I'm not writing another cringe "Fable 5" post

And instead, teaching other "uncool" and useful concepts that actually matter.

Anthropic released Fable 5 a few days ago, and everyone is going crazy over it.

Except me.

I have an unpopular opinion about this release.

Reading its system card taught me nothing about AI engineering or development.

There are some interesting observational findings that might scare or excite you depending on your perspective (I feel neutral about them), but they make no real difference to your knowledge or skills.

Here is what the “Training data and process section” says:

Mythos 5 and Fable 5 were trained on a proprietary mix of publicly available information from the internet, public and private datasets, and synthetic data generated by other models.

Throughout the training process, we used several data cleaning and filtering methods, including deduplication and classification.

…

After the pretraining process, the model underwent substantial post-training and fine-tuning, with the goal of making it an assistant whose behavior aligns with the values described in Claude’s constitution.

Claude is multilingual and will typically respond in the same language as the user’s input. Output quality varies by language. The model outputs text only.

This is the most generic text I’ve ever seen around model training. The team basically reveals nothing about how the model was developed.

You could read the following instead to learn about LLMs are trained.

How LLMs are Actually Trained

In another blog post, they vaguely talk about self-improving systems but never share any meaningful algorithmic or systems-level details, unlike other recent open-source research papers.

Instead, they seem more interested in slowing AI progress (except for themselves) and limiting open-source growth.

So I am deciding not to write about this model and its release, even when I might lose the chance to be viral and gain a few thousand subscribers.

‘Into AI’ is, and will always be, about substance over salesmanship.

I would rather write and tell you about something unexciting yet deeply useful than give in to the hype.

I really appreciate you being here and value your time.

Cheers!