Hacker News new | past | comments | ask | show | jobs | submit | baijum's comments login

That's a very fair and critical point. You're right that we can't change the fundamental, probabilistic nature of LLMs themselves.

But that makes me wonder if the goal should be reframed. Instead of trying to eliminate errors, what if we could change their nature?

The interesting hypothesis to explore, then, is whether a language's grammar can be designed to make an LLM's probabilistic errors fail loudly as obvious syntactic errors, rather than failing silently as subtle, hard-to-spot semantic bugs.

For instance, if a language demands extreme explicitness and has no default behaviors, an LLM's failure to generate the required explicit token becomes a simple compile-time error, not a runtime surprise.

So while we can't "fix" the LLM's core, maybe we can design a grammar that acts as a much safer "harness" for its output.


I would say we have this language already, too. It's machine code or its cousin, assembler. Processor instructions (machine code) that all software reduces down to are very explicit and have no default values.

The problem is that people don't like writing assembler, which is how we got Fortran in the first place.

The fundamental issue, then, is with the human language side of things, not the programming language side. The LLM is useful because it understands regular English, like "What is the difference between 'let' and 'const' in JS?," which is not something that can be expressed in a programming language.

To get the useful feature we want, natural language understanding, we have to accept the unreliable and predictive nature of the entire technique.


Probably Red Hat Universal Base Image would be good enough for development instead of waiting for CentOS; https://www.redhat.com/en/blog/introducing-red-hat-universal...


Nice. I'm currently building an operator and this comes in handy.

Quick question: how do I differentiate between freely available and subscription-only containers on the Red Hat Registry?


If they are in the ubi namespace, they are freely available. The Container Catalog also will tell you if you can pull without a login when you look at the details of a container.


Can these containers be run hosts that are not RHEL? It seems like it's allowed, but I'm not completely sure I read that right.

I'm also looking to know what packages are available in RHEL8 that are not available to UBI containers. I'm not able to find information as to what subset of the RHEL package universe is available to UBI containers. If you're aware of information on this I'd love to be pointed to it.


The article is clear on it being allowed.


There is no photo of the mathematician in the Wikipedia page!


OP here. I posted this here with the exception of getting more feedback on the topic. Thanks!



This issue has been fixed in upstream. https://issues.jboss.org/browse/RHDENG-1320 Please try it again after some time.



Why would I be digging through Red Hat's blog to find this, and not find it on the site of the actual thing itself?


I was expecting to find a video in the link you pasted, almost missed it. Here's the actual link to the video https://vimeo.com/215402513/02867b4aea


That video adresses a bunch of the issues noted. It briefly shows provisioning, the editor, deployment, issue tracking in action, and the narrator overviews the vision & reasoning. It should be touted more prominently.


Sorry for the inconvenience! I have reported this issue here: https://github.com/openshiftio/openshift.io/issues/134


The upstream for openshift.io is primarily in the fabric8 project. fabric8 is found in the following organizations:

https://github.com/fabric8io https://github.com/fabric8-ui https://github.com/fabric8-analytics https://github.com/fabric8-quickstarts https://github.com/fabric8io-images



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: