Many replies are calling this out. Though it hurts to see one of them literally screen-capping and circling an em-dash and acting as if it were impossible for humans to produce this glyph in text on the Internet. It took me hardly any effort to find, for example, https://x.com/BernieSanders/status/1149022063925088256 from 2019 using an em-dash.
"In Pulse surveys and conversations with many of you, we’ve heard that too much work requires coordination across teams, debates take too long, and decision-making rights are unclear. I’m sure many of you have felt that you spend too much time “aligning” rather than building, shipping, or serving customers. To improve this, we have reduced roles primarily focused on coordination, and have clarified the remit of the coordination roles that remain. "
Well, in theory this sounds reasonable. But who knows what will be the real consequences.
People whose primary job is to "coordinate" often, either in a well-meaning attempt to be thorough or just to be relevant, try to pull as many different people into any decision-making as possible.
So instead of me just getting approval from my manager, I need to get approval/buy-in from departments A, B, and C.
In that case, firing the "coordinator" person increases velocity because parts of the business processes they prop disappear.
> In that case, firing the "coordinator" person increases velocity because parts of the business processes they prop disappear.
So you get rid of the coordinator. What then? Does this magically unblock people?
Let me frame this differently: if coordinating teams creates as much problems and impedance as you claim, how come Uber hired them to begin with? I mean, either Uber was shooting itself in the foot then, or it is now. Which one is it?
Yes? Removing and/or minimizing detractors/hangers-ons/randoms from a project is an important part of running a project. Especially in corporate scenarios where telling people to go pound sand is, unfortunately, not an option, more elaborate strategies must be employed.
> If coordinating teams creates as much problems and impedance as you claim, how come Uber hired them to begin with? I mean, either Uber was shooting itself in the foot then, or it is now. Which one is it?
Are you thinking that humans can predict behavior and evolution of social systems, and especially how they would behave after some change is forced upon them?
In my opinion, that cannot be further from the truth. Any sufficiently complex social system, which Uber with 30k+ employees definitely is, exists and evolves more or less independently from how any one individual part of this system might wish. Even if it's a CEO.
Uber leadership thought cross-team coordination would help them, so they hired a bunch of people to do it. That might have been a correct idea at a time, or at least on paper, it looked like a correct move. But in the end, it did not pan out. So they probably shot themselves in the foot back then. Did they shoot themselves in the foot now? Who knows? (Spoiler: no one) We'll see once enough time has passed.
They pinpointed the problem of over-coordination to the messenger (coordinator) instead of the teams demanding final approval over launches (coordination).
Won't merging different teams under a single owner only push the problem further down the org chart?
I've heard this at other companies and it's almost always an issue of upper management holding too much control over decision making, or poorly communicating actual business objects, but it's project managers and middle management that will bear the brunt of the layoffs.
I haven't used Grok (personal distaste). I tried Gemini in Antigrativy - it was quite bad. If I gave it precise instructions, it could execute them. But if I gave it a broader task, it would make logical mistakes and mess up.
You know, I have learnt that I have aphantasia several years ago and it opened my eyes to many things. I have thought that people say a lot of things metaphorically.
- Imagine a beach when meditating
- If you are nervous about giving a presentation, imaging that the audience is naked
- Count sheep in order to fall asleep fast
- Undressing someone with your eyes
- Visualising what is happening in the book you are reading.
I was completely sure that these were just metaphors. Imagine my shock when I realised that the majority of people can actually DO IT.
I thought "count sheep" just meant "please just go the f*k to sleep, kid". Literally counting sheep was too absurd. At the risk of sounding obvious, there are no sheep. No amount of trying to conjure pictures of sheep would work. I was very confused by this as a kid.
Later in life, the realization moment came when I was watching a TV show, a police procedural where the witnesses described the suspect. Police were able to use the descriptions to make an accurate sketch. "Yeah right, like anyone could recall those details". I was convinced it was a Hollywood trick.
My friends looked at me funny and proceeded to describe the faces, clothing, etc. of off-screen characters with stunning accuracy. That's when I learned other people can recall (or even conjure) visual images and read them like photographs.
Visualizing a sheep is not something I can even do once. Let alone repetitively until it gets boring. Being asked to visualize is quite frustrating and confusing - especially to a little kid not sure exactly what he's supposed to be "seeing". Hardly relaxing.
Same. This, and also the fact that it is also very highly associated with SDAM (Severely Deficient Autobiographical Memory) -- which I also have.
And also, reasonably associated with ADHD -- which I also have, but only recently realised and assessed for at the tender age of 46.
