Desk

Written November 8th, 2023

The AI-Native Future:  

Applying Generative AI to High Quality Personal Datasets (HQPDs)

Executive Summary

With the public introduction of Chat Generative Pre-trained Transformer (GPT) by OpenAI in late 2022, generative AI has become the hot new tool that has exploded onto the market after years of fairly quiet research and development. Projections about the impact of Artificial General Intelligence, “a form of AI that possesses the ability to understand, learn and apply knowledge across a wide range of tasks and domains,” span the gamut of “It’s an over-hyped fad” to “It will usher in a new era that could ultimately save or destroy humanity.”[1] Where does reality lie?

That particular question is too large in scope to effectively tackle in brief; however, this paper will focus on exploring how specifically generative AI technology could affect both consumer data recording and data interaction processes and analyze how External Memory Archive Inc. (Ema.ai), a startup operating in the personal consumer data space, might navigate these waters. It will do so by first introducing and analyzing the current AI landscape, projecting out potential applications of currently developing technology, and then applying these insights to the existing Ema.ai business model.

As a disclaimer, no portion of this essay has been generated with the use of a Large Language Model. However, Chat GPT-4 was utilized for research and feedback purposes. These interactions have been included in the Appendix for review.

Analyzing the Current Landscape

When we think about generative AI, we are generally thinking about Natural Language Processors built on Large Language Models. Natural Language Processing (NLP) is “the ability of a program to understand human language as it is spoken and written.”[2] This is what allows Chat-GPT to translate the user input, “Write me a Shakespearean haiku,” into a computer program instruction that it can process and respond to. In this case, the program is using a Large Language Model (LLM), “a model trained on vast amounts of (often) textual data to predict the next word in a self-supervised manner,” to produce the following output (actually pulled from Chat-GPT)[3]:

“Whispers of sweet love

In fair Verona’s moonlight

Hearts entwined as one.”

Chat GPT doesn’t technically “know” what it has produced in the same way that a Professor of English might have intention and knowledge behind creating an output like the text above. The program is just predicting, based on the human-generated instructions, that these are the words that should “properly” follow each other, as influenced by the constraints of the instruction (that it should use Shakespearean language in the syllabic form of a Haiku) when applied to its pattern-recognition training on a particular dataset. Of note, these training datasets can be enormous. GPT-3, the precursor to the more recent and intricate GPT-4 model, was trained on 175 billion parameters.[4]  

There are dozens of companies building AI chatbots on top of proprietary LLMs, with some of the most well-known LLM-powered chatbots currently under development being: GPT-3.5 (OpenAI), GPT-4 (OpenAI), BARD (Google), LlaMA (Meta), Falcon (Technology Innovation Institute), Cohere (Cohere), PaLM (Google), and Claude v1 (Anthropic).[5] Note the major companies playing in this space: OpenAI (who is partnered with Microsoft), Google, and Meta. Something to highlight is that these generative AI programs, while appearing clever, are not a true form of Artificial General Intelligence. Rather, they are a very very clever mimic that can pick out and emulate patterns in human communication. This mimicry is quite valuable and even helpful, but it is not the same as a sentient robot.

Building these LLMs is expensive and time-consuming. There are four main types of costs associated with developing these models: pre-training, tuning, hosting, and inference costs. Pre-training costs are the costs of initially training an LLM from scratch. Hosting is the cost of maintaining a model behind an API that users utilize for tuning or inference. Fine tuning is the process of “adapting a pre-trained model to perform a specific task by conducting additional training with new data.”  Inferencing is the actual act of prompting an LLM to return a response.[6] Inference and tuning costs can be allocated to the consumer or business utilizing the chatbot, and we see these companies beginning to charge for their chatbot’s services at varied rates of dollar amounts per 1000 tokens. A token is a block of text, generally 3-4 characters, resulting in companies charging fractions of a cent per ~750 words of output.[7]

For the companies, we have seen major costs being driven by the initial training and then hosting of these models. To develop a model requires multiple training runs, with a training run on a large data set requiring thousands of Graphics Processing Units in order to provide the necessary processing power.[8] The hardware alone can cost millions of dollars, with each training run bringing on additional costs. Training a model like GPT-3 requires many training runs with a projected cost of $4.6 million per run.[9] This is good news for companies producing GPUs (like NVIDIA), but there’s no such thing as a free lunch and those building services on top of these currently free-to-use models will soon have to pay up.

