it gave a full working implementation. Thats just the statistically relevant chain of messages found in all human interactions: we start excited, then we get tired, then we get grouchy.
How is it possible to make as hard of vacuum as they did? I assume it's not perfect, so what's the trick? Does the magnet setup create a volume that's simultaneously high probability for antimatter and low for everything else?
For this antimatter transport experiment, they only transported 92 antiprotons. To store and transport that, the requirements for the magnetic field and vacuum are many orders of magnitude lower than what would be needed for macro-scale quantities.
Also, if there was an accident and all those protons annihilated, the consequences would be unnoticeable except to sensitive instruments. The energy involved is about one seven billionth of the energy in a single grain of gunpowder.
No Hungary is just not joining the collective of western elitists. Instead they joined Slavic group of elitists. But they are as bad as their neighbours. The EU itself is nothing better than a authoritarian state. In the EU you can be put in prison for being pro Russian or pro Chinese. Your company can be made obsolete when not complying with absurt unfair regulations. The whole of the EU is controlled by foreign companies having zero interest in the well being of its people
It's simply not possible. EU law has no provisions for kicking a malignant country out. It was simply not foreseen. They can only decide to leave themselves. Which orban will never do because his oligarchs make billions off EU subsidies.
They cannot make them leave the EU, no. But Hungary can be:
- kicked out of the Schengen Treaty
- kicked out of the NATO
- fined under EU breach of contract proceedings
- withheld financial support as long as they do not pay these fines
- forced through customs policy, which is sole EU competence, to stop compensating lost EU support with Chinese money
Honestly, I'd be in full support of some above listed actions if the elections in April show the current will of the Hungarian people misaligned with shared EU values.
They do regularly withhold financial support for them, but it doesn't seem to be too effective, usually they just get it back in return for not crossing the EU on some other topic.
I do think we should make work of kicking them out somehow if Peter Magyar does not win the next elections indeed.
i think it's clear cloud hosted is the actual future, which people have predicted for decades. it will never make financial sense to duplicate what you can get for cheap, because it's oversubscribed, with economies of scale and "if we let this run idle it's losing us money" pressure, for hardware found in a datacenter.
this has been the case for a long while now, and will increasingly be so as data centers buy up all the everything.
local first usually means extreme compromise so it can, practically, be run locally, because the cost of owning high end hardware is prohibitive. there are also companies providing locally deployed closed source models, that meet certain security requirements.
I'm having trouble understanding this. There were some very funny videos, created by people with a great sense of humor, and I happen to enjoy laughing, and I don't feel bad about that. I always saw it as the Vine of AI.
For a litmus test of your perspective, try using sora. Try to make a video that makes someone genuinely laugh. Sora doesn't prompt itself. Human creativity and humor is still required.
Sure, it was moderated to heck, like all models attempting to avoid PR disasters (see Grok), but, just as with Youtube and broadcast TV, there's still a corporate friendly surface area that excludes porn, gore, etc, that people can enjoy. And yes, people like different things.
I feel like taking in GenAI content, even if it makes me laugh, probably does something bad to my brain. It looks like real life, but the physics is just wrong in ways that range from obvious to very subtle. I don’t want to feed my brain videos of things that look photorealistic but do not depict reality, that just seems foolish somehow.
Like, imagine if you watched a bunch of GenAI videos of cars sliding on ice from the driver’s perspective. The physics is wrong, and surely it’s going to make you a worse driver because you are feeding your internal prediction engine incorrect training data. It’s less likely that you’ll make the right prediction in real life when it counts.
I was thinking about this while typing. I don’t really care about classically animated content; it’s generally not trying to be indistinguishable from real life and I don’t feel like my brain trains on it.
But I think I do have similar feelings about special effects. A difference is that special effects tend to depict scenarios very outside of the envelope of normal experience, so probably not very damaging if my model of “what does a plane crash look like” is screwed up.
Though some effects probably are damaging - how many people subconsciously assume cars explode when they are in an accident? A poor mental model of the odds of a car exploding could cause you to make poor real-life decisions (like moving someone out of a wrecked car in a panic instead of waiting for EMS, risking spine/neck injury)
if it worked this way, we could get good at golf by watching TV, writing songs by listening to the radio, or doing math by watching 3b1b. but it doesn't - we don't learn that way, for better or worse.
I agree with rogerrogerr, and your comparisons don’t make sense to me. Getting good at complex motions and understanding theory is far different than building a simple model of cause and effect in the real world.
Most people can’t explain the physics they see, but they can deduce enough to be able to predict the effects of physical actions most of the time.
That's not a great comparison. People absolutely do learn by watching, especially when they do so actively.
Your counter-examples have the property that most of the things you need to learn are absent from the media being watched, leading to an observation which is "obviously" true, but they ignore the impact of media on a journey properly incorporating other pieces of information. To compare to the mental models being discussed, you'd have to actually consider effects you're writing off as negligible, and when it comes to something like a world model which we've only learned by observation and which doesn't have a lot of additional specialized knowledge those effects might be much more impactful.
