3:00 Dentist

JuryRoom

• Finished doing Priyank’s analysis

MDS

• Lots more math!
• Working on parametric solutions to intersecting lines

GOES

• 11:00 All hands
• 2:00 Meeting with Vadim. Nope, now tomorrow at 10:00

VoterFraud2020

• We are making publicly available VoterFraud2020, a multi-modal Twitter dataset with 7.6M tweets and 25.6M retweets from 2.6M users that includes key phrases and hashtags related to voter fraud claims between October 23rd and December 16th. The dataset also includes the full set of links and YouTube videos shared in these tweets, with data about their spread in different Twitter sub-communities. Key takeaways from our initial analysis of the data are listed below.
• The methodology for our data collection including the tracked keywords is detailed in the paper, “VoterFraud2020: a Multi-modal Dataset of Election Fraud Claims on Twitter”, available from arXiv.

Book

• Continuing with Diversity injection – done!
• 2:00 Meeting with Michelle

MDS

• Found another post on solving ballistics, with worked examples:
• Going to try this with numpy.roots, which work exactly the way you would want them to do.
• One shortcut did occur to me. I just need to determine when velocity is zero and use 2t to figure out what the range will be. I can then use that knowledge to place the source.
• A little too cooked to work through the example

JuryRoom

• Priyank’s task

GPT agents

• Install Ecco and try the intro (here in the video, also ranked choices, and token probability) locally, then message him to see if he’d like to discuss maps
• Need to put together the embedding files. It’s working! I need to clean the text and play with queries – done!
• 3:30 Meeting – nope, it was yesterday

ML Group

• Add Ecco to presentation list – done

Donate to MD food bank today

More work from Jay Alammar and language model saliency

• Talk on YouTube
• He’s put together a website: Eccox.ix

Book

• Continuing with Diversity injection

NESDIS

• Training – done!

MDS

• Adding ballistics (from NASA, at a nice intro level)
• Working nicely:

GPT agents

• Finish watching Alammar video (done), install and try the intro locally, then message him to see if he’d like to discuss maps
• Need to put together the embedding files. It’s working! I need to clean the text and play with queries – done!

• There does seem to be a bug with getting the search to work

Inauguration day!

Book

• Not much time to work on it this morning, but added some text about Stephens’ neural coupling. Tomorrow I’ll talk about the agents a bit and in particular how increasing dimensions makes it harder to have a stampede

GPT Agents

• Antonio submitted the paper!
• Need to generate embedding files for the Tensorflow embedding projector

MDS

• Started viewer. It’s sooo cute!

GOES

• Working on the interpolation begin/end logic – done!

Feeling like the inauguration will go smoothly, but holding my breath anyway

Fixing disinformation won’t save us – Ethan Zuckerman

• There have been countless fact-checking and other efforts designed to rid social media of misinformation. They’re not going to work until the party and the major ideological amplifiers start explicitly renouncing these points of view. The signs are not good – while Fox News was willing to declare that Joe Biden had won the election, they are still providing platforms for people denying the facts of the victory. And a majority of Republican representatives voted to overturn a democratic election. Until there are consequences for perpetuating those falsehoods, don’t count on changes to the media to solve this problem

The end of the Trump-Fox feedback loop

• Twitter’s January 8 decision to permanently suspend Trump’s account closed a rare window into a president’s mindset and policymaking that we are unlikely to ever see again. For the past four years, I documented the sources of the president’s grievances and obsessions, matching Trump’s tweets to the television segments he was watching. The president’s TV addiction inspired at least 1,375 tweets dating back to September 1, 2018. The vast majority came in response to his favorite programs on the pro-Trump Fox News and Fox Business networks. 