Taking stock - these chatbots are cool, pretty smart, and getting smarter. Chat GPT-4 was able to score in the 90th percentile on the Uniform Bar Exam, 163 on the LSAT, and a combined 1410 on the SAT.[10] The United States leads the world in AI research and development and technology is improving.[11] And now, there are commercial opportunities not only for those who build LLMs, but also those who build tools that apply them. Hundreds of companies have popped up leveraging Generative AI to build co-pilots and various other services for both general consumers and business customers, with 24% of VC investments going into AI companies in 2023.[12]

Where does that leave us? We live in a world where the technical ability of AI-powered programs is growing every day. My hypothesis is that in the future our society will be AI-native, with generative AI integrated seamlessly in our daily life. This next section will explore how that world may work.

 Looking Forward: Theoretical Commercial Applications of this New Technology

To quickly address the elephant in the room, some people believe that we will achieve an Artificial General Intelligence in the near future that will replace humans in their role as innovators and turn into a machine servant that comes up with the solutions to all of humanity’s problems and removes the demand for human labor.[13] Could this happen? Perhaps. Is it likely? That’s hard to say. There are people that are very confident that this will occur. I don’t believe this paper has conducted enough analysis to have an opinion on the matter.

What I am more interested in is analyzing how high-quality generative AI, technology that we are already making measurable and speedy progress towards, could affect the ways that we record and interact with our personal data. To that end, I have three theses. First, generative AI expands the Platform Interaction Universe for users engaging with their personal data. Second, expanding the Platform Interaction Universe reduces interaction friction to the point that it increases user engagement with data collection technology. Third, the creation of high-quality personal datasets will allow for the creation of a platform where tools can be developed that allow users to leverage these datasets for personal enjoyment and/or financial enrichment.

Humans, in the course of our lives, generate data simply through the act of existing. The number of breaths we take, hours we sleep, words we speak, memories we make - these are all data points that we, as living creatures, produce. On top of this, humans, seemingly unique to our species, collect and store our data for temporally or physically distant interactions. We memorize stories so that we can share them years later around a campfire as a way to reminisce about the past or to share important lessons. We write down ideas so that we can transmit them into another person’s brain through their reading of physical scratches on a page or digital markings on a screen. The power of communication really lies in the ability to record and then share data in a way that transmits information across time and space.

To that end, humans have developed increasingly complex methods of collecting and recording data, as well as interesting ways through which to interact with data once it has been stored. In the field of personal memories - some people journal about their emotions and feelings as a practice of self-reflection. Families use photo albums, home videos, memoirs, and scrapbooks as a way to preserve their personal histories. In the field of productivity and professional development - some laboriously note down every task they complete in a day. Others build large rolodexes and make sure to pay special care to logging personal information of relevant contacts.

Suffice it to say, we humans care a lot about our data. Recording it, though, is a laborious and time-intensive process. And then mindfully interacting with it? Most of us won’t even make it to the point where we record a good dataset to begin with, and by the time we do we’re probably too tired to do much with it. This is where generative AI could play a very interesting role.

Before diving into the theses, I would first like to explore the basic building blocks of how we record and interact with our data.

When it comes to the recording of our personal data as we engage with modern technology, we either do it intentionally or unintentionally. Intentional data recording occurs when a user is aware that they are taking actions which result in their user data being recorded. For example, if I decided to post a picture on social media, I am aware that my picture is being stored and even published. Or if I write a journal entry or take a picture on my phone, I am generally doing so with the intent to capture the data that is being stored. Unintentional data recording occurs when a user is unaware that they are taking actions which result in their user data being recorded. For example, Google used to read Gmail users’ emails in order to personalize advertising to that user.[14] Now you might say, “Well, a user is aware that they are sharing and storing textual messages when they send an email,” but Google’s use of those communications to build an ad profile is certainly not in the spirit of how they were intending to use the platform. Companies like Meta and TikTok are going further and collecting quite personal information even when you are not engaging in an act that could be construed as voluntary data storing.

Back when it was just called Facebook, Meta would analyze user interactions (for example, do you like the Facebook page of Ben and Jerry’s ice cream), and use it to tag the user with a political affiliation as part of the user profile that they would sell to advertisers.[15] Tik Tok, a platform where users scroll through an endless stream of short videos, uses a proprietary algorithm to evaluate what a user likes to watch in order to feed them videos that will keep them hooked on the platform. Just by measuring how long you linger on a video, the company is able to identify things like your religious beliefs, political affiliation, taste in music and clothes, personal values, even current emotional state, etc.[16] This might seem terrifying at first, but there is a silver lining. Technology has made it easy to record personal data at no labor cost to the user. Whenever a person is recording data intentionally, there is generally a significant energy cost. Sitting down to write a diary entry takes time and effort. Now, technology records user data while they are doing something else. It’s as if someone else had followed you around while you just went about your day, and then gave you a gift-wrapped journal at the end of the day with everything you had done and thought nicely written down for you. Unintentional data recording reduces the mental and physical energy cost to the user. The main issue here, however, is that the user is not the current beneficiary of their own data profile.

Once data has been recorded, users can interact with it in a variety of ways. Data interactions can range from reading old diary entries to pulling out sleep habit insights from biotrackers to reading through emails to remind oneself about the content covered in a prior conversation. Physically recorded data is more difficult to manage than digital. Take for example the following personal habits tracker. This, along with many other systems, is a way to systemize data collection so as to provide more relevant information trends that will allow an individual to improve their performance over time.

The higher the quality of a dataset the more powerful insights can be drawn. However, there are limitations to this. Without supplemental tools summarizing and helping draw out inferences, humans are not built to keep hundreds of thousands of datapoints in their head to call on at will. It is why we have a need to record data externally in the first place.

Tying in these two pieces together, humans can benefit from recording and interacting with high-quality personal datasets (HQPDs). To fully leverage the power of these HQPDs requires additional tools to assist in both the recording process (preventing garbage in, garbage out), and to analyze and assist in meaningfully interacting with such datasets at scale. A great example of this would be biometric health data. As our recording tools have improved, from intentional recording of notes in a physical log to unintentional / unconscious record of biometric data through tracking devices like the Aura ring or Whoop bracelet, users now have the ability to draw much more powerful insights into their personal health. This begs the question, how will generative AI impact the ways in which we record and interact with our data?

Thesis 1: Generative AI expands the Platform Interaction Universe

The concept of the Platform Interaction Universe (PIU) is as follows. Users engage with the systems around them through various platforms. Their interactions with these platforms can be expansive or limited depending on what the platform allows. A very simple platform interactive universe is that of a light-switch: the interactive universe is binary – the switch is on or off. Contrast this with the interactive universe of the platform of conversation. If you are speaking with someone you can understand, the two participants can analyze and respond to a vast array of varied inputs. As another example, imagine a physical diary. The diary is the platform through which an individual engages with their memories and recollections. Their ability to interact with this platform (the platform interaction universe) is fairly limited- users can draw or write on the pages of the book and engage with this collected data by re-reading previous journal entries. This is an example of a platform that has a physically limited platform interaction universe.

Generative AI has expanded the platform interaction universe for users that are engaging with computer programs, and in this specific case, programs designed to help users record and store their memories. Originally, coding, communicating, and interacting with computers required very specific inputs that generated very specific outputs. Even modern sophisticated customer services limit the users’ platform interaction universe (the dreaded “Press ‘1’ if you have an issue with one of your orders, Press ‘2’ if you’d like to ______” – although this is quickly changing with the introduction of generative AI customer service representatives). Note that when the computer is unable to categorize or understand an input, a human must step in to solve the problem. That is because one of the most expansive platform interaction universes is that of human conversation. Assuming both participants speak the same language, humans can process and understand an almost infinite number of inputs from their conversational partner. I could talk to you about my day, a red ball, an angry dog, the concept of modern democracy, cryptocurrency, or thalassophobia, and humans are good at taking these inputs, processing them, and generating relevant responses.

