Election day! Absolutely no idea how any of this is likely to play out

Also, create a Mastodon account? I probably have enough info at this point. Applied at fediscience.org – done and set up! I have even tooted

SBIRs

• 9:00 planning meeting
• 10:00 MC meeting?
• Paper
  • Finish populating annotated bibliography
  • Add category and “LMN ranking” to spreadsheet
  • Read top 3(?) papers for each
  • Search all papers for Hi/otL statements and add those quotes to the spreadsheets.
  • Tempted to do some embedding clustering, but that’s overkill
• Add a task to Rukan to check out MinGPT as possible NN for out modules? Done

Book

• Roll in changes

GPT Agents

• More documentation – finished TweetEmbedExplorer
• Start Twitter pull?

Move hotel to January

SBIRs

• Adversarial Policies Beat Professional-Level Go AIs
  • We attack the state-of-the-art Go-playing AI system, KataGo, by training an adversarial policy that plays against a frozen KataGo victim. Our attack achieves a >99% win-rate against KataGo without search, and a >50% win-rate when KataGo uses enough search to be near-superhuman. To the best of our knowledge, this is the first successful end-to-end attack against a Go AI playing at the level of a top human professional. Notably, the adversary does not win by learning to play Go better than KataGo — in fact, the adversary is easily beaten by human amateurs. Instead, the adversary wins by tricking KataGo into ending the game prematurely at a point that is favorable to the adversary. Our results demonstrate that even professional-level AI systems may harbor surprising failure modes. See this https URL for example games.
• 9:00 Sprint Review
• More reading
• Used the LMN tools to figure out what to emphasize and find more papers

GPT Agents

• More documenting
• Figure out some keywords for various groups and start pulling tweets. I think 10k per group a week would be manageable.
  • Watching Twitter implde. Maybe I should just use the pushshift API?
• Reply to First line with some examples

Book

• Meeting with Brenda

Sheesh – still don’t feel particularly good

10:00 Dentist

Large Language Models Are Human-Level Prompt Engineers

• By conditioning on natural language instructions, large language models (LLMs) have displayed impressive capabilities as general-purpose computers. However, task performance depends significantly on the quality of the prompt used to steer the model, and most effective prompts have been handcrafted by humans. Inspired by classical program synthesis and the human approach to prompt engineering, we propose Automatic Prompt Engineer (APE) for automatic instruction generation and selection. In our method, we treat the instruction as the “program,” optimized by searching over a pool of instruction candidates proposed by an LLM in order to maximize a chosen score function. To evaluate the quality of the selected instruction, we evaluate the zero-shot performance of another LLM following the selected instruction. Experiments on 24 NLP tasks show that our automatically generated instructions outperform the prior LLM baseline by a large margin and achieve better or comparable performance to the instructions generated by human annotators on 19/24 tasks. We conduct extensive qualitative and quantitative analyses to explore the performance of APE. We show that APE-engineered prompts can be applied to steer models toward truthfulness and/or informativeness, as well as to improve few-shot learning performance by simply prepending them to standard in-context learning prompts. Please check out our webpage at this https URL.

How Online Mobs Act Like Flocks Of Birds

• A growing body of research suggests human behavior on social media is strikingly similar to collective behavior in nature.

Book

• Done rolling in current edits
• Review and sign contract
• Spend some time working on better terrain. Done!

SBIRs

• More reading, maybe a little writing
  • Finished Artificial intelligence and the changing sources of competitive advantage – fantastic
  • 11:30 Meeting

Had a reaction to the latest booster. I feel like an elephant sat on me

Good thread on online radicalization of a primed subject

Graphika leverages AI to reveal and study online communities. We are the best in the world at analyzing how online social networks form, evolve, and are manipulated.

SBIRs

• Twitter dev conference was canceled. Trying to get my funds back
  • Credit for SW
  • Can move the hotel into Dec/Jan
• Working on MORS paper

GPT Agents

• Half a meeting last night. Time zone issues. We might look at the changes in right-wing and left-wing interactions on Twitter pre and post-Musk
• Adding spreadsheet output to tweet counts and Wikipedia counts – Done!

Book

• Rolling in changes
• Promised to have the contract back by Friday COB

I think the quality of Twitter is dropping

SBIRs

• One of the things to add as suggestions is a model-training facility with dedicated staff. The facility exists to train up to very large models that are resilient to attack (think of a GPT-3 ensemble), and staffed with people who study how models fail. The facility also trains faulty models (mode collapse, overfitting, etc) that can be invisibly swapped in for verified (whatever that means) models so that AI pilots can learn to recognize degraded model behavior. Lots of simulators that allow users to be trained in high-stress situations to adapt to failing models.
• Since the facility trains many models, it will be possible to train meta models that can understand which hyperparameters and data sets produce effective models, and how to degrade them. This will be extremely valuable as AI/ML continue to move into more roles that were previously occupied by highly trained and/or experienced people.
• Find chess paper that shows AI/human tams out-perform AI-only

Book

• Roll in edits
• Got Terragen to make some smooth, rugged, and random terrains. I think I figured out enough, but here’s the tutorial: planetside.co.uk/terratuts-getting-started

GPT Agents

• More documentation
• Need to figure out some keywords for watching Twitter pre/post Musk
• 4:00 Meeting

Took some much needed PTO

Tasks

• 1:45 booster
• Vote!

SBIRs

• Write paper
• 9:15 standup
• 10:00 Weekly meeting

Book

• 4:00 Meeting

GPT Agents

• Documentation