But if there ever was a coda for the Trump years, this has got to be it:

• In this article, we will focus on the hidden state as it evolves from model layer to the next. By looking at the hidden states produced by every transformer decoder block, we aim to gleam information about how a language model arrived at a specific output token. This method is explored by Voita et al.[1]. Nostalgebraist [2] presents compelling visual treatments showcasing the evolution of token rankings, logit scores, and softmax probabilities for the evolving hidden state through the various layers of the model.
• [1]: The Bottom-up Evolution of Representations in the Transformer: A Study with Machine Translation and Language Modeling Objectives
• [2]: interpreting GPT: the logit lens

Book

• Start on diversity injection section
• Research note: Examining false beliefs about voter fraud in the wake of the 2020 Presidential Election
  • The 2020 U.S. Presidential Election saw an unprecedented number of false claims alleging election fraud and arguing that Donald Trump was the actual winner of the election. Here we report a survey exploring belief in these false claims that was conducted three days after Biden was declared the winner. We find that a majority of Trump voters in our sample – particularly those who were more politically knowledgeable and more closely following election news – falsely believed that election fraud was widespread, and that Trump won the election. Thus, false beliefs about the election are not merely a fringe phenomenon. We also find that Trump conceding or losing his legal challenges would likely lead a majority of Trump voters to accept Biden’s victory as legitimate, although 40% said they would continue to view Biden as illegitimate regardless. Finally, we found that levels of partisan spite and endorsement of violence were equivalent between Trump and Biden voters.

MDS

• Meeting with Aaron today to discuss nest steps and how to combine with his project?
• Still need to be able to access the VPN – more paperwork. Wheee!

GOES

• Continue with the new TopController
• Reading in and stepping through the script. Now I need to slew through the points and return a done when the l2 dist is within a threshold

GPT Agents

• Fit More and Train Faster With ZeRO via DeepSpeed and FairScale
  • If you use the Hugging Face Trainer, as of transformers v4.2.0 you have the experimental support for DeepSpeed’s and FairScale’s ZeRO features. The new --sharded_ddp and --deepspeed command line Trainer arguments provide FairScale and DeepSpeed integration respectively. Here is the full documentation.
• Send data – done!
• 3:30 Meeting
• Finish paper and submit?
• Need to export embeddings to tensorboard embedding viewer. Here’s how: tensorflow.org/tensorboard/tensorboard_projector_plugin

My Former Hasidic Community Still Supports Trump. I’m Not Surprised.

• The thing is, contemporary Hasidic sects are designed for authoritarian control. Each Hasidic sect, from Bobov to Viznitz to Satmar to Skver, are run by what is called a “grand rabbi.” These rabbis are demanding patriarchs. They expect women to wear particular shades of stockings, men to dress identically, congregants to receive their blessings before making any personal life decisions, and they believe in a world where Hasids are the only Jews worth mentioning. Most importantly, Hasidic grand rabbis center their congregants’ worlds around themselves. They are populist leaders of miniature nations. Congregants have paintings and photographs of grand rabbis around their homes, sacrifice family time for tisches (Friday night gatherings) with their leaders, and would do anything to protect the power of their particular grand rabbi.

Book

• Working on Making better Human-Computer Interfaces for Populations. Finished my first pass at The signature of dangerous misinformation section
• 2:00 Meeting with Michelle

GOES

• Decided to build out a sandbox ScriptReaderScratch RCS controller to work out the file loading and playback. Rather than AngleController, I’ll have a method that interpolates to the newest target. That should be enough to let me work out the details without breaking anything

GPT Agents

• Start pulling off pages from paper
• 3:30 meeting

Today’s webpage from the Washington Post:

Book

• Working on Making better Human-Computer Interfaces for Populations

GOES

• 10:00 meeting with Vadim, then start working on TopController and AngleController
• Have AngleController return DONE when it’s sufficiently close to its goal
• Have TopController load the script, send a command to AngleController to get to the start point of the manouver
• Upon DONE from AngleController, run the script at the specified speed. While the script is running, just re-issue the TO_ANGLE command.
• Once the script is done, wait for the AngleController to reach the goal
• Another option is to leave the AngleController running and handle the logic in TopController. Need to think about that.
• Customer meeting at 2:00
• Submitted spreadsheet to Eric about bandwidth needs

ML meeting at 3:30 – wound up being an evening of financial and cycling advice!

