Comments URL: https://news.ycombinator.com/item?id=46444398

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=41725731

Points: 2

# Comments: 0

Reminds me of this great quote by Terrence Deacon on a podcast (I believe it was one of his Mind & Matter episodes). From memory:

> Engineering is in some sense the opposite of life: Engineering involves assembling components. Life involves differentiating wholes.

I highly recommend Deacon's 'Incomplete Nature', and I'm very psyched about his upcoming book “Falling Up: Inverse Darwinism and Life’s Complexity Ratchet”

Enact's core features include the ability to rewind and replay executions of python functions that are annotated in the framework. The linked notebook shows how this can be used to implement a generic Monte Carlo Tree Search (MCTS) algorithm that can be deployed over any annotated python program and directly searches the space of program executions.

MCTS (used, e.g., in the AlphaZero family of algorithms) searches a tree of generative possibilities using a scoring function in order to find high-scoring branches. The notebook deploys MCTS over a line-by-line poetry generator, allowing the algorithm to backtrack and search different completions for the same starting lines.

Baseline (no MCTS):

> While shadows stir and breathe beneath the moon's licentious glow.

> Nightly whispers wind through trees, long lost in their verdurous flow.

> In the echoing silence, secrets lie dormant, awaiting the raven's crow.

> A path unfolds beneath the thorns, where memories and shadows grow.

MCTS Example 1:

>Under the hush of leaves, ancient secrets pass as whispers.

>Shadows dance as the daylight fades, echoing the rhythm of heartbeats.

>In hallowed silence roots entwine, sharing stories with the undergrowth.

>A shiver of twilight laces through, the phantom pulse of a forest's breath.

MCTS Example 2:

> Under the neon hum, craft recedes into shadow,

> Where artistry and mechanics clash, souls are traded for steam and gas.

> Echoes of creation's heart, usurped by iron’s brazen art.

> Infernal gears grind dreams, artists' tears oiling their ceaseless horseplay.

MCTS Example 3:

> Spun by silent looms of determination ragged from use.

> Eyes search the horizon, fingers till the marrow of invention.

> Under quiet cover of the moon, the night was sown with ceaseless creation.

> Mastering chaos, coaxing stars from a coal-darkened hand.

Comments URL: https://news.ycombinator.com/item?id=37585517

Points: 1

# Comments: 0

It's still early, but the core insight is that a lot of these generative AI flows (whether text, image, single models, model chains, etc) will need to be fit via some form of feedback signal, so it makes sense to build some fundamental infrastructure to support that. One of the early demos (not currently live, but I plan on bringing it back soon) was precisely the type of flow you're talking about, although we used 'prompt refinement' as a cheap proxy for tuning the actual model weights.

Roughly, we aim to build out core python-level infra that makes it easy to write flows in mostly native python and then allows you track executions of your generative flows, including executions of 'human components' such as raters. We also support time travel / rewind / replay, automatic gradio UIs, fastAPI (the latter two very experimental atm).

Medium term we want to make it easy to take any generative flow, wrap it in a 'human rating' flow, auto-deploy as an API or gradio UI and then fit using a number of techniques, e.g., RLHF, finetuning, A/B testing of generative subcomponents, etc, so stay tuned.

At the moment, we're focused on getting the 'bones' right, but between the quickstart (https://github.com/agentic-ai/enact/blob/main/examples/quick...) and our readme (https://github.com/agentic-ai/enact/tree/main#why-enact) you get a decent idea of where we're headed.

We're looking for people to kick the tires / contribute, so if this sounds interesting, please check it out.