When you fall,

Out of sight,

I slip back in my mind.


And I see a long game to play –

The cruelest of loves,

That’s been found and then displaced.


Though it’s not a way to spend your days –

Tracing out memories of a broken vase.


No it’s not a way to spend your days,

But it’s a longwinded way of saying that,

That I love you.

That I love you.

That I love you.

That I love you.


When you fall,

Out of line,

Just slip back in your mind,

And you’ll hear a song, singing plain –

Proof that love has found a home in this place.


Though it’s not a way to spend your days –

Rebuilding memories of a broken vase.


No, it’s not a way to spend your days,

But if you want it,

Here’s my way of saying that,

That I love you.

That I love you.

That I love you.

That I love you.


Using A.I. to Replicate Artisanship

I’ve been experimenting with ideas in interior design, and though I’m not the finest artist, I have a strong intuition for engineering, and so I’d like to think the net results are pretty good, and certainly commercial quality, at least in my opinion. Specifically, I came up with a sketch this morning for a shower with some clever design components, and I was thinking about how to replicate the design on a mass scale, even though it involves some elements of artisanship, in particular the wood cutting near the drain of the shower, which I’ve assumed has been perforated, and sloped, to fasciliate drainage.


Interior of a shower with wood plank floors.

What I’ve realized is, you could have an artisan make the original model of, e.g., the flooring, while the tools used are being monitored and recorded by a machine, effectively recording what’s taking place, to allow for accurate recording of the angles, pressures, and motions made use of by the artisan. This could allow for the mechanization, and mass production of copies, since a machine could then imitate the actions of the artisan. Obviously, this would require the artisan to make use of specialized tools that have been equipped with sensors to allow for the type of observations necessary to actually implement this, but the point being, that with sufficient data, you could probably imitate an expert artisan carpenter, e.g., and then mass produce the results, by having a machine imitate the actions of the artisan. This would probably require a large number of observations (e.g., contact pressure at a large number of points on a surface), and I’ve already written software that can efficiently manage extremely large numbers of observations. Then, perhaps finishing touches could be applied by human workers, but the point being, that an artisan could make use of A.I. to mass produce their products.

This should cause the finished products to have the same handcrafted appearance as the original, since the machine would be imitating the artisan, and you could even include software to ensure that each finished product is unique, despite being mass produced, using something like the data imitation algorithms that I’ve developed. This would lead to the mass proliferation of artisan-quality products, which would almost certainly improve standards of living, allowing ordinary people to enjoy artisan-quality craftsmanship, at prices associated with mass production. This is arguably what IKEA did for design, delivering high-end design at affordable prices, but extending this to the actual manufacturing of the product, allowing high-quality artisan construction to be enjoyed at prices that reflect the economics of mass production. Below are two related sketches using similar elements, and all are assumed to be part of the same house.


A performance space with glass walls, a rug, and a cabinet.


A living space with glass walls, a hanging art piece made of subway tiles with a faux-grass frame, a floor rug intended for seating, and a floor with a partially exposed base filled with large stones.


A wooden chair with shock-mounted panels to allow for contour.

This is the color palette I imagined for the space, generally.

Screen Shot 2020-04-27 at 3.16.18 PM


Prometheus A.I. Physics Engine

It’s been a few weeks since I’ve shared any new software, which for me, is a long time, and though the disruption of the coronavirus obviously had some hand in it, the real cause is that I’ve been working on thermodynamics, coming up with new ideas and writing software that extends the algorithms and ideas that I introduced in a previous note. Specifically, I’ve actually implemented the idea of making use of compressed representations of physical systems, and have managed to reduced complex sets of observations to linear Euclidean embeddings, which allows my core machine learning algorithms to be applied.

In summary, these algorithms takes tens of millions of observations, and reduce them to reasonably short sequences of real numbers, that can then be classified using my core A.I. algorithms, which are of course radically efficient. This in turn allows for the radically efficient analysis of thermodynamic systems, and other comparably complex systems, perhaps biological systems, or actual economies (as opposed to just a single asset price).

I’m very excited about this work, so I’m going to spend a few more days ensuring that I’ve optimized the algorithms, and written an article that is deserving of the quality of work that I’ve done on the matter.