Natural Language Processing expands the ability of computers to categorize, process, and respond to varied human inputs while still producing relevant and high-quality outputs. This means human data, often presented and expressed most naturally conversationally, will be more decipherable to computer programs seeking to compile these datasets and provide value to the user. In short, the application of this technology means it will be easier for humans to engage with computers for both intentional and unintentional data recording.

Thesis 2: Expanding the Platform Interaction Universe Reduces Interaction Friction

To translate this thesis into plainer language, Natural Language Processing will increase user engagement with technology due to improving ease of use. Even more simply put, being able to talk to a computer like you would another person (intentional data collection) makes you more likely to talk to the computer and makes it easier for the computer to decipher your thoughts even when not speaking directly with the computer (unintentional data collection). This analysis is driven by the following concept:

Value Add Δ = Value Add – Interaction Friction

Consumers respond to incentives. They adopt products that they find to have a positive value add delta. Something that, even when accounting for the cost of attaining the product, leaves the consumer better off. The value add component is fairly self-explanatory. It is comprised of the value generated for the user by using the product. Interaction friction can be driven by a number of factors, but it is ultimately determined by the user’s expenditure of resources – time, money, physical labor, mental energy, etc. There might be a restaurant I quite enjoy, but if it is too difficult to get to or too expensive, I might settle for a meal I enjoy less but that is easier for me to cook or buy. The Value Add Δ is greater for the “worse” meal. This heuristic might seem definitionally self-proving. If people only use things at a positive Value Add Δ then everything they do will calculate out that way. What is interesting is to observe edge cases – either within a product space (higher quality product with greater expenses as compared to a lower quality product in the same category with lower expenses) or over time (the development of new technology changing either the value add or the interaction friction such that adoption increases).

As another example of the application of this framework, early computers were notoriously difficult to use, with programmers having to manually feed in programming cards to program the computer. However, the value add of using technology to replace human labor, especially in regards to large number computation, was so great for a portion of the population that some people were still willing to learn how to leverage the platform. As technology improved and it became easier to engage with the technology the formula was attacked on both ends – the value add increased and the interaction friction decreased – leading to widespread adoption of the technology by the general public.

To apply this framework to the personal memory / high quality personal dataset space: personal memories/data clearly have a high value add for individuals. Not only does remembering happy memories actually make a person happier, humans are generally obsessed with themselves. People get a lot of value out of being perceived, applauded, and understood. In spite of this, we do not do a good job of recording or interacting our memories. Why is that? Intentionally recording these memories and data has relatively high interaction friction. The current technology landscape for intentional data recording is fairly simple: users generate content at-will that gets plugged into a recording document of some form: a journal, a spreadsheet, maybe a draft of a memoir or a page on the notes app on their phone, a photo album or home video collection. Unintentional datasets, which are probably more fleshed out and holistic, are not readily accessible to the users as they are owned and controlled by the Big Tech companies. This is where the opportunity lies.

The theory here is that increasing the PIU of memory collection technology through the medium of Natural Language Processing reduces the interaction friction for the user in the memory collection process. An AI-supported personal assistant that logs both intentional and unintentional data collection will promote strong user engagement. This increase in user engagement drives the next point of analysis – that this adoption will lead to the development of larger and better HQPDs.

 Thesis 3: Increased Adoption Results in More High-Quality Personal Datasets

Imagine an AI personal assistant that collects both intentional and unintentional data. It reads your emails and texts, listens to your phone calls and meetings, tracks your web browsing and the content you consume (social media, news articles, videos), and even collects your biometric data. That might sound like a lot if not outright invasive, but it’s already happening (just spread out across a couple different companies). Now imagine that it not only does that, but it also creates a digital avatar that you can talk to and share your feelings with. A personal friend, therapist, mentor, and secretary/scribe all rolled into one. Any time you want to make sure you record a special moment, a memory, picture, or video that you took of your family or friends, this bot is ready and waiting for you to ask for help – and when you do, all you have to do is talk to it like you would a normal person. This is what is on the horizon.

