> persists productively. if agent bankruots its owner or.itself.then that is not productive (who funds paper clip maximizer?)
So, your standard for how risky it is, is simply how competent it is?
That's fine right up until the thing passes an unknown threshold, one which will only be visible in the rear view mirror:
The covid-19 virus itself is not what anyone would call smart, ditto HIV, smallpox, and bubonic plague, but a genome is much the same learning system, and they still killed millions each.
> superficially satisfy their constraints.
The "superficially" part is one of the reasons these things can be dangerous. e.g. hopefully nobody at OpenAI actually wanted their wildly-sycophantic version, but yet they created it.
I mean, the source for claude code was "leaked" by accident so at least some of their processes are not that secure. I feel that they are more like a Startup then a Enterprise (ignoring finances).
Tbh, there really needs to be some legal precedent set that makes model distillation a legal activity. If the model makers can rip everyone else's work and launder information as if it's their own without giving credit back to the original creators, I don't see why it should be illegal to distill the models. It's the same thing the frontier model makers are doing to IP everywhere else.
> Yes, and people will sit there and sip tea while waiting for "someone"? For how long?
Until someone cares enough to do it. This is open source software. When it comes to open source, the golden rule is you either do the things you care about yourself or stfu.
Given the libav fork wasn't all that long ago, it can obviously happen to ffmpeg just as much as it can happen to any other project.
Despite the protests, he admits using AI and then charging his clients full price...
"But maybe I will ask Claude’s opinion, and if one of the suggestions is smart—cutting a paragraph, for instance, or clarifying a sentence—I might accept it.
When I started translating 15 years ago, we used to paste uncooperative sentences into Google Translate to see if it had interesting ways to phrase things differently. Then came DeepL—same idea."
All big generators have an exciter coil that is used to generate the magnetic field. It has the advantage of allowing voltage regulation through adjustment of the field, rather than after the fact, which would be far less efficient.
In both motors and generators, there is an efficiency hit related to the need to supply power in order to generate the field, but when you scale up the system, it actually becomes more efficient to use the electromagnet. With the rare-earth mineral shortage, it makes even more sense.
https://news.ycombinator.com/item?id=48496420: Time and time again at many companies, including well-reputed ones, I have seen that preventing issues gets you no recognition, but building a giant pile of kindling and then putting out the inevitable fire will get you recognition twice. Even in "good" orgs.
Tectonic is a cool project, but hasn't seen any significant changes in a few years---and likely won't anytime soon. It seems we maintainers don't have the time and motivation to put serious work into Tectonic.
I haven't looked at the code in years (and thus may be wrong), but here's a quick overview:
Tectonic's code consists of thin bindings to /harfbuzz/graphite/etc and a vendored XeTeX port, driven by Rust that tries to keep the TeX environment predictable and sane. A few components have been fully ported to Rust (bibtex, spx2html), but the project is very unfinished.
I've looked into the dark corners of TeX when I worked on Tectonic, and it is not pretty. TeX relies on a stack of evil hacks and esoteric behavior that is very hard to replicate, and very difficult to expose in an ergonomic way.
A quick example: code highlighting does not work in Tectonic. The canonical solution is https://ctan.org/pkg/minted, which spawns a python process to style your code. Reproducibility is one of Tectonic's selling points, so we cannot replicate this behavior.
With https://typst.app/ as good as it is, there's little motivation to modernize TeX---especially considering the effort required. Typst _is_ modern TeX, and I'd rather spend my time there.
As I replied to a child comment - this is a nice idea that just isn't tenable in reality. AI hardware isn't just hilariously faster than consumer GPUs, it's also hilariously more power-efficient and has hilariously better connectivity. Every one of these dimensions kills the idea.
The far, FAR superior power efficiency means that even if you did harness every public GPU or GPU-like device on earth, you'd end up consuming so much excess electricity it would be cheaper on net to simply take the money that would have gone to the power bill and spend it on your own datacenter.
And even if electricity was free, having those GPUs spread over the world with internet-level latency will slow everything down by factors of thousands to millions - if it's feasible at all. Regardless, you're not getting fable-oss this century, or maybe even this millenium.
It's a nice idea, like I said, but it just doesn't make sense when you look at the reality of how training works. It would be far, far better for governments to own their own datacenters, maybe as a coalition, and dedicate their operation to the public good.
There is nothing more surreal in AI chat than entering your own name and being told you are a banned topic. Open source models must win. There is no alternative.
Out of curiosity, can that model be trained from the beginning without touching "sensitive" areas and remain useful in others? Will it be able to help in building biological weapons without being trained on articles and books about biology/ medicine?
Given that it's most public use in open source so far is to whitewash GPL code into MIT code, no, I'm sorry, I don't think "open source AI" is particularly important.
I will say they do negotiate it if it’s excessive. I have negotiated flat rates on both public and DC egress due to scale and architecture and they were open to it.
There is an image crisis. Yes, it's not a badly paid profession. But the perception that it's a dead end will lead to a sharp drop off in the student numbers.
Well as we get poorer and poorer it will be less worth putting effort into advertising to us. Im guessing AI will instead focus its effort on convincing rich people of various things.
Not just interoperability with Objective C but with C (full) and C++ (increasingly better but not full) as well.
Swift is also interoperable with different versions of itself courtesy of the Swift stable ABI (Application Binary Interface)[0], which they invested a significant amount of time into at the expense of adding other new features to the language, which have come along later.
Rust offers a different approach: recompile everything and static linking.
So, your standard for how risky it is, is simply how competent it is?
That's fine right up until the thing passes an unknown threshold, one which will only be visible in the rear view mirror:
The covid-19 virus itself is not what anyone would call smart, ditto HIV, smallpox, and bubonic plague, but a genome is much the same learning system, and they still killed millions each.
> superficially satisfy their constraints.
The "superficially" part is one of the reasons these things can be dangerous. e.g. hopefully nobody at OpenAI actually wanted their wildly-sycophantic version, but yet they created it.