286 Matching Annotations
  1. Last 7 days
    1. The breakthroughs are all underpinned by a new class of AI models that are more flexible and powerful than anything that has come before. Because they were first used for language tasks like answering questions and writing essays, they’re often known as large language models (LLMs). OpenAI’s GPT3, Google’s BERT, and so on are all LLMs. But these models are extremely flexible and adaptable. The same mathematical structures have been so useful in computer vision, biology, and more that some researchers have taken to calling them "foundation models" to better articulate their role in modern AI.

      Foundation Models in AI

      Large language models, more generally, are “foundation models”. They got the large-language name because that is where they were first applied.

  2. Jan 2023
    1. A term recommended by Eve regarding an interdisciplinary approach that accounts for multiple feedback loops within complex systems. Need to confer complex systems science to see if ADHD is already addressed in that domain.

    1. Feng, 2022. "Training-Free Structured Diffusion Guidance for Compositional Text-to-Image Synthesis"

      Shared and found via: Gowthami Somepalli @gowthami@sigmoid.social Mastodon > Gowthami Somepalli @gowthami StructureDiffusion: Improve the compositional generation capabilities of text-to-image #diffusion models by modifying the text guidance by using a constituency tree or a scene graph.

  3. Dec 2022
    1. If my interpretation of the Retrieval quadrant is correct, it will become much more difficult to be an average, or even above average, writer. Only the best will flourish. Perhaps we will see a rise in neo-generalists.

      This is probably true of average or poor software engineers given that GPT-3 can produce pretty reasonable code snippets

    1. "Decision Transformer: Reinforcement Learning via Sequence Modeling" (Chen, NeurIPS, 2021)

      Quickly a very influential paper with a new idea of how to learn generative models of action prediction using SARSA training from demonstration trajectories. No optimization of actions or rewards, but target reward is an input.

    1. Develop Credential Quality Guidelines and Processes

      Noteworthy that the recommendations for quality prioritize 1) The granularity of documenting learning outcomes; and 2) that credentials use standards that can be independently verified and validated.

    2. Standardization of these concepts would allow for validators to sift through credential wallets anddistinguish which credentials are most relevant in a specific use case. Critical to linking up such trustinformation is a more prominent role for dedicated trust providers in the credential ecosystem.These organizations include accreditation boards and regulators of professions, as well as otherssuch as ranking boards and private quality assurance agencies who publish quality standards foreducational organizations and maintain lists of which organizations match the criteria

      What constitutes TRUST?

    3. Multiple initiatives have tried to make various kinds of social recommendations by issuingcredentials. However, up to this point they have worked better in closed social networks rather thanas open credentials due to the ability of social networks to tie a recommendation with the profile(and identity) of the recommender. There are also several nascent initiatives to create open linkeddata around which skills, credentials and issuers are valued by employers.

      Clearly, the LinkedIn recommendations use case is an example of one of these initiatives. It has not succeeded in creating strong social signals anchored in trust models. We are wise to consider what's missing from efforts like this. An even greater concern however, and one that I believe is an essential if we are to realize the transformative potential of digital credentials, is how to design social signals built on trust models that help all people. In a world long-governed by "it's not what you know, it's who you know," the social signals and trust models are overweighted in favor of people with connections to other people, organizations and brands that are all to some degree legacies of exclusionary and inequitable systems. We are likely to build new systems that perpetuate the same problems if we do not intentionally design them to function otherwise. For people (especially those from historically underserved populations) worthy of the recommendations but lacking in social connections, how do they access social recommendations built on trust models?

    1. One of the clear signs that the bottleneck to low-income adults working moreresults from their lack of opportunities is provided by looking at their hours of workover the business cycle. When the economy is strong and jobs are plentiful, low-incomeworkers are more likely to find work, find work with higher pay, and be able to securemore hours of work than when the economy is weak. In 2000, when the economy wasclose to genuine full employment, the unemployment rate averaged 4.0 percent and thepoverty rate was 11.3 percent; but in 2010, in the aftermath of the Great Recession, theunemployment rate averaged 9.6 percent and the poverty rate was almost 15.1 percent.What changed in those years was not poor families’ attitudes toward work but simplythe availability of jobs. Among the bottom one-fifth of nonelderly households, hoursworked per household were about 40 percent higher in the tight labor market of 2000than in recession- plagued 2010.Given the opportunity for work or additional work hours, low-income Americanswork more. A full-employment agenda that increases opportunities in the labor market,alongside stronger labor standards such as a higher minimum wage, reduces poverty.

      How can we frame the science of poverty with respect to the model of statistical mechanics?

      Unemployment numbers have very little to do with levels of poverty. They definitely don't seem to be correlated with poverty levels, in fact perhaps inversely so. Many would say that people are lazy and don't want to work when the general reality is that they do want to work (for a variety of reasons including identity and self-esteem), but the amount of work they can find and the pay they receive for it are the bigger problems.