The more an individual opts into this system the greater the quality of HQPDs are produced. These HQPDs are the ultimate generator of value for the operator in this space.

HQPDs are valuable to the user because, with the help of an AI assistant, they serve as the dataset from which holistic and actionable insights can be drawn. Collection of your diet and biometric data could tell you what meals result in the best quality of sleep for you. A fully integrated dataset could build connections between personal relationships and mood – it could notice that every time you speak to a certain individual your heart rate spikes. The power of big data analysis as applied to one individual’s life is great – but the technology has had too much friction for it to be worthwhile for the consumer. This is all possible with existing technology, but the issue is that consumers haven’t been willing to input the data to get the results. It takes a lot of work to track everything you eat and your quality of sleep and then to cross-reference the data and pull out insights. The juice hasn’t been worth the squeeze. Now, you could snap a picture of your meal and have the generative AI categorize it and cross-reference it to the sleep data collected by your wearable biodata tracker. If your AI listens in on your zoom meetings it can build relational databases that tracks who you’re speaking with (which could then be cross-referenced with your biometric emotion indicators).

Ultimately, once the data has been collected, the use cases are almost infinite. Above are a few of the immediate ways that value could be added to the user’s life, but there are interesting fringe cases to consider. With a big enough HQPD, it would be possible to train a digital model on how to speak, sound, and even look like the user generating the dataset. A company could offer a service that builds digital avatars for individuals that their future descendants could speak with. Companies might offer to pay for users’ HQPDs to tailor specific ads, and consumers would be able to sell their own data in exchange for the convenience of receiving recommendations for purchases they are actually interested in. The potential applications are boundless.

The company that empowers users to generate HQPDs, and then creates a platform for companies to sell applications that users can plug their personal datasets into, will occupy an essential role as the marketplace for data applications.

[1] Gartner Glossary, November 9th, 2023, Gartner.

[2] State of AI Report, October 12th, 2023, Air Street Capital.

[3] Id.

[4] Language models are few-shot learners, May 28th, 2020, OpenAI.

[5] Best Large Language Models for 2023 and How to Choose the Right One for Your Site, October 6th, 2023, Hostinger Tutorials.

[6] Decoding the True Cost of Generative AI for Your Enterprise, August 17th, 2023, Maryam Ashoori, PhD.

[7] Ibid.

[8] What Large Models Cost You – There Is No Free AI Lunch, September 8th, 2023, Forbes.

[9] OpenAI's GPT-3 Language Model: A Technical Overview, June 3rd, 2020, Lambda Labs.

[10] Research Chat GPT-4, March 14th, 2023, OpenAI.

[11] 2023 AI Index Report, 2023, Stanford University Human-Centered Artificial Intelligence.

[12] State of AI Report, Slide 115, October 12th, 2023, Air Street Capital.

[13] Artificial General Intelligence will save humanity, February 5th, 2022, Julia Kipperman and Nedim Azar.

[14] Google Will Keep Reading Your Emails, Just Not for Ads, June 23rd, 2017, Variety.

[15] Liberal, Moderate or Conservative? See How Facebook Labels You, August 23rd,2016, The New York Times.

[16] Investigation: How TikTok's Algorithm Figures Out Your Deepest Desires, July 21st, 2021, The Wall Street Journal.

This post is still under construction!

I have reached the stage of life where every once in a while a college or high school student in my network will reach out to me and ask about my pathway to graduate school. I wasn't smart enough to do that when I was applying or thinking about it - I mostly relied on talking to my friends who were also going through the process. It wasn't quite the blind leading the blind (two of my best friends from college got perfect LSATs and went on to Harvard Law), but your fellow 20-year olds only know so much of what goes on behind the curtain. I can say with confidence that I know a lot more about graduate school now than I did from the "outside", and there is often a lot of contradictory and conflicting advice floating around out there - whether from that one uncle of yours or a random reddit post. My response? To add to the cacophony, but hopefully with insight that is a little more fresh and marginally more helpful. I try to make it a habit to take every meeting request from interested students / applicants that reach out, and decided to write down notes to hopefully provide helpful context for future conversations. But, just remember that this advice should be taken with a grain of salt. I always loved the line "It's worth what you paid for it - $0." 

For those who may be missing context, I attended and (through God's grace) graduated from the University of Virginia where I earned both my Juris Doctor and Master's in Business Administration. Those four years in Charlottesville, Virginia contain so many wonderful memories, beneficial but difficult times of struggle and hardship, and lessons learned that all the writing I could do on that short time wouldn't fit into a single essay (at least not one of a tolerable length).  There are many topics about graduate school that I could write about (the admissions process, the specifics behind choosing to get a JD/MBA, tips for how to do well in grad school, and insights into full-time recruiting), but to start I will focus specifically on the admissions process as it is most broadly applicable to the conversations I have recently been having.

Applying: Do you know why you want to go?

They say that there are three reasons not to go to graduate school. (1) Your friends are doing it. (2) Your parents want you do it. And (3) You have nothing better to do. This will be relevant later on.

So there I was, a fresh-faced, energetic 21-year old staring down the barrel of graduating from college with no clue what to do with my life. I remember looking at the people around me and trying to pick out what options I had in front of me. I had gone to some consulting informational sessions - but no luck getting an interview. I could return to the non-profit I had worked at during one of the previous summers which had been a great learning opportunity, but I had a nagging suspicion they wouldn't pay me enough to live on. I also took the Foreign Service exam for fun (it was free and my former and current roommates at the time had mentioned they were going to sit for it so I decided to tag along). I actually made it past the multiple choice test before flaming out on the "experience essays" (What relevant experience did I have to be shaping American foreign policy as a 21-year old? Not much as it turns out). But then I learned a whole host of my friends and classmates were doing this thing where they were applying to law school - and it turns out that more than a few of them had been admitted and were going to very prestigious law schools that sounded cool and impressive. And on top of that, jobs out of those schools supposedly paid very well! With my parents approval, the chance for "cool points" with my friends and classmates, and a hefty dose of "Well, this beats everything else," I merrily set off on this track. Note the three reasons listed above which are not reasons to go to graduate school.

To be fair to myself, those weren't the only reasons I decided to apply. Attending law or business school was also the most consistent experience shared among all the people whose Wikipedia pages I read that I thought they had sick careers.

It sounded like the perfect way to kick the can of making a difficult career choice down the road, so I took some practice LSATs (not nearly enough), and sat for the exam. All that to say - there are still many reasons to go to graduate school (which I will outline in a few paragraphs). And if I could go back in time I wouldn't change my decision. A broken clock is still right twice a day, I just wish I made the right decision for the right reasons, and not just because I got lucky / had providence on my side.

I ended up scoring decently well on my first go (a 167/180 which was approximately the 93rd percentile), but for the schools I was targeting this put me well below median. My grades weren't as competitive (I had taken some exploratory econ classes early in my collegiate academic career that did no favors for my GPA) - and I knew I needed a good LSAT score to even get looked at by the admissions officers. At the end of the day, the admissions process is a bit of a numbers game as schools compete to have high GPAs and LSAT scores for their incoming class as this is a major factor in their U.S. News and World Report rankings. As such, being "above median" for the school's previous year's admission class's GPA and LSAT scores is considered a good indicator of your chances of getting in. 

A brief aside here, law schools and lawyers are unhealthily obsessed with prestige, with the top 14 schools in the rankings, (colloquially referred to as the "T14"), being viewed as a golden ticket to high-paying jobs at even more prestigious law firms that work with the most elite clients on the most difficult issues (which is often just making sure that Wall Street financial transactions meet their regulatory due diligence requirements). Needless to say, I've realized that it's not where you go, it's who you are and what you do.

