aryehof

> … hard for me to understand Google's redefinition of "world models"

I’ve also wondered what they mean by “world model”, whereas I interpret them as being about predicting outcomes - something an LLM cannot reliably do.

Is this new model an LLM or a predictive outcome model (as you say a la Yann Lecun), or some hybrid?

You can’t control or enable your children …

You can, however, be a good role-model.

> Code review does two jobs: 1. Find and fix issues before they ship 2. Build shared context and codebase understanding across the team

So code review has become newspeak for “testing” and “education”?

Well not really testing. We the users are the testers and it falls to lucky us to really discover and find the errors, edge cases and security issues. The consequence of the rush to continually release…

Is the AI advance in the field of software testing nothing more than “now an AI agent can do a code review?”.

> Computer programming is, fundamentally, about two things: 1. Problem-solving using computers 2. Learning to control complexity while solving these problems

Pretty sure your forgetting more than half of computing…

… 3. The modeling of external systems into code.

Modeling isn't solving a problem, it’s representing an external systems with a degree of fidelity. That’s different from solving a problem through transformation, albeit far harder to teach.

Most software that isn't trivial is a big ball of mud. AI is trained on that and produces the same (or worse) - faster.

Are you following more of an “aider” type approach, or something different?

When I was a child, schools didn’t need to communicate on a daily, weekly (or even monthly) basis with parents. How times change. Heck, we went for a whole school day without having to talk to Mummy or Daddy. So this agent is technology to help the mess enabled by technology?

As a zero-shot classifier, I expect that effectiveness is dependent on the data trained upon.

Jev input … > Unstructured data (e.g. text) with an emphasis on structured program state.

What pre-training data/model is Jev based on? Surely result effectiveness is dependent (outside of one’s own input as “state”) on that?

> Without alignment, further improvements in capability turn LLMs into wanton felony generators

Isn’t “full” alignment and guardrails a task that can never be generally achieved? One will always discover and realize the need for new guidelines and guardrails? What about vague, incomplete, inconsistent nuances and generalizations - that make this a losing battle?

I suggest ultimate safeguards need to be outside the models?

Are the Silo shots too dark to see, like when attempting to watch it via an Apple sub?

On reading the title, the first thing that came to mind was “a good book”.

Albeit, that I view many of the things in the list as “ideals” or “values”, rather than “luxuries”. The greatest thing one can aspire to in terms of an ideal, is to have “peace of mind” in my opinion.

So true.

> I am kind of amazed how negative the comments are here, especially on HN.

Sadly, I attribute this to a decline in having “good will” towards others, and an increase in cynicism, prejudice, and meanness.