Blog posts:

Summer Weeks 6 - 10

2021-5-31

# Week 6 I just moved to Bloomington to work on [secret project] with Luke and Charlie. I only got my computer set up today (Thursday) but now I should have daily work day blogs. We're working mo...

Summer Weeks 1-5

2021-4-26

# Week 1 This is the end of the end of my schooling, and the beginnning of my first summer off in 5 years. What to do? How to keep from drifting into the abyss? ...

Thesis Week 15 (Spring)

2021-4-19

Monday: - 10-12: finished adding related works slides, went outside to practice reading over my slides and streamlining - rest of day: kept editing and practicing slides - Tuesday:

Thesis Week 14 (Spring)

2021-4-12

Monday: () - 1-4: Paperwork and email writing - Tuesday: () - 11-12: more planning, and editing for final copy...

Thesis Week 13 (Spring)

2021-4-5

Monday: () - 10-12: 1 hour reading AI failure material that Dr. Yampolskiy sent me, - 1-2: Analyze the failures for common themes. - Tuesday: ()...

Syntax & Semantics, Plants & Ecosystems

2021-4-5

It's been too long since I've been backpacking. So long that I've aquired a number of perspectives and ideas that I hadn't had in the past that change how I look at the natural world. I finally un...

Thesis Week 12 (Spring)

2021-3-29

Monday: () - 10-12: writing about the two interpretability papers - 1-3: Reading and writing about "The Interpretation Problem" - Tuesday: ()...

Thesis Week 11 (Spring)

2021-3-22

Monday: () - No work today - Tuesday: () - 11-12:...

Thesis Week 10 (Spring)

2021-3-15

# Work log: ...

Thesis Week 9 (Spring)

2021-3-8

# Work log: ...

Thesis Week 8 (Spring)

2021-3-1

It's Monday! Starting on time today. Or maybe a minute late, I had to fix a loose wire in my computer bumping against the GPU's fan. Anyway, what do I need to do? I remember my AI safety section h...

Thesis Week 7 (Spring)

2021-2-22

I'm starting on time today! The first full week of the schedule. It's not perfect because I slept in so I'm going to have to break in a second to eat some toast and make some tea. Today I...

Thesis Week 6 (Spring)

2021-2-18

Despite the lack of blog posts, I have been working this whole time, and sending updates to Dr. Yampolskiy. Sending him weekly updates was redundant with making these updates, but now I'm feeling...

Thesis Update Log Week 3

2020-11-20

### Thesis Advisors Update on the thesis advisors siuation: I have one CSE proffessor on board, and I need a non-CSE prof. I emailed an ECE prof who I really enjoyed his classes, and was quite su...

Thesis Update Log Week 2

2020-11-10

This week I sent out emails to recruit members for the thesis committee (my defense won't be for months but my academic advisor wanted this done ASAP). Being so official and interacting with profe...

Thesis Update Log Week 1

2020-11-2

### Getting started This is the first week of keeping updates on my thesis. I've been working on it since the start of the semester, and now I'm keeping track of my progress each week for Dr. Ya...

Robust First String Matching

2020-7-1

Since writing the last post I've made a lot of new operations on string - reversal, eqality testing, splitting on spaces. I've learned two things: operating on a string without knowing its len...

Robust First String Manipulation

2020-6-19

(read Living Computation Primer first) Representing Strings ...

Living Computation Concepts

2020-6-18

In reading David Ackley's work and thinking about and implementing these ideas, a few concepts have really stuck with me as significant. Bespoke physics for living computation ...

Living Computation (aka Indefinitly Scable Computing) Primer

2020-6-17

(incomple, unedited and mostly incoherent) Background ...

Infinite Grid Computer

2019-10-10

Luke mentioned this idea of ininitely expandable computing, where a computer network is extended to a size where speed of light becomes significant, and you have to design algorithms that run on...

Tarot Card Reading Mnemonics

2019-8-11

Tarot card decks have two kinds of cards: the Minor Arcana, which are almost exactly the same as a regular deck of cards, with cards like "5 of spades" or "ace of cups", and Major Arcana, which mo...

Summer at PNNL

2019-8-2

I spent this summer working for Pacific Northwest National Laboratory. It was a great experience! I worked with lots of people who do deep learning and data science full time, and, even better, th...

Multi-label Classification

2019-5-14

I'm submitting my Optical Illusions Dataseet paper to INSAM journal - horray! However, they have a 3000 word requirement, and the body of my paper is only 2200 or 2000 if you don't count the appen...

Adversarial Transfer to Humans

2019-5-1

I'm currently working on the topic of "Adversarial Examples Transfer from Recurrent Neural Networks to Finite State Automata". Recurrent Neural Networks (RNNs) can be made to misclassify a sequen...

Communications Speech 4

2019-4-8

It needs to be about policy - suggest the audience that a problem exists and that the audience should do something about it. Most of the class seems sympathetic to ecological issues and resource...

Holy Salvation Meme

2019-4-5

Newsgroups: talk.religion.misc,alt.memetics From: ai815@FreeNet.Carleton.CA (Greg Erwin) Subject: This is the Holy Salvation Meme Reply-To: ai815@FreeNet.Carleton.CA (Greg Erwi...

Four Leaf Clovers

2019-4-5

I searched for four leaf clovers yesterday and today, and doing so brought a lot of ideas into my head about intelligence and probability. # Zero shot classification ...

An Incomputable Number

2019-4-3

I heard that the computable numbers are countable, which seems to lead to a paradox using

Artificial Intelligence Safety And Security Independent Study Week 11

2019-3-18

This week I will be doing a comprensive, critical review of open-ended curiousity driven exploration, and other open-ended learning techniques. That's a bit of a vauge subject: AlphaGo could be co...