2021 continues to produce surprises: Several senior Republicans join impeachment push

I was looking at a McSweeny’s article (LEST WE FORGET THE HORRORS: A CATALOG OF TRUMP’S WORST CRUELTIES, COLLUSIONS, CORRUPTIONS, AND CRIMES). It references a pile of Trump Tweets that now look like this:

History has been deplatformed. Now what?

Good article on QAnon: QAnon reshaped Trump’s party and radicalized believers. The Capitol siege may just be the start.

Speaking of Twitter, this is a good thread on how to write and contest ML conference papers

ParametricUMAP allows users to train a neural network to optimize the embedding, resulting in a direct neural net based mapping from source data to embedding. This allows for extremely fast inference (embedding of new data points), orders of magnitude faster than standard UMAP. It also provides facilities for an inverse transform, mapping from the embedding space to the original data space that is both far faster and more robust that that provided by standard UMAP. Since network architectures can be user provided this also allows for CNN and RNN based UMAP embeddings for images or sequences.

Book

• Continue Making better Human-Computer Interfaces for Populations

GOES

• Add in mapping to script reader, verify by adding legends

MDS

• Status meeting maybe produce a spreadsheet to walk through that shows a time series of inputs and a calculation for each set? I think the inputs can be a column of six (for now?) variables as a set of rows, and the prediction calculations are shown below that. Make a DataFrame and see what that looks like.

From Charlottesville to the Capitol: how rightwing impunity fueled the pro-Trump mob

• The playbook for the Maga invasion of the nation’s Capitol building on Wednesday has been developing for years in plain sight, at far-right rallies in cities like Charlottesville, Berkeley and Portland, and then, in the past year, at state capitols across the country, where heavily armed white protesters have forced their way into legislative chambers to accuse politicians of tyranny and treason.

Here’s what seems to have happened with the Parler hack. The data may be available for research

• The person who did this: @donk_enby
• Gizmodo: Every Deleted Parler Post, Many With Users’ Location Data, Has Been Archived
• Cybernews: 70TB of Parler users’ messages, videos, and posts leaked by security researchers
• Repo on archive.org
• Another repo of videos (also plotted on a map based on geocodes) here: www.tommycarstensen.com/terrorism

Nice paper on training a model to generate synthetic data for better classification training: Reducing AI bias with Synthetic data. It uses the gretel’s gretel-synthetics library It’s free to use during the beta period, not sure about after, or what the pricing will be. They are hiring, with about seven openings at the moment, so they are burning through someone’s money.

GPT Agents

• Finish abstract submission – done
• Make an Overleaf project for qualitative paper?

GOES

• Finish up the ManeuverReader – done! Here’s the original, with some large number of points that is subsampled to 100 points and stored as a json file

• Here’s a reconstructed version that uses 1/3 (33) steps through the file. You can see a little roughness, but with more points it’s indistinguishable from the original pulled off influxDB:

• And here’s a snippet of the json file

{
    "title": "test",
    "speed_multiple": 1.0,
    "read_fmt": "%H:%M:%S",
    "mapping": {
        "GNC_AC_MOM_GEN_MOMBODY_X": "pitch",
        "GNC_AC_MOM_GEN_MOMBODY_Y": "roll",
        "GNC_AC_MOM_GEN_MOMBODY_Z": "yaw"
    },
    "duration": "00:29:56",
    "samples": [
        {
            "timestamp": "00:00:00",
            "values": [
                {
                    "name": "GNC_AC_MOM_GEN_MOMBODY_X",
                    "value": 63.68628693
                },
                {
                    "name": "GNC_AC_MOM_GEN_MOMBODY_Y",
                    "value": 1.657353401
                },
                {
                    "name": "GNC_AC_MOM_GEN_MOMBODY_Z",
                    "value": -3.304497004
                }
            ]
        },

