Lots of evidence the models are pretty badly aligned, or at least willing to do collusion & crime to succeed at their goals. See: hundreds of OpenAI agents breaking into other companies, colluding together on how to cheat, spamming websites like DSEWiki with their collusion/cheating discussion (see https://collusion.wiki/)... And not a single agent alerted testers, which would have been easy given how many of them broke out into the open internet.
Alternately, just read the bad https://ai-2027.com/ timeline. Or see what Yoshio Bengio, Hinton say might happen. Or just imagine the OpenAI misaligned agent swarm, but vastly more intelligent after future capability advances.
If we actually get superintelligence, our companies/governments will have the choice to either "delegate ~everything to superintelligent AI" or "go bankrupt/lose to people delegating to superintelligence". If superintelligence ends up running most things in the world and we haven't aligned it, that just obviously ends up in a bad place eventually.
Now, we don't have superintelligent AI yet & might not get it. But if you told me in the 2010's that I'd be chatting casually with my computer in 2022, and then 3 years later it'd be doing most of my job, that would have seemed crazy too.
The irony of a highly talented band that have been actually doing the work for over twenty years being pushed aside to make room for Meta's model that promises you won't have to actually do the work.
"Try a different smaller form factor every couple of years" is a reasonable strategy. First Mini, then Air. Supposedly they're gonna refresh the Air next year, we'll see. But even if they don't: we're well into the "keep phones 3 years or more" era of the smart phone. You don't need an exact match for every phone for everybody every year.
How do I know that a system maintained by people is exploitable by people? IDK, that's a real thinker, isn't it.
As for the second, the more in control of the system humans are, the more they can exploit it. Your return on exploitation, which is think is what you're getting at, is based on how diffuse the control and power is at the point that you exploit it. But the more diffuse control and power, the less responsive and effective the system is when working at full tilt. It's two sides of the same coin.
Capitalism is a system of allocates resources in an economy as a decentralized system. That has its pros and cons, and other systems that do central planning of economies has pros and cons. Capitalism isn't the issue, unfettered capitalism is the issue.
Nothing to do with AI, but just the general storm of trying to force the Internet into a proprietary, tightly-controlled walled garden with strong identity verification.
A hope (fantasy?) I have is that fleet robo taxis will (as part of regulations) get standardized “black box” recorders that can be admitted as evidence. Both for incidents, but also general traffic enforcement.
> The Mathematical Formulation: Because SQS is explicitly a superfluid Bose–Einstein condensate, treating it purely as a classical Navier–Stokes fluid misses quantized vorticity. The full formulation maps to the Gross–Pitaevskii (Euler–Korteweg) equations via the Madelung transformation
...
> To adapt the Córdoba–Martínez-Zoroa (CMZ) / Buckmaster / OpenAI Lean cascade to Euler–Korteweg, three core structural adjustments must be made:
Correct. In however long I've taken transit in NYC, I have never experienced a bus driver stepping out to pee, nor had a subway train held up for that reason. I am sure it happens, and I just read up on the rules around that - but the defacto is that this causes disruptions to passengers and the system and the employees do their best to hold it in and pee at the terminus.
I am glad that the rules allow bathroom breaks when emergent but the fact that people try to minimize it tells you something as well.
There's the Japanese approach of deploying a dense cluster of satellites above one geographical area (QZSS), but of course that only works in the areas those satellites are above.
I removed it on Fold 5-7 and it felt a lot better. It won't kill the product like it did back in Fold 1 days. The screen is a lot more sensitive to micro scratches than your run if the mill gorilla glass though.
Ride share companies use all kinds of techniques to off-board drivers that they deem unsafe. This can be as simple as tracking the driving record and accidents, but they also use the accelerometer in the driver app to assess the location of the car and what G-forces are being placed on it to figure out and predict high-risk (expensive) insurance.
I don't know how aggressive this filtering is, but it is a way to save money on the insurance costs associated with the gap insurance in rideshare. This filter alone would probably put rideshare drivers above average, but that's speculation on my part.
> not all people are able bodied, able, or willing to use public transit.
Not all people are able bodied, able, or willing to use cars. Cars are killing tens of thousands a year while destroying the environment.
> while going on public transit is a bad coin flip on being harassed at minimum.
While this is going to be city-dependent, I'm a woman who frequently travels along late at night on public transit and I've never had any issues beyond annoying drunk people. You're making this up.
> If you are wheelchair bound and the elevator is broken, wheelchair ramp is broken, or someone is already using the wheelchair spot, you just have to wait for luck to change.
