I don't think anyone should really believe in Rails at this point. It was a monumental leap forward at the time but other stacks have shown that there are better ways.
Think of it like spending $1 trillion to become the next Google. That hardware itself may never turn a profit. But if 10 years for now you're the software provider that owns the ecosystem around too cheap to meter tokens you've got a money printing machine.
They're the next step in UI for computers. Telling codex to add a network interface is easier than a GUI which is easier than a terminal program which is easier than manually editing a config file. None of those are a crutch. They are increasingly useful tools.
Frankly I find terminal applications considerably easier than most GUIs for most purposes, but I agree that natural language is generally an easier interface layer than either, and that an interface layer to the functions of a computer in general seems to be LLMs' destiny as a technology
I’ve been using command lines for over 25 years now and the thing that rubs me the wrong way about them is that nobody can agree on conventions with regard to how a CLI utility should work, which limits the amount of knowledge that’s transferable between them. Each effectively has to be learned individually which gets old fast.
I mean I think there's about as much convention and overlap as there seems to be for GUIs at least, and with much more discoverable clarification to boot
The issue with manpages is that they need to be pulled up independent of the program, they’re overly verbose, and often don’t include practical usage examples or if they do, bury them under miles of flags I’ll never use.
Sometimes CLI utility help commands fill the gap, but often they’re too spartan or don’t even exist.
Sure, documentation could always be better. My first attempt is usually tab-completion in the shell, then I'll try --help or something, then I usually go to either the archwiki or web search, which increasingly these days is equivalent to asking a chatbot, or asking an llm in a TUI. I find I'm generally not getting past the --help part of that more than 10% of the time, and the web or wiki search solves my problem with one query 99.99% of the rest of the time, but I'm willing to believe this is aided by experience.
No amount of experience using GUIs stops me from about 40% of the time I don't already know how to do something having to find a youtube video that explains what arcane path I need to take through a menu to find a setting, and if the menus have changed I might have to find the part of the video that explains what I'm looking for, watch it, realize it's changed, and hope I can find a more recent one. This goes for both native and web applications, for mobile and desktop, for every operating system I've ever touched. Sometimes there's a text explanation I can find, usually via a web search or llm. Sometimes I can guess at where something might be (Tools > Preferences? Wait maybe it's settings? Is this one a view setting? Hamburger menu? O wait no that one's in account settings. Ah, that one's under a different specific tab because the org that makes the software considers it an integration and also they A/B tested that it's more discoverable in this other place so they moved it since the youtube video I saw about it). On a web browser it's helpful to be able to dig into the page via debug consoles, sometimes I can search for some string I know is the name of the setting, or a value that's probably currently set. I can do all that, if necessary, but I'm certainly not comfortable claiming it's "easier"
In no way is a stochastic system that has zero ability to understand anything "the next step in UI". That's a recipe for disaster and even greater frustration.
Plenty of people support Flock or at least its mission. They're just not vocal because the opposition is vociferous. Like most things the country is fairly divided. This poll has it 46% opposed and 38% support:
But MAD means you can't use them. It might save you from an all out killbot invasion but it doesn't stop asymmetric warfare or being cutoff from the rest of the world.
MAD is a belief system. It doesn't matter whether AI can effectively stop the retaliation, it is enough that people making the decisions believed that it can (are we there already?).
They didn't want his name removed from any paper. The invitation was to write a new joint paper between Buckmaster and OpenAI. An invitation Alpoge couldn't accept and OpenAI wouldn't make since he worked for a competitor.
This seems unsupported. OpenAI has access to internal models that the general public doesn't have and a compute budget that dwarfs what an NYU professor would have.
it's very possible they only had to use the massive compute budget because they were trying to plagiarize his work before he published it though, e.g. autonomously do things in ~7 days what he had likely been thinking about for ~1 year.
Of course you do if you're a) only given a partial solution b) racing against someone else using a competing AI.
The open question was whether their LLM got the nudge in the right direction because it got access to the chat somehow (e.g. automated training that scraped his chat logs) or just a high level "Navier stokes can be solved through LLM". It sounds like the former may have happened although right now we just have an accusation and a weak denial.
wasn't it reported elsewhere that they used the equivalent of $22M (street) in Astra tokens? obviously it's not the same when you own the machinery but still.
reply