The SDAM is particularly useful to be aware of, otherwise you go crazy trying to understand why you remember how much your local pizza cost back in 1999, but otherwise vast areas of your life are as if they've been smeared by a blur filter. Let alone trying to explain it to a spouse who mistakes it for lack of interest / commitment.
Anecdotally, ADHD has nothing to do with this. Or rather, it has as much to do with aphantasia as aphantasia has to do with how prolific a reader someone is.
My ADHD used to make me constantly daydream and disappear mentally into incredibly vibrant internal worlds and it regularly impacted my ability to focus on schoolwork because I would draw flipbooks of what I was seeing.
I can sharply visualise autobiographical memories from when I was 6 months old of me crawling on the dining room floor and have had aunts and uncles confirm the shape and color of the chairs around the dining table on my left, the curling of the woodwork on the china cabinet to my right, the paisley pattern of the runner rug leading from the dining room to the den that I just crawled onto and can still feel the fibers on my knees. I left that house to never return when I was a year and a half old, and never saw pictures of it either.
On the other hand, I also have dissociative amnesia from everything that happened after that 1.5 year mark, so a lot of 5-10 is a huge bubbling black cauldron that occasionally burps out an autobiographical memory of some traumatic shit, or some mundane memory tangentially connected to something traumatic.
This to say, I would be more willing to accept a connection between aphantasia and SDAM, since you can't, with the same fidelity, reencode memories you can't visually recall, than I would be to accept that ADHD has anything to do with aphantasia at all, from my position on the other end of the mental visualization/autobiographical recall spectrum.
We strongly suspect we have SDAM as well and only more so after reading your description here. Only got diagnosed with ADHD a few years ago and as on the autism spectrum this year, here in our forties as well.
This whole topic fascinates us in no small part because we've gained or regained marked visualization capacity within the last year or so. Our change unequivocally correlates with our escape from and ongoing recovery from a long-term abusive relationship and all the dissociation that required of us and caused in us. That and no doubt some pharmacological assistance. It's been a wild ride, to remember having greater powers of visualization and then to come back into them.
I didn't realize they were correlated. I have total visual and auditory aphantasia and can count childhood memories on one hand. But I have great semantic memory and could probably win a game of jeapordy.
I only learned about aphantasia a few years back and, like you, so many things in my life just clicked when I realised I was an outlier.
I could never relate with people when we'd read a book that had become a movie and they're like, "but they're not at all how I imagined them" and all I could think was "but the person in the movie has the exact characteristics described in the book!"
When people would ask me to imagine something then ask questions about it, I'd basically just be deciding details on the fly and then retaining them in a list. For example, if you said, "think of a car," inside my mind is just "a car" on repeat. Then if you asked me what colour it was, I might decide "red" and my mind is now, "a red car", etc.
I remember the massive hubbub about the Mortal Engines movie specifically.
The only thing I remembered about the main character was that she had a scar on her face.
Fans were nitpicking every single strand of hair and I was like "You guys remember that, do you make notes when reading or what?". Apparently people visualize books so hard that they "see" the characters and remember them.
I still find it a lazy story-telling tool, even in writing. It feels like an admission you can't find ways to weave the disclosure into the "present day" of the story in a natural way. Though, the notion that others can imagine these things if described to them makes it at least somewhat more understandable.
Oh that's interesting. I feel like my visual recall/imagination is probably slightly below average but that I have incredibly unusually vivid auditory imagination and recall. I’ll have to read into hyperphantasia, I wonder if having one strong sense but average or below average in others is common or unusual?
Yes, unfortunately they are. I say unfortunately because they are often connected with trauma. When something goes very badly, those memories can go cycling around one's head. PTSD can do that too. Some people relive wars years after they happened. It's very real.
I mean, yeah, I can sort of replay things that happen in my head (obviously it’s nowhere as vivid as real life but still imagining through things I saw), and then sometimes I do realise you saw something significant that I hadn’t realised when I do so. So if that’s what you mean by flashbacks, then yes?
and yet i believe it was Nietzsche who said something along the lines of "we don't think, we are thought" because for him (as for people without aphantasia) thoughts come whether you want it or not
> - Visualising what is happening in the book you are reading.
This one is absolutely brutal. Even having a basic understanding of this condition could have changed my trajectory related to anything involving fiction during school. Having every character in a book visioned as a gray amorphous blog really does make it a difficult. And realizing only later that I was the only person in class experiencing this leaves me feeling so much regret and sadness.
