Summer solstice!

Tix and hotel – verify airport shuttle – done

SBIRs

• Sprint review yesterday
• MDA meeting. Spent most of the time with Chris I. going over the file. Turns out there is a manual, so I need to read that
• Sprint planning
  • Research monolithic model to compare with RCSNN
  • Continue with RCSNN tool. Start eval?
  • MDA mgmt
  • Support simaccel
  • Set up acct for remote dev on the new server
  • Support Rukan

GPT-Agents

• Contact Nabil Galini to set up a chat
• Add option to clamp to sample size rather than always downloading all the smallest sample
• Make sure to pull in the full day – done

Book

• Pull out extra detail from overview – done!
• Reword “examine” in chapter detail – done!
• Submit! -Done

Safer is a supertanker in advanced state of decay that will break apart or explode if the world does not act. The result will be an environmental and humanitarian catastrophe centered on the coast of a country already devastated by seven years of war and affecting the entire region. The UN is ready to stage an emergency operation to address this threat, but work will only begin when we have the necessary funds.

TOKEN is a MASK: Few-shot Named Entity Recognition with Pre-trained Language Models

• Transferring knowledge from one domain to another is of practical importance for many tasks in natural language processing, especially when the amount of available data in the target domain is limited. In this work, we propose a novel few-shot approach to domain adaptation in the context of Named Entity Recognition (NER). We propose a two-step approach consisting of a variable base module and a template module that leverages the knowledge captured in pre-trained language models with the help of simple descriptive patterns. Our approach is simple yet versatile and can be applied in few-shot and zero-shot settings. Evaluating our lightweight approach across a number of different datasets shows that it can boost the performance of state-of-the-art baselines by 2-5% F1-score.

Book

• Finish proposal? Yes!
• Gita Manaktala (Information Science and Communication editor) oversees the MIT Press’s book acquisitions and works closely with our other editors. She acquires her own list of books in the areas of information science, communication, and internet studies. Her interests include networked communication, news and information, privacy, data security, and access to knowledge.
• Katie Helke | Editor: I acquire trade books, professional books, crossover books, and (very occasionally) textbooks. Head here if you’d like to learn more about those different book types and some other random publishing stuff that may or may not be useful to you. Head here if you’d like to learn more about the MIT Press, its history, and some of its current initiatives.

GPT-Agents

• Continue working on balanced pull. I think I finally got the math right

SBIRs

• Demo slides

Book

• Working on Stripe proposal

GPT Agents

• Sent a note on belief maps to Dr. Deb Roy and Dr. Nabeel Gillani about their Wired Article
• Working more on the Twitter downloader

SBIRs

• 9:15 sprint review
• 2:00 Meeting with Ron about managing the Lambda server
• 2:30 Meeting with Rukan about model behavior

Overview and key findings of the 2022 Digital News Report

• A clear throughline in this year’s report is the changing habits of younger groups, specifically those under 30, whom news organisations often struggle to reach. Throughout this Executive Summary, and in a separate chapter, we find that this group that has grown up with social media is not just different but more different than they were in the past. We also explore their use of newer visual networks for news such as TikTok and Instagram, with support from a detailed qualitative study in three countries (UK, US, and Brazil).

GPT Agents

• Met with Shimei and Jimmy.
  • added ” OR ” to the input list to handle things like “chinavirus OR china virus”
  • Still working towards initial corpus creation. Need to store the queries in the db too
  • No meeting next week

Book

• Working on Stripe proposal. Set up the root tex file and reworked the OUP parts. I need to do a first person author bio and a lite version of the comparables
• Need to write a marketing plan

SBIRs

• Work on RCSNN App
  • Filtering – done
  • Color coding dictionary entries by type – got colors working. It is not obvious!

    def color_text(self, target:str, c:str='red'):
        self.tk_text.tag_remove(target, '1.0', tk.END)

        start_pos = self.tk_text.search(target, '1.0', stopindex=tk.END)
        spl = start_pos.split('.')
        index = int(spl[1])
        end_pos = "{}.{}".format(spl[0], index+len(target))
        print("{} pos = {}, end = {}".format(target, start_pos, end_pos))
        self.tk_text.tag_add(target, start_pos, end_pos)
        self.tk_text.tag_config(target, foreground=c)

Did travel insurance!

Book

• Backed up to svn
• Sent letters to Oxford, Cambridge, and Stripe
  • Got a response back and template from Stripe! (marketing plan)
• If that doesn’t work, I think it’s time to hire an editor

GPT Agents

• Still working on collecting balanced data. I think the trick will be to look for the lowest number of tweets per day starting at the first day of collection and work forward, collecting that many tweets from each keyword, then repeat
• For unbalanced, just make the one request and go forward in time until the corpora size is reached?
• 3:30 Meeting

SBIRs

• 9:15 standup. Need to get reacquainted with the RCSNN codebase and tool
• 1:00 server status meeting
• 1:30 resync meeting

Book

• Finished? The proposal rewrite

SBIRs

• More on FMDS. Need to write an abstract
• MDA Meeting at 2:00

GPT Agents

• Calculate rough tweet rate per keyword. Actually used the count interface and that works well.