Artificial Intelligence Safety And Security Independent Study Week 9

2019-3-4

This week I won't be doing any exploratory reading. Instead, I will be studying curiosity rewards in open ended problems, such as Game of Life or Minecraft. I will start by getting a reference imp...

Comm speech 2

2019-3-3

I have to do another speech, so I'm preparing it here before doing an editorial copy to notecards. I interviewed Deborah Denenfeld about her dance career. ...

Artificial Intelligence Safety And Security Independent Study Week 8

2019-2-26

Last week I spent a lot of time experimenting with GPT-2 (well, the tiny pretrained model that OpenAI shared with the public). It's an amazing creative tool, like how surrealist games can help peo...

Fun With GPT-2

2019-2-20

So I downloaded the "small" GPT-2 model. It's still enourmous, and takes 3-4 minutes to run a prediction on my laptop. I need a GPU to speed this up. Copied below are transcipts of my interactions...

Computer Generated Dessert

2019-2-15

A cake recipe based on analysis of GP-2's output! The AI is truly amazing, but its creators refuse to share it. They tempt us with...

Artificial Intelligence Safety And Security Independent Study Week 6

2019-2-11

We're going to focus more on reinfocrcement learning this week, such as inverse reinforcement learning and other variations. ...

Code Katas!

2019-2-8

Luke sent me a link to a page with some "code Katas", which are like Euler problems except focused on software design and extensibility. I figured I might as well try out the first few, even if I...

Artificial Intelligence Safety And Security Independent Study Week 5

2019-2-4

Coninuing the trend on artificial life, I found a quantum evolution experiment and the "Innovation Engines" paper. # Week 4 Wednesday ...

Communications Class Speech 1

2019-2-4

I have to create and deliver several speeches this semester for my communications class. I'll keep my drafting and miscellaneous notes here, and a final copy of the speech at the bottom. This fir...

Column Prose

2019-2-4

I had the idea for a kind of restricted poetry/storytelling, where by following a strict line length limitation, certain columns would have ...

Puzzle Pieces

2019-2-3

I am in a room facing a vast board. Scattered on it are thousands of small colorful bits of cardboard, puzzle pieces. Scattered about are coherent continents, with ragged edges waiting for more pi...

Artificial Intelligence Safety And Security Independent Study Week 4

2019-1-30

We had another meeting this week, and started a trend on open ended exploration by evolution, and artificial life. # Week 4 Monday ...

Artificial Intelligence Safety And Security Independent Study Week 3

2019-1-23

I'm still very interested (curious, even) in curiosity driven AI. I didn't finish "Responses to Catastrophic AGI Risk: a survey", I need to make a high level review of the paper sometime soon.

Response to "Meditations on Moloch"

2019-1-18

A few months ago, I read Meditations on Moloch by Scott Alexander and thought it was the most absolute and true piece of...

Curiosity Maximizer

2019-1-17

Last week, we read Curiosity-driven Exploration by Self-supervised Prediction. Their success in a powerful curiosity reward signal raised a lot of...

Bananagrams AI!

2019-1-16

I just finished the code for my Banangrams playing AI! I've added it under the "fun" tab. Type in all of the letters you have and it returns an html table with a board you can play. Try it out he...

Artificial Intelligence Safety And Security Independent Study Week 2

2019-1-13

Last week was a good start, and our exploration rate was fairly high. This week we are starting with a review of AI safety literature, which will serve as a reference for the rest of the semester....

Artificial Intelligence Safety And Security Independent Study

2019-1-6

This semester at UofL, Luke Miles and myself are working with our professors (Brent Harrison and Roman Yampolskiy, respectively) on an independent study. We will be reading 2-3 papers concerning a...

Capacity for Boredom

2018-12-9

### Disclaimer: this is mostly missspelled ramblings are unfinished. Read at your own diescretion Everything, when repeated enough time, becomes boring. If you listen to enough Motzart, you might...

Emacs Defun Custom Syntax

2018-11-1

In common lisp, it's fairly common to come up with your own variations on defun. For instance, I might make the following macro, which prints out a function's name before calling

Attacks On NN Proxies

2018-10-27

I'm working on a way of controlling arbitrary physical systems using low energy inputs. The idea is to create a NN (neural network) model that emulates the real world system, develop adversarial...

My Paper Got Media Attention!

2018-10-16

I published the optical illusions paper, and got lots of interesting media attention. This is really motivating, I feel a lot of imposter syndrome but really I should feel proud that this went s...

Preprint of Optical Illusions Paper

2018-10-1

With the help of Dr. Yampolskiy, I put the information in the series on optical illusions together into an academic paper, which can be found here: htt...

Data Download and Initial Neural Network Training Results

2018-4-10

I downloaded and cleaned all of the files from Mighty Optical Illusions and ViperLib into JPEG...

Long Term Universe Outcomes

2018-2-14

The Fermi Paradox is a huge problem for humanity. The simple question, "Where are they?" has no easy answer. Are there really no aliens out there colonizing the universe? Maybe life is unlikely, o...

Fun With (Fun With Quines in Python) Quines In Python

2018-2-9

I've seen quines before, as well as quines that aren't really quines, like this dumb bash trick: ` echo $BASH_COMMAND ...

Data Collection Results

2018-2-4

None of the website owners replied to my emails, so I collected the images myself. All of the content on these sites has been collected from other sources, so I don't think there is any issue with...

Optical Illusion Datasets

2018-1-15

I have been digging around for optical illusion datasets. They don't seem to exist, so I will be laying the grounds for the creation of one and starting to collect data. I don't know all that much...

Adversarial Examples for human vision

2018-1-9

Adversarial examples are very revealing about a neural net's inner workings and weaknesses. This wonderful post by open AI discu...