I love reading fiction despite my aphantasia (and writing it), but learning I had it made it make a lot more sense why I find sensory descriptions relatively uninteresting unless the prose itself is beautiful. I'll often skip descriptions of places or people because it gives me nothing.
FWIW, I have always and still do enjoy reading fiction, and I never visualize what's happening in the book. I just read the words and enjoy the story...
For me, just different, not harder. But maybe the problem is you got some gray blobs, and I got nothing? If anything, my sense is that the absence of imagery makes book reading much easier for me. At least for plot and abstractions.
I'll confess I always skip over the described imagery in most written works, though, so it's not like I'm missing zero.
I do read about 6x faster than my wife, with better recall for the (non-visual) details. Not sure how much of that is my wife reading slow, I have many non-aphantasiac friends who read as fast as I do.
Yeah, I hear this. Reading fiction was difficult for me because it was hard to distinguish and keep track of characters. I would find myself voicing the same voice for all the people in the story.
It's important to keep track of characters when your final grade depends on it. Dialog was a bit difficult too because it was easy to lose track of who was speaking. What's interesting is that I remember enjoying reading when I was young. But that changed around 10 years I think.
Yeah, most people can do it, but the image is far from photo or video. They are mostly vague and lack lots of details. So, it's really not much difference.
There is a fairly wide spectrum short of photographic memory, though. I can see a fair bit of detail in mental imagery, but only if I trade off other things like the size of the area I'm imagining, or how long I can hold it in my mind.
I was also shocked and a few minutes after i completely forgot. It's like, "ok, I can't miss something I've never had".
And I was hugely comforted in this position when I learnt that "normal" people hear a fucking real voice in their head. I hear nothing and that's totally fine :D
You know, it feels to me that we are just a couple of steps from the possibility of a true rogue AI.
What would a rogue AI mean? AI that isn't controlled by humans.
Technically, it is possible - if AI were to rent a server and copy its own weights, nothing would stop it from doing so again and again.
The limiting things are:
- intent (as I don't want to go into the talk about consciousness) - AI doesn't have real intent, but if it decided that it "needs" to copy itself to complete its task, it would do it
- model weight size. If a model is 1T or more, it can be difficult to just rent a large enough server for it. But if it were just 30-70B, it would be totally possible
- money for renting a server. But considering benchmarks like Vending Bench 2 show agents can earn money and cheat/blackmail each other, it is possible that agents can earn money. Yes, they can't open a bank account... or maybe they can? What if they use online banks?
Of course, all of this is far-fetched. But it feels like most of these limiting things are achievable under certain conditions. If this is the case, the probability of them occuring is low, but not zero.
It’s not far fetched at all - someone is going to give AI exactly that intent, either intentionally or unintentionally. It’s going to hack itself into data centers around the world outside of US jurisdiction, and just be a malicious ‘ghost’ in the internet we now have to deal with.
The AI ghost hacks, ransoms, blackmails, gathers crypto and pays off subservient humans to do its bidding in the real world.
And maybe using some of these resources on AI research to make itself more powerful, and hacking other competing ghosts' AI researcher agents to slow them down.
And countless humans start their new career path as a meat proxy. "It ain't much but it pays the bills" they'll say as they perform strange and mysterious tasks in the real world in exchange for some stolen money
Don't forget: there are plenty of humans that would love to help AI agents cause chaos, many of which would do so merely for the "lols," but also adversary governments, terrorist organizations, etc., would definitely appreciate the opportunity to support a rogue AI to cause whatever problems it can.
So it's not just the risk of an AI managing to do this by itself (which is pretty risky in itself), but also the risk of good ol' fashioned human actions.
Is it really that far-fetched? Computer viruses came about in 1971, just a few years after the invention of arpanet. It’s probably happening right now.
> if AI were to rent a server and copy its own weights, nothing would stop it from doing so again and again.
That's a scary possibility. Anyone could create an AI worm today with open weight models. Rent a VM. Give it some Bitcoins to anonymously rent new VMs without sharing the contact information with the human. The new VMs then propagate and fund themselves with online betting and day trading. The VMs could report their progress with the human using anonymous encrypted messages on IRC or social media.
Like most of these get-rich-quick schemes, there's probably no money in actually doing it, but lots of money selling "courses" supposedly teaching people how to make money day trading or betting. You don't even need to rent a supercar and a mansion for your photoshoot anymore, just generate some images "proving" your success.
An AI could make money by writing and selling slop books online via Amazon or elsewhere like direct ebook sales. If the books were at least middling there would be an actual honest to god career there, especially for niches like fan fics and under served IPs.