The town crier

• Six years into the grass-roots movement unleashed by Donald Trump in his first presidential campaign, Angela Rubino is a case study in what that movement is becoming. Suspicious of almost everything, trusting of almost nothing, believing in almost no one other than those who share her unease, she has in many ways become a citizen of a parallel America — not just red America, but another America entirely, one she believes to be awash in domestic enemies, stolen elections, immigrant invaders, sexual predators, the machinations of a global elite and other fresh nightmares revealed by the minute on her social media scrolls. She is known online as “Burnitdown.”

Book

• Working on the proposal

Twitter counts:

I tuned in to the Jan 6 committee hearing thinking I’d have it on in the background, but it was riveting. I wound up watching the whole thing

The Internet Needs You-Are-Here Maps

• Nabeel Gillani is a Postdoctoral Associate at the MIT Center for Constructive Communication, and will lead the Plural Connections Group as an Assistant Professor of Design and Data at Northeastern University starting Fall 2022.
• Deb Roy is a professor and director at the MIT Center for Constructive Communication; he is the co-founder, CEO, and chair of Cortico.

Trump Fan Confesses To FBI That He Electroshocked D.C. Cop During Capitol Attack

• The video of the FBI’s March 31 interview of Rodriguez, released to members of the media by federal prosecutors this week after an order from U.S. District Judge Amy Berman Jackson, is a remarkable look at how a radical Trump supporter came to engage in an act of domestic terrorism in hopes of keeping the former president in office for a second term. An emotional Rodriguez explains how he actually believed Trump’s dangerous lies about the 2020 election, referring to himself as “fucking piece of shit,” “so stupid,” “an asshole,” and “not smart” as he confesses his crimes.
• Really good example of stampede behavior

Book

• Reworking the proposal

SBIRs

• Fix broken things? Not sure what will be needed today

GPT Agents

• Continue on tweet app as per here

Call about bike! Does it pass inspection? Nope. Selling it to Bobs for very little

Book

• Finished up the last edits and tweaks for this version and sent out some copies to readers.
• I need to look at what is needed to submit to Cambridge and Stripe. Might as well try Oxford again. I will need to update the proposal

SBIRs

• More FMDS

GPT Agents

• While waiting for Aaron, start getting the pulls working.
  • If NOT randomized, then pull the tweets in sequential order. I think that this can use the query token and just stop when the end is reached.
  • If it is randomized, then randomly select within the span and then pull in sequential order.
  • Make sure that the beginning of the NEXT sequence does not re-use rollover tweets from the previous sequence. If it does, then throw a warning and use the timestamp of the last tweet in the previous pull.
• Set up the schema and tables. We can start with tweet_table, and add a user table later

Towards Learning Universal Hyperparameter Optimizers with Transformers

• Meta-learning hyperparameter optimization (HPO) algorithms from prior experiments is a promising approach to improve optimization efficiency over objective functions from a similar distribution. However, existing methods are restricted to learning from experiments sharing the same set of hyperparameters. In this paper, we introduce the OptFormer, the first text-based Transformer HPO framework that provides a universal end-to-end interface for jointly learning policy and function prediction when trained on vast tuning data from the wild. Our extensive experiments demonstrate that the OptFormer can imitate at least 7 different HPO algorithms, which can be further improved via its function uncertainty estimates. Compared to a Gaussian Process, the OptFormer also learns a robust prior distribution for hyperparameter response functions, and can thereby provide more accurate and better calibrated predictions. This work paves the path to future extensions for training a Transformer-based model as a general HPO optimizer.

SBIRs

• Head down on FMDS

This is a very insightful thread on social dominance:

Book

• Finished with puppetmasters and Jews, started on Diana and Elvis

SBIRs

• Sprint planning
• FMDS writing?

GPT Agents

• 3:30 Meeting
• Good chat with Jimmy. Going to randomize the times within the sampling period

It was a weekend of perfect weather

Book

• Continuing through the conspiracy section

GPT Agents

• Work on getting keyword tweets in fine granular samples (e.g. 5 minutes four times a day at semi-random intervals. Mostly done, though there are all kinds of odd behaviors that involve making too many requests. Working through the options.
• Compare proportions of samples to full counts for the same time periods
• Train a model!

SBIRs

• Sprint review – done
• Write up stories. Probably go back to RCSNN – done

Four levers of reciprocity across human societies: concepts, analysis and predictions

• This paper surveys five human societal types – mobile foragers, horticulturalists, pre-state agriculturalists, state-based agriculturalists and liberal democracies – from the perspective of three core social problems faced by interacting individuals: coordination problems, social dilemmas and contest problems. We characterise the occurrence of these problems in the different societal types and enquire into the main force keeping societies together given the prevalence of these. To address this, we consider the social problems in light of the theory of repeated games, and delineate the role of intertemporal incentives in sustaining cooperative behaviour through the reciprocity principle. We analyse the population, economic and political structural features of the five societal types, and show that intertemporal incentives have been adapted to the changes in scope and scale of the core social problems as societies have grown in size. In all societies, reciprocity mechanisms appear to solve the social problems by enabling lifetime direct benefits to individuals for cooperation. Our analysis leads us to predict that as societies increase in complexity, they need more of the following four features to enable the scalability and adaptability of the reciprocity principle: nested grouping, decentralised enforcement and local information, centralised enforcement and coercive power, and formal rules.

Doing this today!

Diversity injection example:

Book – working through the mapping section