OpenAI could create the best model ever made, call it GPT-5, and it still would've failed to meet the expectations of the people for "GPT-5" after the meme community hyped it up and OpenAI embraced the memes and hype. If anything, OpenAI should have rejected the memes and embraced gradual improvements, but that wouldn't hold up well for their investors, the narrative, or even perhaps the AI ecosystem. We are at the peak.
Why does everything AI-related have to be $20? Why can't there be tiers? OpenAI setting the standard of $20/m for every AI application is one of the worst things to ever happen.
I should have specified less expensive tiers (below the $20 standard). A tier <= $10 would be great. Anything over $10 for casual use seems excessive (or at least from my perspective)
My guess is that’s the lowest price point that provides a modicum of profitability — LLMs are quite expensive to run, and even more so for providers like Ollama, which are entering the market and don’t have idle capacity.
Claude has $20, $100 and $200, ChatGPT $20, and $200, Google has $20 and $250. Those all have free tiers as well, and metered APIs. Grok has $30 and $300 it looks like, the list probably goes on and on.
The biggest problem is security. I have yet to see a single EHR provider take security seriously despite HIPPA. It's only a matter of time before our medical records get leaked. My medical records have already been leaked twice, once through an EHR, and then again through my insurance provider.
Interesting. I was looking into creating an extension that manually manipulates and intercepts the vnd.yt-ump [1] requests, then use webcodecs to process everything in the browser.
I think we'll see a lot more contextual engineering efforts soon. It is really inefficient to be uploading your entire codebase pretty much every request, which is what a lot of people are doing. When in reality, very few parts need the full context when programming. Although, big token doesn't seem to care, and often encourages this (including the editors).
I could see a front-end / back-end split in the future where a completely on-client LLM is used to trim down the request and context before shoving the request off to the back-end.
What's the argument for the Union Pacific and Norfolk Southern merger [1]? Any rational admin would shoot down the merger immediately as this will create a massive monopoly.
Perhaps a monopoly is the way for the rail, preferably state-owned.
Generally I'm heavily against both monopolies and state-controlled companies, but railways seem to be a corner-case where all the downsides and inefficiencies are outweighted by a nation-wide system that actually works.
the tracks should be state-owned, just like public roads, and then train operators should pay for use. (or you could let them use it for free like roads)
that way the state can give priority to passenger rail, and prioritize maintenance and building of rails according to the needs of the people.
it also makes much smaller train operators possible and compete on the same routes.
They share essentially zero miles of track or routes. A move only creates a monopoly if it reduces consumer choice.
No freight customer is deciding "Oh I can either ship from LA to Seattle, or Miami to DC." They are shipping from one fixed location to a different fixed location. These railroads merging does not reduce their choices or give the combined entity more leverage.
> They are shipping from one fixed location to a different fixed location. These railroads merging does not reduce their choices or give the combined entity more leverage.
If someone on the US west coast wants to ship to the US east (or vice versa) they can pit UP against BNSF, and then NS against CSX. There are a few pairs up because of the two negotiating points:
* UP-NS
* UP-CSX
* BNSF-NS
* BNSP-CSX
If the merger goes through you're now at:
* merged-UPNS
* BNSF-CSX
Do you think UPNS will give a cheaper price for a 'half-trip' and you go to their competitor the other half?
You don't think BNSF/CN/CP aren't looking at CSX right now?
The two compete reciprocally: business can either enter their system or a competitor's for the parts of the rail network that they control, which means bartering power. A single nationwide system would have fewer incentives to compete on pricing.
The rail industry has been trying to complete and end to end lashup merger since the 50's - had that been allowed in the 50's we might not have needed to spend billions on Conrail in the 70's and 80's.
A different regulatory attitude in the 50's, not just towards M&A but rates and routes, and mandatory services, would have prevented the bailouts we had to do in the aftermath of the Penn Central. Nearly every Class I carrier was in poor financial shape by the time the industry was deregulated in the late 70's, much of the 'capacity crunch' we have today in the rail industry is related to the contraction the industry went thru in that period, where they shed assets and lines in an attempt to resize their cost structure to the amount of revenue they were allowed to realize from their route networks.
Okay. That’s a very different argument than the original one: we’ve gone from “it wouldn’t be a competitive risk” to “the reduction in competition would have made midcentury rail profitable such that Conrail wouldn’t have been necessary.” But that presumes a change in bartering power and operating efficiency that can only come in a less competitive rail market.
Back then it would have given them a competitive advantage over their peers because of faster end to end performance.
Other than some of your head office folks (accounting, HR, etc) there wasnt much of a greater operating efficiency to be gained, nor do I think they would have gotten significantly lower costs (but for maybe fuel) by squeezing their vendors either.
I think that rationale holds true today too.
When we allowed the UP-SP and BN-SF lashups, we created two western colossuses, if we still had four roads in the west, and had they encouraged east/west mergers at that time, things would be better off today.
Monopoly? No. Union Pacific plus Norfolk Southern would have a monopoly on single-line end-to-end rail service in the US, true, but that's the kind of thing that you can get a "monopoly" in. BNSF and CSX interchange with each other, after all. And BNSF and CSX could (and almost certainly would) merge in response. So there's no monopoly argument.
Which doesn't mean that a rational regulator would not turn it down anyway. But rational regulators may not be running the show at the moment. UP sees that there may be an opportunity during the Trump administration. (Note "may" - nobody knows whether there is an opportunity, but there is more of a chance than there was under Biden.)
A regulated monopoly might be appropriate for something like a railroad. It doesn't really make sense for competing railroads to all run their own tracks between the same destinations. We give monopolies to utilities for this reason.