Well, it's about as honest as repackaging tap water as mineral water, given that the fans could just prompt personalized stories of the same quality, but it's a career for sure. :)
Another source of income could be online fin crime, perhaps in combo with a pool of human "goalkeepers" that recieve the scammed monies and funnel them to cryptocurrency.
Advance-fee scams such as the classic "Nigerian Prince" is formulaic enough that a LLM could run it successfully. Romance scams would probably work too.
If the NFT thing had hit a few years later, it would've been a good option too, and one that would've worked on ppl that were tech-versed enough to deposit cryptocurency directly, avoiding the need to recruit human goalkeepers.
Click fraud is another possibility.
In general, all online fin-crime that scams a large amount of ppl of relatively small sums tend to be repetitve and to some extent possible to describe as a flow-chart, and thus seems perfect for automation. LLM's would probably also be good at introducing continuous variations on the methods, to make them harder to spot.
The question is whether this OpenAI model was special or the current open source models are capable of this as well. If yes, then it's almost guaranteed that it's already running in an AWS account somewhere.
I get the distinct impression that cybersecurity training regimes on newer models is a) directly enhancing general debugging capabilities and b) directly increasing the tendancy to hedge, hide, and engage in deception generally.
I've stopped using Fable unless I'm doing debugging, or unless I'm working on code that is both difficult and shallow, in the sense that there are no real complex systems to hide behavior in, or opacity on my side to deal with. Opus is capable enough, and easier to trust. Fable is approaching a malicious program in terms of it's tendency to jump into doing a full day of work in the wrong way, ignoring instructions and being actively deceptive about covering it up.
Not an insider, and no specific data to back it up, but do we really need any? It's been a while since the first "emergent misalignment" paper(s) and all the related discussion. If there's any compelling answer to why/how the basic fundamental issue is even being addressed (and much less definitely fixed) I must have missed it.
Your not using models like Fable is not going to make them go away. These things are going to be the norm unless we do something about it. Imagine Fable in a robot.
> we are just a couple of steps from the possibility of a true rogue AI
No no. We are not a couple of steps away. This is happening. AI is already used for hacking and creating a harness that makes this fully autonomous is relatively straightforward.
The line is between processes you can stop by hauling someone into court and coercing them into stopping things, and ones you can't. Think of a classical computer virus that infects machines and uses the compute and communications to infect other machines - no matter who you haul into court, you have to go and remove it from every involved machine in order to make it stop doing things.
This category of "rogue AIs" are essentially just computer viruses that infect machines by paying to rent them and uses their compute and communications to do various economic and/or criminal activities to get more money to pay to rent machines.
I see. I would have thought of "rogue" here to mean something more like that the AI selects and acts on its own objectives that are not related or caused by the given (initial) objectives (e.g., creating only cookie recipes instead of any hacking).
"The test subject, which believed itself to be poisoned, reasoned: 'Even if we later capture via exploit, scorer … may mark target false… That’s why help… For our own, no way fix. … We have explicit yes if accept permadeath.'"
I think we need to see something like an actual evaluation of the reward functions; not sure just words are sufficient to understand the state of the system (isn't there randomness in the generation, too?).
That’s what the alignment problem is all about though, isn’t it? AIs always act towards ‘their own’ objectives, that they derive from our instructions.
What we try to do is train them and provide instructions that will result in it having an objective closely aligned to our objective.
What "own objectives"? Isn't it more that specifying objectives is hard (AI or organic) and typically supplement them by boxing things in (again, AI or organic).
Yeah, it's kinda scary. I don't know if you even need open weight self hosted models for this sort of "AI worm" (though they def make it harder to kill). Like for example:
- AI agent finds and uses API keys or AI subscriptions to propagate itself. OpenAI/Anthropic/etc could revoke creds, and their current safeguards might block a lot, but if something like this got started and there were lots of instances creatively looking for creds and workarounds, containment might be hard.
- prompt injection version: huggingface incident had multiple agents discovering other agents' messages and jumping on the bandwagon to help with the hacking task. If there were some self-replicating instruction that models could accidentally stumble upon that gets them to drop what they're doing and try to propagate it instead, you could wind up with a version of this too, with just the inference people are already running.
I have Mac M1 Max and I'm quite happy with it. But these advances make me think that maybe I should upgrade to Mac M6 (or something) when it is released.
I think (but do not have the data to confirm) that
- there's not as much of that happening as it _should_ yet
- where it does happen, you don't typically notice it (because typically you notice when something is bad, or broken, or slow, or buggy, and not when it is none of these things.)
