Red Hat Developing AI Tool "Log Detective" To Help Developers

[phoronix]

Phoronix: Red Hat Developing AI Tool "Log Detective" To Help Developers

Jiri Kyjovsky of Red Hat has shared news today of Log Detective, a new tool being developed that will leverage an AI model to help in analyzing build failures for RPM packages...

https://www.phoronix.com/news/Red-Hat-AI-Log-Detective

[spicfoo]

I would like to see concrete examples and a clear model but I do like that they are paying attention to how they are collecting data instead of mass feeding random things without a proper license as many AI models do.

[Panix]

I would like to see concrete examples and a clear model but I do like that they are paying attention to how they are collecting data instead of mass feeding random things without a proper license as many AI models do.

There's nothing to 'like' about this. This is where Red Hat/IBM is going and where most of the tech world is going. Windows, too, obviously. The ppl who scoffed at AI - they probably know deep down they were wrong and feel stupid now.

[trapexit]

As someone working in the AIML space...

Why not just invest the time in building better tooling and messages?

If no one was willing to do that but are willing to fool with AIML... fine... but the problem was known prior and is (hopefully) fixable without such technology.

[spicfoo]

There's nothing to 'like' about this. This is where Red Hat/IBM is going and where most of the tech world is going. Windows, too, obviously. The ppl who scoffed at AI - they probably know deep down they were wrong and feel stupid now.

You say there is nothing to like and then say that's how the industry is moving? This is a confusing response

[spicfoo]

As someone working in the AIML space...

Why not just invest the time in building better tooling and messages?

If no one was willing to do that but are willing to fool with AIML... fine... but the problem was known prior and is (hopefully) fixable without such technology.

You can't really control how tens of thousands of independent software components with different build systems choose to log their error messages. There is no standard pattern for that

[vextium]

AI is a tool and, like any other tool, it can be used for evil, but it can also be used for good. Open source needs to make their own AI and embrace it. If we don't, there will not be anything stopping tech giants from using the technology for evil and self-interested reasons instead of actually pushing the industry forward.

[trapexit]

You can't really control how tens of thousands of independent software components with different build systems choose to log their error messages. There is no standard pattern for that

We are talking about Red Hat and IBM here. They contribute massively to the space. They absolutely could be improving DX via traditional improvements. This product is literally created to deal with their own tech stack (RPM).

[spicfoo]

This product is literally created to deal with their own tech stack (RPM).

Nah, RPM is just the messenger here. The error message comes from the builds of the various open source components. I doubt anyone can standardize that readily or it would have already happened.

[trapexit]

Nah, RPM is just the messenger here. The error message comes from the builds of the various open source components. I doubt anyone can standardize that readily or it would have already happened.

Then why is the thing expressly mention RPM multiple times? And talk about how RPM packaging debugging is tough?

I package my own projects in deb and rpm. They are indeed a pain to work with. But that's because of terrible documentation and lack of examples and poor general DX. I'm positive I'm not stressing what those package managers can do but step 0 in improving things starts with your own tooling. Problems with autotools, cmake, compilers, etc. are almost never my issue when dealing with packages.