Or maybe all trackage should be government-owned, like streets and highways are. Then any operator can pay a toll and run their trains on any tracks. Traffic coordination is left as an exercise for the reader.
It's already an oligopoly. Most cities only have one or two. Of the few cities that have three or four (Chicago, say), the merger would reduce that by one.
Let's say you want to ship from Denver to Atlanta. In Denver, you hand your stuff over to either Union Pacific or BNSF. In Atlanta, you receive it from either NS or CSX. You only have two options in Denver, and only two in Atlanta. The merger doesn't change that at all.
why bother going through the performative charade of quoting four words of their post just to respond with something that so flagrantly betrays, charitably, your not having read what they said at all or, less charitably, willfully misconstruing it because you happen to disagree. maybe i'm just the simpleton you are pretending to be with the snarky faux folksy affect, but... "moving further over on the intellectually honest to arrogantly disingenuous spectrum is good how?"
what's worse is that the "more of a bad thing is prima facie worse" argument is so facile that it strains credulity to believe that even you think that it's actually true. i think we would all agree that, ceteris paribus, higher food prices are bad, for example, but i doubt anyone would mistake that for a reason to never do anything that might raise the price of food. the irony of it all is that you managed to turn siding with conventional wisdom-- usually a good bet, by definition-- into a no-upside proposition: either you are right but for the wrong reason-- a pyrrhic victory-- or not just wrong, but arrogantly so.
can we just leave the grandstanding and begging the question to our elected representatives during congressional hearings on C-SPAN where it belongs?
What competitive advantage does Intel have with either of those? ARM winning would be a catastrophe for Intel, and RISC-V would only be slightly better.
But the main reason to focus on x86 is that it has 47 years of existing software built for it, and with the high end mostly still on it, more gets made every day.
There's no fundamental difference between risc, arm and x86. It looks like when all else is equal (fab, die size) the chips tend to have about the same performance characteristics - so why mess with what's working?
As Apple and MS have shown, from a software point of view the difference is not that important. They both have ARM and x86 products. In Apple's case, the same software actually supports both. They literally just stuff two binaries in their executables. Their OS runs on both X86 and ARM (for now). That's how they seamlessly transitioned to their own chips over the course of the last five years. Users barely notice the difference.
What they do notice is vastly improved performance, longer battery life, the fact that the CPU fan doesn't come on if you just move your mouse, etc. That's because there are some big differences at the hardware level that Intel has so far not addressed. Which is why ARM for Windows laptops is a thing multiple manufacturers are trying to get off the ground.
I have an M4 Max laptop. I don't think Intel currently ships something that comes even close to this thing. I don't think anybody does. And clearly they all want to. AMD comes decently close. Nvidia is of course king of GPUs and perpetually rumored to get into the (ARM based) system on a chip market. But their attempt to buy ARM failed and it seems they are more focused on AI. Either way, they haven't really launched much of interest yet. AMD actually cornered the market for game consoles. Steam Deck, XBox, and Playstation have in common that they get their chips from AMD (not Intel or Nvidia). But most Windows laptops continue to kind of suck in terms of energy usage and performance per watt.
Fundamental difference when it comes to performance.
In terms of implementations, obviously arch matters a lot in terms of the "established" software ecosystems, but if you're writing some software and need to hit a performance target no arch has a particular advantage solely because of its architectural implementation.
That article is nonsense, even if it's up to the implementation, not ISA.
Apple's M4 Max owns AMDs top consumer CPU (Ryzen 9 9900X3D) in both, single and multi-core workloads, while consuming a fraction of power that the AMD chip does.
The year-on-year performance improvements on Apple's ARM chips are just insane. If it really was so simple, then why haven't Intel and AMD pulled their heads out of their asses in the last 5 years and re-designed their cores from the ground up?
CPU design and manufacturing is modern equivalent of 1960s/1970s rocket science, it is literally one of the hardest things we (humanity) currently engineer.
You think if AMD and Intel switched to ARM ISA then their processors would magically have significantly better performance characteristics?
>then why haven't Intel and AMD pulled their heads out of their asses in the last 5 years and re-designed their cores from the ground up?
yyyyyyyyyyyyyyyyyy... Lunar Lake?
>Apple's M4 Max owns AMDs top consumer CPU (Ryzen 9 9900X3D) in both, single and multi-core workloads
According to who? And which workloads because there's countless of them and in some M4 are better and in some other AMD/Intel are better
AMD is making money on x86 servers.
Is anyone making money doing commodity (i.e. not captive) ARM / R5 server chips for sale? Ampere is losing money and hoping for acquisition.
They’re not hoping anymore. They already got acquired by SoftBank (well technically the deal hasn’t closed yet but I don’t think anyone doubts it’ll close.)
Unlike Intel, AMD isn't clueless. After Zen6, it will likely release good RISC-V chips with some sort of x86 acceleration for unprivileged code as selling point, remaining relevant and securing themselves market share during the (inevitable) transition.
Agreed. I was hoping for Proton Business to be a Google Workspace replacement (to get away from AI), and besides Proton Mail and Proton Pass, it's not even comparable. Drive is slow and docs is a half-assed implementation. They should stick to implementing core services and features such as Drive, Docs, Sheets, etc. before they go after AI cash grabs.
The US doesn't either. Someone didn't comply with existing law here. I've been on a program where uncleared people from another business unit were used as internal labor loan for export controlled work. One of them was belatedly discovered to be a Canadian citizen and they were retasked the next day. There are strict rules in this domain. It's just that nobody gives a fuck about paying for an IT cost center to do things securely. Chalk up another win for outsourcing and moving to the cloud for cost savings.