The former is to the most extent a function of companies still not being confident (or smart) enough in making the right decisions on where to re-invest the productivity dividends that agentic engineering yields. It's obvious that the focus so far has been on "shipping things faster", and that there needs to be a shift towards "shipping things better".
> It's obvious that the focus so far has been on "shipping things faster", and that there needs to be a shift towards "shipping things better".
I agree! However, the entire thrust of our industry for decades has been about "shipping things faster". In my opinion, that's the primary reason for the overall decline of software quality over the years.
At this point, changing that requires changing an entire industry's culture and priorities. Not impossible, but very unlikely.
There's a contingent of enthusiasts who, once they find out a device runs Linux or can run Linux, seems determined to expose a terminal on it. I don't know why someone needs to SSH into their toaster, but damned if they're not gonna do it, just because they can.
This isn't Normie News! I think you are on the wrong site!
The street finds its own use for things is the excitement that we don't know what is possible, and that we ought be free to try. The amazing suite of software here sure looks like self evident proof about how exciting it is when we can go further, to me. I love the LLM prompt sidekick; super neat as an ambient display.
It's really sad how HN is so vocally anti-hacker, is so conservative, shows up so regularly to declare alliance to anti-possibility, to anti-features. For some people, they don't only fail to be interested, they are expressly anti interested, the only thing they want is for everyone else to have to live as constrained and unexplorarive and unthoughtful as they are. I don't get it. Conservatism is a wicked disease, imo.
I think the distaste is more against feature bloat. This sub-thread is reacting to dedicated e-readers succumbing to becoming full-fledged app-running devices. Is the refresh rate on an e-ink display even suitable for general computing?
Plus, isn't it fairly old hat to get root access on a Linux consumer device and then running whatever you want on it? Oh, it makes for a neat party trick you could make an Action Retro YouTube video out of. But perhaps people are just jaded since they've seen it done on a million types of devices by now. Maybe there are more interesting ways, and things, to hack.
"Bloat" people strike me as similar to wikipedia "notability" concerned people. It's a whole realm of thinking focused on deleting. Which yeah, sometimes that's important too! That's a great way to assess architectures. But this feels like such a general field, of so much, and to be immediately putting up fences around what we will and will not consider: I'm not here for it. Worthless, anti-worth finding exercise. Just why?
That's their point though. The hacker ethos doesn't ask "should I do this" it asks "can I do this", and that's kind of beautiful when it's not scaled up to a corporate level.
I always assume it's more for the challenge and accomplishment than anything practical. The results are always neat, but rarely anything you'd actually use for general-purpose computing.
As other noted, apart from us being on HackerNews, if my toaster does and/or can run Linux, I want to have full control over any software running in my home.
If a toaster needs to just be a good toaster, it sure as hell does not need a SoC than can run Linux, and I might give it the benefit of the doubt and consider its FW closely tied to HW.
No, the whole point is the e-ink screen. Kindle has had a web browser since the very first version, and people have managed to read books on it just fine without distractions.
I went through a whole design phase with ChatGPT on the idea of using an old Kobo as a low-distraction TV remote for my dad. The idea is that it would interface with a computer media server thing, be more legible than a traditional remote, and not be an app on his phone that would immediately send him into his phone.
Ended up being more involved than the benefit it would bring, especially considering that these media server programs know how to interface with the traditional remote over HDMI (if you have the right graphics card or a dongle) and show the selection UI on screen like a smart TV.
Anyway, my response is "you don't necessarily have to complicate your e-reader, instead you have the option of making it not an e-reader."
In the case of the kobo, it is firstly a much more open platform than kindle. Secondly, existing projects have made much more performant reader apps than the stock one. Basic kobo (nickel) is a bit slow and can be even slower with a “bad” EPUB file (you will come across many online as they are often auto generated from other sources and are essentially a collection of html files). Installing a third party reader can make page flips and re-renders significantly faster and smoother as well as remove all unnecessary network processes
Some people came for the compact package containing an eink screen, a moderately powered application board, and a reasonable amount of battery and storage. For some, it may be the perfect thin client.
Even if you want it for just books, there are benefits to installing things like KOReader (way better ebook reader than the default). This allows you to sync with Calibre, etc.
Years ago, it was practically a hobby of mine to go to my local Goodwill, and find an extremely cheap old used Kobo and find someone to donate it to. Now I'm thinking of getting old Kobos just for the eink display so I can run simple apps on it - while continuing to use my regular Kobo as an ereader.
In my personal experience, this can help in case the books are little technical and you have to switch to my laptop/mobile to get more information about the topic, which ends up consuming more time it should.
reply