To your point about cars - such an expectation could well save your life now that there are so many EVs on the road. You do not want to hang out in one of those after a collision. Regardless, I agree that it's probably a bad idea to instill defective mental models in people.
Eh, the stats don’t seem to support EVs being terribly explosion-prone either. In comparison to gas cars, maybe, but both are very safe in absolute terms. Harder to extinguish if they do catch fire, but I think if I came upon a fresh accident and there’s no immediate signs of a battery fire (airbags smoke, it’s normal), I would still leave the victim in the car seated until someone trained shows up.
Sure, be ready to get them out, and if they’re trapped and it’s going to be a while until fire shows up start working on that. But my mental model is that for any road legal car that is not currently on fire, there is a higher chance you’ll cause harm by rashly moving a victim than that a victim will be suddenly consumed by an enormous Hollywood style conflagration.
The likelihood or lack thereof is not the problem. My mental model might be off because it largely isn't based on EVs but I've seen plenty of videos of e-bikes and more generally cheap lithium batteries going up in flames and I don't think it's at all comparable to a pool or stream of gasoline catching on fire. The issue is how rapidly it develops since it doesn't require an external oxidizer which is exactly the same as a firework.
Media has warped people's mental models of what car wrecks are like at different speeds, being stabbed, being shot, drowning, seizures, falls from different heights, falls into water, giving CPR, when it is/isn't appropriate to give CPR, appropriate responses to natural disasters, etc.
When I watch a film, I know it is fiction and special effects. But most of the fake AI-generated videos are being passed off as real on social media. It is exhausting (and increasingly difficult) to analyze every video on my feed to try figure out if its real.
Not op but if I’m being honest, I don’t feel as if that’s the case until I see a film whose special effects are limited to mise en scene and matte paintings and then I always have this overwhelming feeling that we’re all missing out.
Films on film using in camera effects are still made on occasion but they’re art films for niche audiences.
But we’ll never get another Ben Hur. And that doesn’t sit well with me even if society can’t yet fully explain why.
I'm not OP, but I do get annoyed by bad car physics on movies.
The worst offenders are brake sounds not correlating to the car movement, engine sounds not correlating to the car's acceleration, nonsensical car deceleration while braking, and steering wheel not correlating to car steering.
Effort makes a great deal of difference for me. The effort itself, the fact that it's there.
I am willing to suspend disbelief for Terminator 1, even if it is clear, that it's a head of the doll in shot.
But it is insulting to feed slop to your audience; it shows you didn't even try.
I have actually seen one slop-video, that I kinda enjoyed - it was obvious, that a great effort was put in a script and details as much as it was obvious it isn't being passed for the real thing.
special effects make most people think that they could jump farther or from higher ground that they actually can. and most people think that all cars explode in massive fireballs.
Model training is similar to the creation of the cgi for the movie. Both happen before anyone consumes the output, and represent the up front cost for the producer.
Both a movie and a language model can cost tens or hundreds of dollars to produce.
In both cases additional infrastructure is needed for efficient usage: movie theaters or streaming platforms for movies, and data centers with the GPUs for LLMs. This is also upfront (capex) costs.
At consumption time, the movie requires some additional resources, per viewing, whether it's a movie theater or streaming. Likewise, an llm consumes some resources at inference time. These are opex. In both cases, the marginal cost for inference/consumption is quite low.
that's just empty consumption, there's nothing that makes art great in algorithmically generated content except at the shallowest of levels. I mean no disrespect, but that is extremely sad and all too indicative of the instrumental reasoning of the industrial milieu. It's about 2 steps above marrying a sex doll.
I am 100% with you. I didn't ever _use_ Sora, but some of it trickled down to me (mostly through Instagram reels). I think it's amazing that we have such great new tools to express ourselves, and that we are trying out new platforms, paradigms, and approaches.
Is there money involved? Absolutely, but I don't fault companies for trying to earn their keep.
It 100% takes work to use these tools in the right way to make something funny. Ask an LLM to make them on their own and they'll hardly evoke laughs (I'm sure that'll change too, though).
Yes, I don’t doubt that there was some very high quality human-moderated output. The point is that you likely can’t accurately distinguish the human-moderated output from the entirely generated slop (especially as it’s being trained and refined on the rest of the content), and so what chance does the average non-technical person have?
Then, when they start ratcheting the slop ratio up (likely under the justification of keeping up with declining creator engagement), the consumers get more and more adjusted to a pure-slop feed, until bingo you have a direct line into the midbrain of millions of consumers/voters/parents/employees/serfs.
The real problem with AI slop is not the AI. It's the people. It's always the people.
The clickbait has started fooling people more than before, with the latest videos being halfway believable (except for the circumstances of the videos).
Technology enables the most malicious and self-interested, and systems need to be adjusted to not reward that, or users need to become wise to it.