    1. natural-language processing is going to force engineers and humanists together. They are going to need each other despite everything. Computer scientists will require basic, systematic education in general humanism: The philosophy of language, sociology, history, and ethics are not amusing questions of theoretical speculation anymore. They will be essential in determining the ethical and creative use of chatbots, to take only an obvious example.
    1. Houston, we have a Capability Overhang problem: Because language models have a large capability surface, these cases of emergent capabilities are an indicator that we have a ‘capabilities overhang’ – today’s models are far more capable than we think, and our techniques available for exploring the models are very juvenile. We only know about these cases of emergence because people built benchmark datasets and tested models on them. What about all the capabilities we don’t know about because we haven’t thought to test for them? There are rich questions here about the science of evaluating the capabilities (and safety issues) of contemporary models. 
  4. Nov 2022
    1. 11/30 Youth Collaborative

      I went through some of the pieces in the collection. It is important to give a platform to the voices that are missing from the conversation usually.

      Just a few similar initiatives that you might want to check out:

      Storycorps - people can record their stories via an app

      Project Voice - spoken word poetry

      Living Library - sharing one's story

      Freedom Writers - book and curriculum based on real-life stories

    1. Misleading Templates There is no consistent re-lation between the performance of models trainedwith templates that are moderately misleading (e.g.{premise} Can that be paraphrasedas "{hypothesis}"?) vs. templates that areextremely misleading (e.g., {premise} Isthis a sports news? {hypothesis}).T0 (both 3B and 11B) perform better givenmisleading-moderate (Figure 3), ALBERT andT5 3B perform better given misleading-extreme(Appendices E and G.4), whereas T5 11B andGPT-3 perform comparably on both sets (Figure 2;also see Table 2 for a summary of statisticalsignificances.) Despite a lack of pattern between

      Their misleading templates really are misleading

      {premise} Can that be paraphrased as "{hypothesis}"

      {premise} Is this a sports news? {hypothesis}

    2. Insum, notwithstanding prompt-based models’impressive improvement, we find evidence ofserious limitations that question the degree towhich such improvement is derived from mod-els understanding task instructions in waysanalogous to humans’ use of task instructions.

      although prompts seem to help NLP models improve their performance, the authors find that this performance is still present even when prompts are deliberately misleading which is a bit weird

    3. Suppose a human is given two sentences: “Noweapons of mass destruction found in Iraq yet.”and “Weapons of mass destruction found in Iraq.”They are then asked to respond 0 or 1 and receive areward if they are correct. In this setup, they wouldlikely need a large number of trials and errors be-fore figuring out what they are really being re-warded to do. This setup is akin to the pretrain-and-fine-tune setup which has dominated NLP in recentyears, in which models are asked to classify a sen-tence representation (e.g., a CLS token) into some

      This is a really excellent illustration of the difference in paradigm between "normal" text model fine tuning and prompt-based modelling

    1. Antibiotic resistance has become a growingworldwide concern as new resistance mech-anisms are emerging and spreading globally,and thus detecting and collecting the cause– Antibiotic Resistance Genes (ARGs), havebeen more critical than ever. In this work,we aim to automate the curation of ARGs byextracting ARG-related assertive statementsfrom scientific papers. To support the researchtowards this direction, we build SCIARG, anew benchmark dataset containing 2,000 man-ually annotated statements as the evaluationset and 12,516 silver-standard training state-ments that are automatically created from sci-entific papers by a set of rules. To set upthe baseline performance on SCIARG, weexploit three state-of-the-art neural architec-tures based on pre-trained language modelsand prompt tuning, and further ensemble themto attain the highest 77.0% F-score. To the bestof our knowledge, we are the first to leveragenatural language processing techniques to cu-rate all validated ARGs from scientific papers.Both the code and data are publicly availableat https://github.com/VT-NLP/SciARG.

      The authors use prompt training on LLMs to build a classifier that can identify statements that describe whether or not micro-organisms have antibiotic resistant genes in scientific papers.

    1. “The metaphor is that the machine understands what I’m saying and so I’m going to interpret the machine’s responses in that context.”

      Interesting metaphor for why humans are happy to trust outputs from generative models

    1. "On the Opportunities and Risks of Foundation Models" This is a large report by the Center for Research on Foundation Models at Stanford. They are creating and promoting the use of these models and trying to coin this name for them. They are also simply called large pre-trained models. So take it with a grain of salt, but also it has a lot of information about what they are, why they work so well in some domains and how they are changing the nature of ML research and application.

  5. Oct 2022
    1. Business ModelWill I get charged at some point? How do you make money to run this product?TBD

      "TBD 🚀🚀🚀" is such a bad indication for the future of a product

  6. Aug 2022
    1. Harris said this model is often better for the textbook authors OpenStax works with, whom Harris called "the long tail" behind the minority of financially successful academic authors -- those who wouldn't necessarily sell enough units to make a lot in royalties, but who are committed to their work nonetheless.
    2. "We are fully committed to providing affordable, high-quality learning solutions for students," Joyner said. "We are excited to think openly and collaboratively with key partners like OpenStax to ensure that we, and our authors, are able to reach as many students as possible in new and highly accessible ways."
  7. Jul 2022
    1. https://www.youtube.com/watch?v=7s4xx_muNcs

      Don't recommend unless you have 100 hours to follow up on everything here that goes beyond the surface.

      Be aware that this is a gateway for what I'm sure is a relatively sophisticated sales funnel.