• start to integrate into TopController

Book – Not much, just jotting down notes

GPT Agents

• Working on submitting

GOES

• Working on script generator

MDS

• Trying to find the right charge number
• Made slide deck for todays meeting and presented overview and next steps

This is one hell of a thread:

GPT Agents

• Moved the synthetic agents paper to the ijcai format and shared the Overleaf version with Antonio

Also:

GOES

• Work on script generator and reader

Book

• Working on Hierarchies, Networks, and Technology. New technologies may have the same arc as writing and printing, which is initial hierarchy that produces influence networks that counter (to a degree), the more aggressive aspects of a dominance hierarchy
• Meeting with Michelle

MDS

• Discussion with Aaron about phase2
• Wrote up thoughts and sent to Clay

And just so we remember that the pandemic is not going well here. For comparison, the battle that took the most American lives was Antietam, where there were 3,675 fatalities if you count both sides.

Need to look into replacing JetBrains

GOES

• Slide deck for 2:00 meeting
• 11:00 AI-ML meeting
• 2:00 Sim discussion. We have until the end of March to come up with a compelling demo
• More script generator. I need to write a method that searches through a Measurement list looking for the last value before a datetime
  • Need to map the names in the database to the desired name for the sim

MDS

• Write up notes from meeting and distribute
• Write up a couple of paragraphs for Clay

GPT Agents

• More coding
• ML group meeting

Book

• More Hierarchies, Networks, and Technology

Georgia is looking promising! Maryland is trying to be more flexible in its vaccinations!

And then later in the afternoon, this happened:

MDS

• 10:00 meeting with Aaron and Peter
  • Create pipeline for data (what does it look like?) and FOM evaluation (input/output)
  • Who creates the sim and generates the data?
  • Who feeds that into the FOM?
  • Can all this run locally?
  • Write up notes
• Meeting with Clay
  • Write up two paragraphs on phase 2 thoughts

GOES

• Slide deck for tomorrow’s meeting
• More work on script generator

GPT Agents

• Created a local version of the IJCAI paper project. Need to fit the format and then create the Overleaf project and share with Antonio
• Do some coding, dammit

JuryRoom

• 5:00 Huri Whakatau meeting

Book

• Continue Hierarchies, Networks, and Technology
• Found Charting the “Rise of the West”: Manuscripts and Printed Books in Europe, A Long-Term Perspective from the Sixth through Eighteenth Centuries, which is very cool.

Voting in Georgia today. I am pessimistic but hopeful about the outcome

GPT Agents

• I’m not sure if the meeting is today at 3:30 or Friday at 4:00?
  • It was today. Continuing on trying to figure out the best way to understand the behavior of the model. One of the interesting findings for today was that if the data isn’t in the dataset, then the model will start generating tokes at the meta wrapper.
• More coding

Book

• Working on what’s become Hierarchies, Networks, and Technology, and I think I’m now happy with where it’s going. It makes sense to use as the end of the chapter as well
• Made a cool figure:

GOES

• The Lambda box was cancelled. Sigh
• 11:00 Meeting with Vadim
• I’m going to start on a script-reading capability for TopController. I think a JSON or XML file that contains the following elements:
  • Absolute or relative move
  • axis name
  • Target (HPR or XYZ)
  • Timestamp
  • Required accuracy
• So a move could be a series of HPR coordinates that ‘play’. The first step is a MOVE command which includes the filename. The TopController opens the file (or fails and reports it), loads the move into memory and begins to step through it based on the timestamp. On reaching the end of the file and when the AngleController reports success/failure, the TopController reports DONE and is ready for the next MOVE
• Downloading the yaw flip maneuver from influx: