Robert Miles discusses recent developments in AI Safety

Across the dunes, in the waning light, The rising moon pours her amber rays, Through the slumbrous air of the dim, brown night The pungent smell of the seaweed strays— From vast and trackless spaces Where wind and water meet, White flowers, that rise from the sleepless deep, Come drifting to my feet. They flutter the shore in a drowsy tune, Unfurl their bloom to the lightlorn sky, Allow a caress to the rising moon, Then fall to slumber, and fade, and die.

White flowers, a-bloom on the vagrant deep, Like dreams of love, rising out of sleep, You are the songs, I dreamt but never sung, Pale hopes my thoughts alone have known, Vain words ne’er uttered, though on the tongue, That winds to the sibilant seas have blown. In you, I see the everlasting drift of years That will endure all sorrows, smiles and tears; For when the bell of time will ring the doom To all the follies of the human race, You still will rise in fugitive bloom And garland the shores of ruined space.

The prompt was "Trees." If you know how to format line breaks instead of paragraph breaks, please let me know.

Timber

I have heard that wood will warm you many times,

When you chop it,

Split it,

Burn it and cook-

I find a standing dead.

It's no good taking the fallen,

The wet gets in so quick around here.

Leave those for the beetles.

Abraham Lincoln said

If he had six hours to fell a tree,

He would spend four

Sharpening his axe.

My father once asked me

"Why not use a chainsaw?"

I could let another man

fuck my wife.

Sweat slick and

Sore muscles

Never felt so good.

Life shorn of its artifice.

I wrote a comment recently in response to a senator describing himself as "pro-choice" in defense of his refusing to repeal child marriage statutes. I recognized a poetic rhythm and have adapted it to hopefully fit as a submission here.

Pro-Choice

Yeah, he's pro choice-

Pro choice like a gunman asking, "which kneecap?"

Pro choice like a lioness stalking a herd of gazelle.

Pro choice like a homeless man can choose a bridge.

Pro choice like deciding between financial and medical ruin.

Pro choice like a rapist asking, "ass or cunt?"

Pro choice like choosing which bill to put off this month.

Pro choice like a backalley crackhead choosing another hit.

Pro choice like forcing a mother to choose between an impossible baby and an illegal abortion.

Yeah. He's pro choice.

>Maybe gravity is no more fundamental than the force of a stretched elastic band. Maybe gravity is just an entropic byproduct—an emergent effect of the universe’s tendency to disorder.

>Abstract: > >Researchers have developed artificially intelligent (AI) and expert systems (ES) to assist in the formulation, solution and interpretation of generic mathematical programs (MP). In addition, researchers also have built domain-specific systems either modeled around a mathematical program or which include a mathematical program module. In these systems, the specificity of the domain allows researchers to extend the interpretation or formulation beyond that available from the generic set of assumptions about mathematical programming. Further, researchers have begun to investigate the use of mathematical program formulations of expert systems. The purpose of their research has been to, e.g., understand the complexity of the expert systems and also to examine the feasibility of mathematical programming as an alternative solution methodology for those expert systems. This paper surveys and extends some of that literature that integrates AIlES and MP, and elicits some of the current research issues of concern.

>Maybe AI systems would be safer if they avoid gaining too much control over their environment? How might that work?

>This video discusses the second stage of the machine learning process: (2) collecting and curating training data to inform the model. There are opportunities to incorporate physics into this stage of the process, such as data augmentation to incorporate known symmetries.

I believe I've found a bug. I intermittently go back multiple pages when pressing back button. It frequently happens on youtube, and occasionally happens elsewhere.

In the picture, find my history where I started with the google search for "the chain" (please forgive me for having history from multiple tabs all mixed up). From here I went back and forth with the expected behavior a couple of times looking for the version I remembered. When I found the correct version, I settled in to listen to it in the background for a few songs minutes until it switched to gypsy which wasn't what I wanted to hear. Naturally enough, I went back once- straight from gypsy to the google search, skipping rhiannon and dreams and the chain. This is reflected on the rightclick drop down from the forward button- observe the three intervening pages which are in the history are not in the forward/back list.

In general, I frequently notice this unexpected and unwanted behavior when I go to youtube. I often go to the homepage, select a video, and then try to back to the homepage only to find it take me back to the blank tab, and when I try to forward it takes me to the video without the intervening homepage.

I seem to recall times when I experienced similar behavior, but cannot put a pin in when or where and cannot at this time, despite trying, reproduce it anywhere but youtube. I haven't tried downloading chrome or shudder using edge, mostly because fuck that, but also slightly because I am very lazy.

I also didn't notice this behavior on my laptop which had win10 and firefox (that is, before I switched to linux as a trial over there, but I digress). I tried to search for this error, but google search is terribly corrupt and difficult to get useful results beyond eg song names. I also searched this forum for "back pages" and "back page" and didn't find anything relevant going back a year.

Do I set my computer on fire for its disobedience, or...?

Abstract:

>Rapid progress in machine learning and artificial intelligence (AI) has brought increasing attention to the potential impacts of AI technologies on society. In this paper we discuss one such potential impact: the problem of accidents in machine learning systems, defined as unintended and harmful behavior that may emerge from poor design of real-world AI systems. We present a list of five practical research problems related to accident risk, categorized according to whether the problem originates from having the wrong objective function ("avoiding side effects" and "avoiding reward hacking"), an objective function that is too expensive to evaluate frequently ("scalable supervision"), or undesirable behavior during the learning process ("safe exploration" and "distributional shift"). We review previous work in these areas as well as suggesting research directions with a focus on relevance to cutting-edge AI systems. Finally, we consider the high-level question of how to think most productively about the safety of forward-looking applications of AI.

>We can expect AI systems to accidentally create serious negative side effects - how can we avoid that? >The first of several videos about the paper "Concrete Problems in AI Safety".

A brief overview of the concept of generality in AI systems.

Abstract

For stable and efficient fusion energy production using a tokamak reactor, it is essential to maintain a high-pressure hydrogenic plasma without plasma disruption. Therefore, it is necessary to actively control the tokamak based on the observed plasma state, to manoeuvre high-pressure plasma while avoiding tearing instability, the leading cause of disruptions. This presents an obstacle-avoidance problem for which artificial intelligence based on reinforcement learning has recently shown remarkable performance. However, the obstacle here, the tearing instability, is difficult to forecast and is highly prone to terminating plasma operations, especially in the ITER baseline scenario. Previously, we developed a multimodal dynamic model that estimates the likelihood of future tearing instability based on signals from multiple diagnostics and actuators. Here we harness this dynamic model as a training environment for reinforcement-learning artificial intelligence, facilitating automated instability prevention. We demonstrate artificial intelligence control to lower the possibility of disruptive tearing instabilities in DIII-D, the largest magnetic fusion facility in the United States. The controller maintained the tearing likelihood under a given threshold, even under relatively unfavourable conditions of low safety factor and low torque. In particular, it allowed the plasma to actively track the stable path within the time-varying operational space while maintaining H-mode performance, which was challenging with traditional preprogrammed control. This controller paves the path to developing stable high-performance operational scenarios for future use in ITER.

> This video discusses the first stage of the machine learning process [identified in the previous post]: (1) formulating a problem to model. There are lots of opportunities to incorporate physics into this process, and learn new physics by applying ML to the right problem.

>This video describes how to incorporate physics into the machine learning process. The process of machine learning is broken down into five stages: (1) formulating a problem to model, (2) collecting and curating training data to inform the model, (3) choosing an architecture with which to represent the model, (4) designing a loss function to assess the performance of the model, and (5) selecting and implementing an optimization algorithm to train the model. At each stage, we discuss how prior physical knowledge may be embedding into the process.

>Physics informed machine learning is critical for many engineering applications, since many engineering systems are governed by physics and involve safety critical components. It also makes it possible to learn more from sparse and noisy data sets.

>Abstract

There is no proof that black holes contain singularities when they are generated by real physical bodies. Roger Penrose claimed sixty years ago that trapped surfaces inevitably lead to light rays of finite affine length (FALL’s). Penrose and Stephen Hawking then asserted that these must end in actual singularities. When they could not prove this they decreed it to be self evident. It is shown that there are counterexamples through every point in the Kerr metric. These are asymptotic to at least one event horizon and do not end in singularities.

>It's not too often that a giant of physics threatens to overturn an idea held to be self-evident by generations of physicists. Well, that may be the fate of the famous Penrose Singularity Theorem if we're to believe a recent paper by Roy Kerr. Long story short, the terrible singularity at the heart of the black hole may be no more.

>3D prints still suffer from bad layer adhesion due to their 2.5D slicing and printing approach. I investigated if a novel slicing method that interleaves the layer could improve the strength of 3D prints.

> > > 3D prints still suffer from bad layer adhesion due to their 2.5D slicing and printing approach. I investigated if a novel slicing method that interleaves the layer could improve the strength of 3D prints. > >

Abstract

We argue that the striking similarity between the cosmic abundances of baryons and dark matter, despite their very different astrophysical behavior, strongly motivates the scenario in which dark matter resides within a rich dark sector parallel in structure to that of the standard model. The near cosmic coincidence is then explained by an approximate ℤ2 exchange symmetry between the two sectors, where dark matter consists of stable dark neutrons, with matter and dark matter asymmetries arising via parallel WIMP baryogenesis mechanisms. Taking a top-down perspective, we point out that an adequate ℤ2 symmetry necessitates solving the electroweak hierarchy problem in each sector, without our committing to a specific implementation. A higher-dimensional realization in the far UV is presented, in which the hierarchical couplings of the two sectors and the requisite ℤ2-breaking structure arise naturally from extra-dimensional localization and gauge symmetries. We trace the cosmic history, paying attention to potential pitfalls not fully considered in previous literature. Residual ℤ2-breaking can very plausibly give rise to the asymmetric reheating of the two sectors, needed to keep the cosmological abundance of relativistic dark particles below tight bounds. We show that, despite the need to keep inter-sector couplings highly suppressed after asymmetric reheating, there can naturally be order-one couplings mediated by TeV scale particles which can allow experimental probes of the dark sector at high energy colliders. Massive mediators can also induce dark matter direct detection signals, but likely at or below the neutrino floor.