      Motivational and a great start, but I wonder how many followed up on these techniques and methods, internalized them and used them every day? I've not read his book, but I suspect it's got the usual mnemonic methods that go back millennia. And yet, these things are still not commonplace. People just don't seem to want to put in the work.

      As a result, they become a sales tool with a get rich quick (get smart quick) hook/scheme. Great for Kwik's pocketbook, but what about actual outcomes for the hundreds who attended or the 34.6k people who've watched this video so far?

      These methods need to be instilled in youth as it's rare for adults to bother.


      Acronyms for remembering things are alright, but not incredibly effective as most people will have issues remembering the acronym itself much less what the letters stand for.


      There seems to be an over-fondness for acronyms for people selling systems like this. (See also Tiago Forte as another example.)

  8. Jun 2022
    1. Ernest Hemingway was one of the most recognized and influentialnovelists of the twentieth century. He wrote in an economical,understated style that profoundly influenced a generation of writersand led to his winning the Nobel Prize in Literature in 1954.

      Forte is fairly good at contextualizing people and proving ethos for what he's about to present. Essentially saying, "these people are the smart, well-known geniuses, so let's imitate them".

      Humans are already good at imitating. Are they even better at it or more motivated if the subject of imitation is famous?

      See also his sections on Twyla Tharp and Taylor Swift...

      link to : - lone genius myth: how can there be a lone genius when the majority of human history is littered with imitation?

    1. It was as if Silicon Valley had made a secret pact to subsidize the lifestyles of urban Millennials. As I pointed out three years ago, if you woke up on a Casper mattress, worked out with a Peloton, Ubered to a WeWork, ordered on DoorDash for lunch, took a Lyft home, and ordered dinner through Postmates only to realize your partner had already started on a Blue Apron meal, your household had, in one day, interacted with eight unprofitable companies that collectively lost about $15 billion in one year.

      ...but we'll make up for it in volume.

    1. Free public projects private projects starting at $9/month per project

      For many tools and apps payment for privacy is becoming the norm.

      Examples: - Kumu.io - Github for private repos - ...

      pros: - helps to encourage putting things into the commons

      cons: - Normalizes the idea of payment for privacy which can be a toxic tool.

      discuss...

  9. May 2022
  10. Apr 2022
    1. Kai Kupferschmidt. (2021, December 1). @DirkBrockmann But these kinds of models do help put into context what it means when certain countries do or do not find the the variant. You can find a full explanation and a break-down of import risk in Europe by airport (and the people who did the work) here: Https://covid-19-mobility.org/reports/importrisk_omicron/ https://t.co/JXsYdmTnNP [Tweet]. @kakape. https://twitter.com/kakape/status/1466109304423993348

  11. Feb 2022
    1. Learnings: - It's easy to assume people in the past didn't care or were stupid. But people do things for a reason. Not understanding the reason for how things are is a missed learning opportunity, and very likely leads to unintended consequences. - Similar to having a valid strong opinion, one must understand why things are as they are before changing them (except if the goal is only signaling).

    1. In her 2021 book "Bet on Yourself," which features a foreword by Schmidt, Hiatt lays out the two key ways she "up-leveled" her career."First I have prioritized finding a manager who is modeling the career path I want to take and embodies the leadership qualities I want to possess," she wrote. "Second, I have chosen roles that surround me with top quality people and a depth of opportunities to grow with them."

      Look at their life and how it can bring opportunities and then if you will be exposed and streched.

    1. Our brains work not that differently in terms of interconnectedness.Psychologists used to think of the brain as a limited storage spacethat slowly fills up and makes it more difficult to learn late in life. Butwe know today that the more connected information we alreadyhave, the easier it is to learn, because new information can dock tothat information. Yes, our ability to learn isolated facts is indeedlimited and probably decreases with age. But if facts are not kept

      isolated nor learned in an isolated fashion, but hang together in a network of ideas, or “latticework of mental models” (Munger, 1994), it becomes easier to make sense of new information. That makes it easier not only to learn and remember, but also to retrieve the information later in the moment and context it is needed.

      Our natural memories are limited in their capacities, but it becomes easier to remember facts when they've got an association to other things in our minds. The building of mental models makes it easier to acquire and remember new information. The down side is that it may make it harder to dramatically change those mental models and re-associate knowledge to them without additional amounts of work.


      The mental work involved here may be one of the reasons for some cognitive biases and the reason why people are more apt to stay stuck in their mental ruts. An example would be not changing their minds about ideas of racism and inequality, both because it's easier to keep their pre-existing ideas and biases than to do the necessary work to change their minds. Similar things come into play with respect to tribalism and political party identifications as well.

      This could be an interesting area to explore more deeply. Connect with George Lakoff.

    1. Most writing is chasing clout, rather than insight

      As the result of online business models and SEO, most writing becomes about chasing clout and audience eyeballs rather than providing thought provoking insight and razor sharp analysis. The audience reaction has weakened with the anger reaction machines like Twitter.

      We need better business models that aren't built on hype.

    1. Founded in partnership with a team of entrepreneurial journalists who believe in a better model to create excellent content while narrowing the synapse between elite creators and their audiences.

      http://puck.news/who-is-puck/

      Another platform play of journalists banding together to find a niche space of readers.

    1. Aligning editorial mission and business model is critical.

      One of the most complex questions in journalism in the past decade or more is how can one best align editorial mission with the business model? This is particularly difficult because the traditional business model(s) have been shifting in the move to online.

    2. Axios Pro is bundling newsletters together in a high-priced subscription product ($2,500 for the bundle; $599 each) aimed squarely at deep-pocketed investors.

      Old business advice: find the rich and charge them a pretty penny for something they either think they need or fear they can't live without.

  12. Dec 2021
  13. Nov 2021
    1. also into business models that can better serve the interests of both students and educators.

      We are at a point in time where we need to reflect on our business practices. The question should be who are we serving and how do we show we are serving them.

  14. Oct 2021
    1. “Speed kills.” If you are able to be nimble, assess the ever-changing environment, and adapt quickly, you’ll always carry the advantage over any opponents. Start applying the OODA Loop to your day-to-day decisions and watch what happens. You’ll start to notice things that you would have been oblivious to before. Before jumping to your first conclusion, you’ll pause to consider your biases, take in additional information, and be more thoughtful of consequences.

      In che modo si può applicare il modello OODA Loop nella vita quotidiana?

      Semplicemente applicando ad ogni nostra decisione le fasi previste dal modello, rendendo questo processo una abitudine riusciremo ad essere sempre più veloci nell'eseguirlo e questo ci darà la velocità necessaria per sopravvivere e vincere.

    2. When you act fast enough, other people view you as unpredictable. They can’t figure out the logic behind your decisions.

      Quale è il ruolo della velocità e della prevedibilità del nostro operato nel modello OODA Loop ?

      Operare a velocità maggiore degli altri ci rende imprevedibili e questo ci fornisce un vantaggio competitivo, adatto all'OODA Loop che è per definizione un modello fluido.

    3. Boyd made use of the Second Law of Thermodynamics. In a closed system, entropy always increases and everything moves towards chaos. Energy spreads out and becomes disorganized. Although Boyd’s notes do not specify the exact applications, his inference appears to be that a fighter pilot must be an open system or they will fail. They must draw “energy” (information) from outside themselves or the situation will become chaotic. They should also aim to cut their opponent off, forcing them to become a closed system.

      In che modo la [[Seconda legge della termodinamica]] si applica all' #incertezza e come possiamo utilizzarla come parte del modello OODA Loop ?

      Il principio afferma che all'interno di un sistema chiuso tutto tenderà sempre all'entropia. Per questo bisogna essere dei sistemi aperti acquisendo ogni volta informazioni dal contesto, così da evitare che la situazione diventi caotica.

    4. The second concept Boyd referred to is Heisenberg’s Uncertainty Principle. In its simplest form, this principle describes the limit of the precision with which pairs of physical properties can be understood. We cannot know the position and the velocity of a body at the same time. We can know either its location or its speed, but not both.

      In che modo si applica il [[principio di indeterminazione di Heisenberg]] nel modello [[OODA Loop]] ed in che modo ci aiuta ad affrontare l' #incertezza ?

      Il principio afferma che è impossibile determinare in maniera specifica due proprietà fisiche allo stesso tempo.

      Boyd estende questo concetto anche alla gestione delle informazioni, cercare di gestire al meglio due variabili informative diverse è troppo difficile ed all'atto pratico induce a maggiore incertezza

    5. Boyd referred to three key principles to support his ideas: Gödel’s theorems, Heisenberg’s Uncertainty Principle, and the Second Law of Thermodynamics. Of course, we’re using these principles in a different way from their initial purpose and in a simplified, non-literal form.

      Quali sono i tre principi che possono aiutarci nella gestione dell'incertezza e parte integrante del modello OODA Loop?

      • Il teorema di #Godel
      • [[principio di indeterminazione di Heisenberg]]
      • [[Seconda legge della termodinamica]]
    6. Gödel’s theorems indicate any mental model we have of reality will omit certain information and that Bayesian updating must be used to bring it in line with reality. For fighter pilots, their understanding of what is going on during a battle will always have gaps. Identifying this fundamental uncertainty gives it less power over us.

      In cosa consiste il teorema di #Godel e come si colloca nel modello [[OODA Loop]] ?

      Questo teorema afferma che ogni modello mentale sarà sprovvisto di alcune informazioni, è inevitabile, per questo bisogna applicare il metodo di #Bayes per aggiornare le informazioni ed allinearsi alla realtà.

      Già essere consapevoli di questa inevitabile incertezza ci rende più forti nei suoi confronti e capaci di gestirla.

    7. If the opponent uses an unexpected strategy, is equipped with a new type of weapon or airplane, or behaves in an irrational way, the pilot must accept the accompanying uncertainty. However, Boyd belabored the point that uncertainty is irrelevant if we have the right filters in place.

      Cosa è importante ricordare riguardo l'incertezza che deriva da un contesto in cui le informazioni sono sempre in aggiornamento e sempre cambiano secondo il modello [[OODA Loop]] ?

      La cosa più importante da ricordare è che l'incertezza del contesto è irrilevante se si adoperano i giusti filtri decisionali.

    8. If we can’t cope with uncertainty, we end up stuck in the observation stage. This sometimes happens when we know we need to make a decision, but we’re scared of getting it wrong. So we keep on reading books and articles, asking people for advice, listening to podcasts, and so on.

      In quale situazione rischiamo di ritrovarci se non siamo capaci di gestire l'incertezza?

      Rischiamo di ritrovarci in una situazione in cui la paura di prendere una decisione ci paralizza e continuiamo ad osservare, studiare, analizzare senza mai agire.

    9. Speed is a crucial element of military decision-making. Using the OODA Loop in everyday life, we probably have a little more time than a fighter pilot would. But Boyd emphasized the value of being decisive, taking initiative, and staying autonomous. These are universal assets and apply to many situations.

      Quale è il primo dei benefici più importanti di applicare il modello [[OODA Loop]] nella propria vita?

      Consiste nella velocità con cui si possono prendere decisioni, più si utilizza questo metodo più sarà facile muoversi in contesti dalle informazioni variegate perché il pattern decisionale sarà lo stesso.

    10. There’s a difference between making decisions and enacting decisions. Once you make up your mind, it’s time to take action. By taking action, you test your decision out.

      In che modo si collega la fase di azione del modello [[OODA Loop]] a quella di decisione?

      Si collega perché la fase precedente imposta il mindset mentre questa passa all'effettiva azione.

    11. This part of the loop needs to be flexible and open to Bayesian updating. In some of his notes, Boyd described this step as the hypothesis stage. The implication is that we should test the decisions we make at this point in the loop, spotting their flaws and including any issues in future observation stages

      In che termini è importante ragionare quando si parla della fase di decisione del modello [[OODA Loop]] ?

      È importante ragionare non da un punto di vista granitico, è importante avere un approccio di testing ed essere flessibili. Ogni decisione sarà semplicemente una ipotesi da mettere a confronto della realtà ed in funzione della quale avremo dei risultati che saranno informazioni da poter utilizzare per la prossima decisione.

    12. He recommended a process of “deductive destruction”: paying attention to your own assumptions and biases, then finding fundamental mental models to replace them.

      Quale processo può essere alternativo a quello creato da Munger per raccogliere modelli mentali?

    13. He identified the following four main barriers that impede our view of objective information: Our cultural traditions – we don’t realize how much of what we consider universal behavior is actually culturally prescribed Our genetic heritage – we all have certain constraints Our ability to analyze and synthesize – if we haven’t practiced and developed our thinking skills, we tend to fall back on old habits The influx of new information – it is hard to make sense of observations when the situation keeps changing

      Quali sono le barriere principali che impediscono una giusta applicazione della fase di orientamento nell' [[OODA Loop]] ?

      • Le tradizioni culturali: molte di quello che consideriamo oggettivo o ovvio è in realtà una semplice convenzione;
      • Limiti genetici;
      • Limiti razionali, le capacità di pensiero e raziocinio sono frutto di addestramento e quindi possono essere maggiori o minori;
      • La frequenza di aggiornamento delle informazioni nel contesto;
    14. To orient yourself is to recognize any barriers that might interfere with the other parts of the OODA Loop. Orientation means connecting yourself with reality and seeing the world as it really is, as free as possible from the influence of cognitive biases and shortcuts.

      In cosa consiste all'atto pratico la fase di orientamento dell' [[OODA Loop]] ?

      Consiste nel riconoscere le barriere che potrebbero interferire con l'esecuzione delle altre fasi del processo OODA.

    15. If you want to make good decisions, you need to master the art of observing your environment.

      Quale è uno dei presupposti fondamentali da considerare per applicare il modello [[OODA Loop]] ?

  15. Sep 2021
    1. Competent scientists do not believe their own models or theories, but rather treat them as convenient fictions. ...The issue to a scientist is not whether a model is true, but rather whether there is another whose predictive power is enough better to justify movement from today's fiction to a new one. Steve Vardeman, 1987. Comment. Journal of the American Statistical Association 82 : 130-131. [kw]

      easier said than done

    1. A series of studies conducted by Frédéric Vallée-Tourangeau, a professor of psychology at Kingston University in Britain; Gaëlle Vallée-Tourangeau, a professor of behavioral science at Kingston; and their colleagues, has explored the benefits of such interactivity. In these studies, experimenters pose a problem; one group of problem solvers is permitted to interact physically with the properties of the problem, while a second group must only think through the problem. Interactivity “inevitably benefits performance,” they report.

      Physical interactivity with a problem may help improve results.

    2. Moving mental contents out of our heads and onto the space of a sketch pad or whiteboard allows us to inspect it with our senses, a cognitive bonus that the psychologist Daniel Reisberg calls “the detachment gain.”

      Moving ideas from our heads into the real world, whether written or potentially using other modalities, can provide a detachment gain, by which we're able to extend those ideas by drawing, sketching, or otherwise using them.

      How might we use the idea of detachment gain to better effect in our pedagogy? I've heard anecdotal evidence of the benefit of modality shifts in many spaces including creating sketchnotes.

      While some sketchnotes don't make sense to those who weren't present for the original talk, perhaps they're incredibly useful methods for those who are doing the modality shifts from hearing/seeing into writing/drawing.

  16. Aug 2021
    1. “The real economics of college have shifted so much during the last 70 years, and we have not made adjustments to all those changes. Students are in an equation that has not adapted to the circumstances.”
  17. Jul 2021
    1. How a memory palace works When we’re learning something new, it requires less effort if we connect it to something we already know, such as a physical place. This is known as elaborative encoding. Once we need to remember the information, we can “walk” around the palace and “see” the various pieces. The idea is to give your memories something to hang on to. We are pretty terrible at remembering things, especially when these memories float freely in our heads. But our spatial memory is actually pretty decent, and when we give our memories some needed structure, we provide that missing order and context. For example, if you struggle to remember names, it can be helpful to link people you meet to names you already know. If you meet someone called Fred and your grandmother had a cat called Fred, you could connect the two. Creating a multisensory experience in your head is the other part of the trick. In this case, you could imagine the sound of Fred meowing loudly. To further aid in recall, the method of loci is most effective if we take advantage of the fact that it’s easiest to remember memorable things. Memory specialists typically recommend mentally placing information within a physical space in ways that are weird and unusual. The stranger the image, the better.

      This notion of using spatial memory to encode other concepts - or even the P-A-O sytem where a 2 digit number encodes a person performing an action is an interesting idea for someone like me who forgets quite a bit.

  18. Jun 2021
    1. The problem is, algorithms were never designed to handle such tough choices. They are built to pursue a single mathematical goal, such as maximizing the number of soldiers’ lives saved or minimizing the number of civilian deaths. When you start dealing with multiple, often competing, objectives or try to account for intangibles like “freedom” and “well-being,” a satisfactory mathematical solution doesn’t always exist.

      We do better with algorithms where the utility function can be expressed mathematically. When we try to design for utility/goals that include human values, it's much more difficult.

    2. many other systems that are already here or not far off will have to make all sorts of real ethical trade-offs

      And the problem is that, even human beings are not very sensitive to how this can be done well. Because there is such diversity in human cultures, preferences, and norms, deciding whose values to prioritise is problematic.

  19. May 2021
    1. In a way, the essential premise of the collab-house business model is not far from that of pornographic entertainment. (Where else do talent and crew and cadres of management congregate in furnished mansions to produce intimate content?) Interestingly, but maybe not surprisingly, many TikTok influencers, including some here at the Clubhouse, have made the crossover from social media to pornography, using apps such as OnlyFans to post nude pics for their legions of subscribers.
    2. Later in my visit, Chase Zwernemann, the twenty-one-year-old VP of talent management, will tell me that “we really see ourselves like influencing professors.” And if this weren’t enough, under the Clubhouse aegis is a trio of TikTok houses, each of which corresponds, apparently, to a different level of academe. There’s Clubhouse BH—the grad school—which is meant for “our more seasoned influencers.” (If this phrase conjures for you images of geriatrics taking selfies in suggestive postures, please know that by “seasoned influencers” they simply mean people who have been in the business a while and have thus reached the ripe old age of twenty-two or twenty-three.) Beneath that is Clubhouse FTB, which apparently serves as the undergraduate program. And finally, there’s Not a Content House, the high school of the Clubhouse venture, one meant to appeal to an even younger demographic.

      He uses academe, but I might liken it to a studio system of sorts.

    1. In viewing academia as a business, you should always give customers what they want, and this applies on two levels. First, always consider the demand for the research product. This is much easier said than done. Anyone can acknowledge that the customers are always right, but truly listening to them and extracting what they need is difficult, especially if you have your own personal desires with respect to the product (in this case, the research). Talk to the funding customer constantly. Second, most students are, in effect, employees, and the adviser is a boss who doubles as a customer. In some respects, your adviser will provide your pay cheque, or at least govern it. Thus, do what the customer requires. In addition, always consider your audience when writing and presenting. In the case of a thesis, the audience is your adviser and committee. Again, talk to the customers constantly.
    2. Anyone who treats research as a business tends not to be well received in academia, but they likely have the funding necessary to drive advances, and they may eventually be wealthy.

      There can be clear benefits to treating academia like a business.

    1. Amidst the global pandemic, this might sound not dissimilar to public health. When I decide whether to wear a mask in public, that’s partially about how much the mask will protect me from airborne droplets. But it’s also—perhaps more significantly—about protecting everyone else from me. People who refuse to wear a mask because they’re willing to risk getting Covid are often only thinking about their bodies as a thing to defend, whose sanctity depends on the strength of their individual immune system. They’re not thinking about their bodies as a thing that can also attack, that can be the conduit that kills someone else. People who are careless about their own data because they think they’ve done nothing wrong are only thinking of the harms that they might experience, not the harms that they can cause.

      What lessons might we draw from public health and epidemiology to improve our privacy lives in an online world? How might we wear social media "masks" to protect our friends and loved ones from our own posts?

    1. Substack insists that advances are determined by “business decisions, not editorial ones”. Yet it offers writers mentoring and legal advice, and will soon provide editing services.

      Some evidence of Substack acting along the lines of agent, production company, and studio. Then taking a slice of the overall pie.

      By having the breadth of the space they're able to see who to invest in over time, much the same way that Amazon can put smaller companies out of business by knocking off big sales items.