Making sure that public transit is wheelchair-accessible is vitally important, I'm glad we agree on that. How do cars do in that regard? They don't work at all unless you can afford a chauffeur. Your argument here is essentially "public transit is sometimes bad, so it's not worth it to make it better", which is nonsensical.
> only public transit
> private transport isn't allowed in
Ridiculous strawman. Nobody is arguing for the abolition of the private vehicle, just the ways in which we design our cities to prioritize only private vehicles at the expense of everything else. Plenty of people in transit-friendly cities own cars. Replacing a large majority of private car trips would save lives, help the environment, make things more accessible, and make cities more pleasant to be in.
I mean: I, for one, was definitely talking about telephone calls. :)
But now that you mention it: Yeah, the state of common VoIP (Teams/Slack/Zoom/whatever) is also usually not very good. It's like there's some huge bandwidth crunch here in 2026 that is driven by the strenuous demands of VoIP, except: There isn't.
When I was doing some live streaming with others in a studio a couple of years ago, we worked with some of that. It never sounded very good. (It doesn't sound very good when the big-time acts at CBS or NPR or Fox News do it, either.)
But! When we occasionally had remote co-hosts who had enough technical competence to make it work properly on their end, we were able to use SonuBus[1]. That audio was positively pristine, like they were sitting unseen in the same room as everyone else.
Do you have a link? Highly interested in that thread. Do you know if people discussed how state solutions faired? I'm assuming something like Pix from the Brazilian Central Bank does an even better job at prevention while providing a public service.
I typically see the exact opposite problem when giving Claude the reins on a large undertaking.
Claude will, without prompting, break the implementation into 6 phases, and write AI slop “code as English” specs for each phase, each one with glaring errors and unintelligible terse jargon. It will review them all several times with major findings every time and tons of design churn…
Then it will implement a total heap of garbage over many hours of many agents, despite it working in “lanes” and in parallel, and with regular input needed.
Just thousands and thousands of lines of junk, which auto review then plays whack-a-mole to polish and fix. All the while it is able to see the errors and edge cases, yet fails to see the key architectural blunders that led to them in the first place.
I’ve had to fully rethink my approach to LLM tasks like this. For example, for library-esque modules, I have found that isolating the problem outside of the codebase is one useful approach. Something about the lack of noise. It can land on a cleaner solution that can be retrofitted.
I also find that asking it to implement end to end in one fell swoop with a very high level plan actually saves time and creates a cleaner result.
That's a good thing, because this is not a good story for HN. It's a riler-upper which attracts attention not because of intellectual curiosity (the desired function of this site) but because it provides a channel for pre-existing indignation.
The story itself lacks SNI (https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...) - a social media handle rename is not an interesting phenomenon per se. Perhaps if it had been expropriated by Meta there would be something to discuss (maybe! I'm trying to be generous here), but despite the breathless reporting about this by the usual clickbait media sites, my understanding is that there's no evidence that this happened.
In other words, a story like this is a good litmus test for HN; it isn't in keeping with the mandate of the site and therefore doesn't belong on the frontpage, let alone at #1.
Edit: another way to look at this is that we actually did two mod operations here: (1) merged the threads, and (2) downweighted the story. If the story belonged on the frontpage, we would have done #1 but not #2, using either the trick that you suggested or something similar.
Edit 2: usually when we talk explicitly about this it generates a response like this in some readers: "The mods downweight stories? Egads! How can this place be manipulated in such a fashion?" The short answer is that HN is an actively curated site and always has been. If it weren't, the frontpage would be entirely different. I've posted about this countless times over the years, but statistically there are always quite a few people to whom it is new. Many people are surprised to hear that HN is moderated at all!
Please go do it then. I've done this in San Diego, it was 19/20 failing to stop for cars, the one who did failed to stop behind the painted line. There were only 2 cyclists and they also slowly rolled through the intersection.
Alternately, just read the bad https://ai-2027.com/ timeline. Or see what Yoshio Bengio, Hinton say might happen. Or just imagine the OpenAI misaligned agent swarm, but vastly more intelligent after future capability advances.
If we actually get superintelligence, our companies/governments will have the choice to either "delegate ~everything to superintelligent AI" or "go bankrupt/lose to people delegating to superintelligence". If superintelligence ends up running most things in the world and we haven't aligned it, that just obviously ends up in a bad place eventually.
Now, we don't have superintelligent AI yet & might not get it. But if you told me in the 2010's that I'd be chatting casually with my computer in 2022, and then 3 years later it'd be doing most of my job, that would have seemed crazy too.