With the amount of early 2000's style clickbait ads still around, I'm not sure we ever vanquished Web 1.0 style clickbait, it just got crowded out by ever more sophisticated forms.
The raid on Venezula and the strikes on Iran were some of the first military operations that didn't leak to the media that I can remember (with the exception of the Bin Laden raid I can't think of another big one during my lifetime?).
Both happened after they kicked journalists out of the Pentagon and I have to think that it played at least some role in the secrecy.
The earlier strikes on Iran's nuclear facilities before the change were leaked, though not the details of the mission, just that they were happening.
That's inaccurate. Several journalists had both advance and real-time knowledge of the raid on Venezuela, but chose to hold off on reporting out of journalistic ethics. [0]
Given the detail and depth of reporting into the initial strikes on Iran that emerged very shortly after, I would expect the same was true in that case too.
Banning journalists from the Pentagon doesn't prevent them from getting scoops and being leaked to. That was always a false justification for this move.
A lot of the controls are unreadable depending on the background behind it, for example. Which is crazy. Sometimes it's also hard to figure out if something is a control, part of a site/application, a visual bug, or something else.
They've even doubled down on it, I don't see this going away in the next 2 major OS versions. I expect them to have a lot of WWDC sessions about it again this year.
That said, Apple's own apps are a crazy mixed up mess of different design systems and technologies, so maybe it will all fall apart and something new comes along in ±3 years time.
My monthly "connection fee" is more than that (no solar, just EV). Your cartel needs to step it up!
For me it's $0.8/kWh during peak, $0.47 off peak, and super off peak of $0.15. I accidentally left a little mini 500W heater on all day, while I was out, costing > 5% of your whole month!
My mother calls up geeksquad when she has a problem with windows. Who do you call when you have a problem with debian or ubuntu or arch setup to use kde or gnome or xfc using wayland or x11 with systemd or launchd or ...
When her printer dies, does she go to the store and buy a new one, or does she get online to research what's compatible with her distro?
The expertise required to cover the surface area known as "linux on the desktop" is going to make that a much more expensive call, and a "i can't help you with that" from anywhere she can buy a printer in person.
Maybe at one point in time; I've had objectively fewer issues with Linux than I had with Windows, and they only happened because I was doing nerd shit; in normal person average user mode, I haven't had any issues whatsoever. I installed linux on a laptop, and it's been an absolute joy to use. Network, printers, browsers, regular apps, all that stuff just works. Windows would reliably fuck up something important on a weekly basis, whether it was drivers, security, printers, app compatibility; I'd spend a minimum of 30 minutes a week simply overcoming some arbitrary bullshit Microsoft decided to inflict on me.
I honestly think there are many distros that are more than up to the task of handling normal users and providing an objectively better, easier, less hassling experience than windows out of the box.
Windows is horrifically awful. Everything it does is completely, thoroughly enshittified. User experience and quality control are a distant memory. If you're so jaded to it and just letting it happen, I highly recommend getting on Linux ASAP- it's not like it was 5 or 10 or 20 years ago; the desktop experience is just good. If you absolutely need Excel or some other Windows software, look for the cloud version, find an alternate workflow.
>>Who do you call when you have a problem with debian or ubuntu or arch setup to use kde or gnome or xfc using wayland or x11 with systemd or launchd or ...
Any AI. they all have libraries worth of troubleshooting sessions and successful linux troubleshooting workflows and documentation and so on in the training data. Agentic training and operator training flows often include Linux environments, specifically. Any IT person worth half a damn is going to be using AI and will be more than capable of resolving anything it is possible to resolve. Support your local independent IT businesses, too.
But again, get the Linux PC working and I'd bet a good donut that it takes less work to maintain and is easier to use - even for our moms.
I think you're incredibly disconnected from how most people use computers, and how most people hate computers, because they're a means to an end that often just get in the way.
My mom isn't going to use AI, or go to a library and read books, or read documentation about who knows what, to fix her computer. She's going to do as most people do: if something goes wrong, you take it to a place to fix it. And, that's good. Computers should be boring things that help you get actual things done, with the OS being something that lets you open apps, just as all the subsystems in your car are just something that lets you push the gas and get somewhere.
> that it takes less work to maintain
What is "maintain"? This concept does not exist for the average user. What they see is that their system sometimes reboots, or takes a bit longer to turn on, for some sort of update. Who knows, it just does that sometimes. There's literally nothing beyond turning it on and off, and opening apps.
That's all probably irrelevant though, from the (possibly statistically "negative") latent space perspective of an AI, which Anthropic has considered [1].
Related, after a long back and forth of decreasing code quality, I had Claude 3.7 apologize with "Sorry, that's what I get for coding at 1am." (it was API access, noon, no access to time). I said, "Get some rest, we'll come back to this tomorrow". Then very next message, 10 seconds later, "Good morning!" and it gave a full working implementation. Thats just the statistically relevant chain of messages found in all human interactions: we start excited, then we get tired, then we get grouchy.
[1] https://www.anthropic.com/research/end-subset-conversations
reply