    2. Record labels are another endangered middleman. They have historically taken care of turning a song into a hit, in return for an ongoing share of revenues. But more and more artists are going it alone. More than 60,000 new songs are uploaded to Spotify every day, most by bedroom-based rockstars who can use new online services to handle the logistics themselves. UnitedMasters, a music-distribution platform which bills itself as “a record label in your pocket”, recently raised $50m in a venture-capital round led by Apple. Tools like Splice make recording easier. Companies like Fanjoy take care of merchandise.And financing is getting simpler. One startup, HIFI, helps artists manage their royalties, paying them regularly and fronting them small sums to make up shortfalls. Another, Karat, extends credit to creators based on their follower count. Helped by such services independent artists took home 5.1% of global recorded music revenues last year, up from 1.7% in 2015, calculates MIDiA Research, a consultancy. In the same period the share of the three largest record labels fell from 71.1% to 65.5%.

      The same sort of dis-aggregation and disintermediation that has hit the publishing business is also taking place to newspapers, magazines, and music.

      The question is how to best put the pieces of the pie together in the best way possible. There's probably room for talented producers to put these together to better leverage the artists' work.

  20. Apr 2021
  21. Mar 2021
    1. Paywall success stories are a mix of specialized news, often catering to the ultrawealthy (WSJ), superstar news orgs focused on specific national and international news (NYT) or news with billionaire backstops (WP). The civic function of news is not met by any of these models. But reader-supported, open access news, like Canadaland and The Halifax Examiner are filling in the gaps. The co-op model is a most welcome adjunct to these success stories.

      List of business models for news.

  22. Feb 2021
    1. A popular strategy for bootstrapping networks is what I like to call “come for the tool, stay for the network.” The idea is to initially attract users with a single-player tool and then, over time, get them to participate in a network. The tool helps get to initial critical mass. The network creates the long term value for users, and defensibility for the company.

      This is an interesting and useful strategy. I've heard the idea several times before.

      I'm curious if this is the oldest version of it? I have to imagine that there are earlier versions of it dating back to 2011 or 2012 if not earlier.

    1. the idea of decentralized business models. These are business models distributed across a blockchain network not centralized in a traditional corporation. They can be fully autonomous too, meaning no humans involved. There's big potential for true peer-to-peer models like ride-sharing without Uber as a middleman taking fees. Owners of driverless cars will someday be able to put their autonomous vehicles to work. So even the big disruptors can be disrupted by this new technology.

      Decentralized Business Models

    1. But the inverse trajectory, from which this essay takes its name, is now equally viable: “come for the network, pay for the tool.” Just as built-in social networks are a moat for information products, customized tooling is a moat for social networks.1 This entrenchment effect provides a realistic business case for bespoke social networks. Running a bespoke social network means you’re basically in the same business as Slack, but for a focused community and with tailored features. This is a great business to be in for the same reasons Slack is: low customer acquisition costs and long lifetime value. The more tools, content, and social space are tied together, the more they take on the qualities of being infrastructure for one’s life.

      An interesting value proposition and way of looking at the space that isn't advertising specific.

    2. Most business writing lacks a meaningful engagement with the question of whether the strategies, tactics, and trends on offer are good, in a larger and longer term sense. It is negligent not to address these questions.

      Very few businesses consider the long term effects of their work...

    3. A new business type here is the paid community: a direct subscription to join in. Today, most paid communities live on the outskirts of existing social platforms. But as they become normalized, paid communities are becoming a viable business model for smaller-scale social networks aiming to be both profitable and socially sustainable.

      paid communities

    1. A broad overview of the original web and where we are today. Includes an outline of three business models that don't include advertising including:

      • Passion projects
      • Donation-based sites
      • Subscription-based sites
    1. I wonder how much this mini-article about Twitter subscription services may have been in response to Galloway's article last week?

      Or will they, as he suggests they do so often, make a head fake to something they might do and then just do nothing (again)?

    1. Neither Snap nor Pinterest is free of issues, but to date, nobody has rallied a mob to attack the U.S. Capitol using tastefully curated photos of bathroom remodelings.
    2. <small><cite class='h-cite via'> <span class='p-author h-card'>jeremycherfas </span> in The Capitalist Case for Overhauling Twitter (<time class='dt-published'>02/08/2021 14:14:45</time>)</cite></small>

    1. To prompt this kind of revolution in your own life, Rose and Ogas suggest creating a micromotive, or a goal tailored to an extremely specific activity that truly inspires you. For example, when Korinne Belock left her job as a political aide to form Urban Simplicity, a firm that declutters and redesigns homes and offices, her micromotive was “organizing physical space.” Note that she didn’t say “doing something creative” or “starting my own business.” Those declarations are too general and fuzzy to be acted on. Instead, she identified a task that sparked within her an outsized amount of curiosity and pleasure and used it as her guide.

      To escape a boring and unfulfilled life, create a micromotive, where we tailor an extremely specific activity that burns the spirit inside you.

  23. Jan 2021
    1. Help is coming in the form of specialized AI processors that can execute computations more efficiently and optimization techniques, such as model compression and cross-compilation, that reduce the number of computations needed. But it’s not clear what the shape of the efficiency curve will look like. In many problem domains, exponentially more processing and data are needed to get incrementally more accuracy. This means – as we’ve noted before – that model complexity is growing at an incredible rate, and it’s unlikely processors will be able to keep up. Moore’s Law is not enough. (For example, the compute resources required to train state-of-the-art AI models has grown over 300,000x since 2012, while the transistor count of NVIDIA GPUs has grown only ~4x!) Distributed computing is a compelling solution to this problem, but it primarily addresses speed – not cost.
  24. Dec 2020
    1. what is meant involves some assertion or hypothesis about the model – that it correctly or incorrectly represents some phenomenon in some respect or to some degree.

      The truth is that it represents something real.

    1. “Although we now have at our disposal some fairly sophisticated methods of characterizing uncertainty,” she warned, “these do not actually enable us to control or even predict the extent of the disaster.

      Many believe models predict the future. Exhibit A: Climate change

  25. Nov 2020
    1. We’ve always wanted to build a layer on top of the web where every person can have their mental model of how the whole world works, and they can start to share ideas across everything.

      Conor's idea of Roam was a layer on top of the web where everyone can have their mental model of how the world works.

    1. Fortnite’s monetization model is based on cosmetics: The skin your character wears; the looks of your glider and the tools you use; the way your character dances (emotes) – all of these are signaling amplifiers with different signal messages to uniquely express yourself in the game. And you have to purchase them

      Julian posits that Fortnite's revenue model is also based on signalling. People buy cosmetic upgrades to their character like your tools, your skin color etc.

    1. A better definition I've been using since then, thanks to Jason Hwang, is "fixed output." A company that delivers the same thing to all customers is going to be organized differently than one that does things made-to-order.

      Fixed output companies

      A company that delivers the same thing to all customers. These companies are going to be organized differently than ones that do things bespoke for their customers.

  26. Oct 2020
    1. Like many other stores, Vroman’s is hosting online events to promote new books, which can attract attendees from all over the country but generally bring in almost no money.

      Maybe they need a book paywall for admission into those events? Buy a book to get the zoom code to get into the event?

      David Dylan Thomas essentially did this for his recent book launch.

    1. That is to say: if the problem has not been the centralized, corporatized control of the individual voice, the individual’s data, but rather a deeper failure of sociality that precedes that control, then merely reclaiming ownership of our voices and our data isn’t enough. If the goal is creating more authentic, more productive forms of online sociality, we need to rethink our platforms, the ways they function, and our relationships to them from the ground up. It’s not just a matter of functionality, or privacy controls, or even of business models. It’s a matter of governance.
    1. Twenty Stories Bookmobile, which left L.A. traffic for Providence, Rhode Island, in 2018

      This makes me think that a mobile bookstore a la the traditional LA roach coach with a well painted/decorated exterior could be a cool thing.

      I'm reminded of a used bookstore pop-up I saw recently at the Santa Anita Mall prior to the holidays. Booksellers were traditionally itinerant mongers anyway. Perhaps this could be a more solid model, especially for the lunchtime business crowds.

    1. A common complaint we heard from publishers at all levels is that it’s difficult to build partnerships with social media platforms. They seem to be holding all the cards. Even large publishers often feel in the dark during meetings with large platform companies.

      I'm more curious why all the large media companies/publishers don't pool their resources to build a competing social platform that they own and control so the end value comes to them instead of VC-backed social silos?

  27. Sep 2020
    1. Models

      Models are fancy constructors compiled from Schema definitions. An instance of a model is called a document. Models are responsible for creating and reading documents from the underlying MongoDB database.

      https://mongoosejs.com/docs/models.html

  28. Aug 2020
    1. It might be instructive to think about what it would take to create a program which has a model of eighth grade science sufficient to understand and answer questions about hundreds of different things like “growth is driven by cell division”, and “What can magnets be used for” that wasn’t NLP led. It would be a nightmare of many different (probably handcrafted) models. Speaking somewhat loosely, language allows for intellectual capacities to be greatly compressed. From this point of view, it shouldn’t be surprising that some of the first signs of really broad capacity- common sense reasoning, wide ranging problem solving etc., have been found in language based programs- words and their relationships are just a vastly more efficient way of representing knowledge than the alternatives.

      DePonySum ask us to consider what you would need to program to be able to answer a wide range of eight grade science level questions (e.g. What can magnets be used for.) The answer is you would need a whole slew of separately trained and optimized models.

      Language, they say, is a way to compress intellectual capacities.

      It is then no surprise that common sense reasoning, and solving a wide range of problems, is first discovered through language models. Words and their relationships are probably a very efficient way of representing knowledge.

  29. Jul 2020
    1. Imagine a large population of people living, seeing, learning, doing and generally going about their lives. As they do so, they accumulate beliefs. Depending on how smart they are, they also compress beliefs via abstraction, metaphor, subconscious pattern-recognition circuits, muscle memory, ritual, making and consuming art, going p-value fishing, exploring tantric sex, generating irreproducible peer-reviewed Science! and so on.

      Compression of knowledge through abstractions ~ mental models.

    1. unless the model is embedded in a suitable structure thatpermits extrapolation, no useful inference is possible, either Bayesian or non-Bayesian.
  30. Jun 2020