I studied a bit more, retook the exam, and with a score that put me right at or above median for the schools I was targeting (I had totally fallen for the prestige trap), I decided to throw my hat in the ring. One other important thing to note about the law school admissions process - applications are accepted on a rolling basis. This means that most schools begin accepting applications in the fall with a submission deadline in the springtime of the following year (some closing at the end of January, others in February, and so on). If you weren't trying to retake the LSAT in the spring for a last chance to bump your score before submitting your packet, you would have to be a fool to wait until the deadline for each school to apply because this would significantly reduce your chances of admission or receiving a scholarship if you were admitted (at the end of the admissions cycle, spots will have filled up and scholarships will have been given out - and these are both in very limited supply). Cue the clown music and entry from stage left for good ol' Pete.

Now, it's not quite that bad. I fully expected with my applicant profile (below median GPA, moderately competitive LSAT, and some good career experiences but nothing like curing cancer or winning a gold medal at the Olympics) that I wouldn't get in that particular cycle and my best shot would be to take two years, gain some relevant work experience, retake the LSAT again after studying more seriously, and then re-apply. With that in the back of my head, I thought it would be best to take my time and work on creating the most compelling applicant packet possible as I could reuse most of it in a few years - and therefore I didn't feel any particular rush to submit. This resulted in me sending in my collection of essays, test scores, resumes, and letters of recommendation right at the deadline for pretty much every school I applied to. If you are interested in going to law school,

Do. Not. Do. That.

It worked out for me (one of the friends who ended up at Harvard law that I mentioned earlier would jokingly call me the "unicorn of the cycle"), but past results are not indicative of future performance. I ended up wait-listed at two schools and admitted to two others with comparatively equivalent small amounts of scholarship money (an all too familiar experience from my college admissions journey, but that's a story for another post), with one of them obviously being the University of Virginia. And I do have a small amount of pride in the fact that it was mostly a "do-it-yourself" operation. I had some of my friends look over my essays and of course relied on my mentors' support for their letters of recommendation, but I didn't pay for an admissions consultant or call in any favors from relatives or my parents' friends to boost my application (not that we had anyone in our personal networks that I could have asked even if I wanted to). That being said, a better-crafted application strategy may have resulted in more admissions or more scholarship money - and the money does matter. As of 2024, attending graduate school can set you back well over $250,000. 

Choosing the right school for you.

I was now faced with a choice between UVA and an unnamed Ivy League school of comparable ranking but slightly more "lay prestige" (a term used in the law school admissions online forums meaning a school has a more elite reputation among the general population). Obviously to the vast majority of people it doesn't and shouldn't matter what specific program you went to. On top of that, the quality of education and career outcomes are almost indistinguishable once you reach a certain level of school.

Not sure how to decide, I asked the admissions offices of both programs if they could put me in touch with current students in order to conduct some personal reconnaissance. They kindly obliged, and I went on to speak with a host of students from both schools. Almost every call that I had with the Ivy League students ended with them expressing the following, in so many words, "It's really hard here. I don't really enjoy it. But you get a good job at the end of it!" Meanwhile, the UVA students would say, "These have been the best three years of my life! I love it here and have made some of my closest friends!" 

Ok. That definitely put one school ahead. So I went back to the Ivy League students and asked, "What do you actually like about your law school?" The answer sealed the deal. "Every year we get to go to UVA to play in their law school softball tournament." 

Obviously a little more thought went into the decision, and I was particularly influenced by a conversation with a specific alum who had graduated from both UVA's law and business programs. We hit it off in our conversation as we discussed the role his faith and values had played in his graduate school experience, and his belief that UVA was the best school to go to in the country for someone of this background. The connection I felt with the student culture and community, knowing that I would have a good shot at admissions to the MBA program, and the strong ties to Virginia, a state I have grown to love deeply, all helped me make my ultimate decision to attend UVA - and I don't regret it for a second. My one word of advice here is to go to the school where you feel most at home - especially if costs work out to be around the same. The outcomes are fairly similar across the T14, and so the biggest factor in your success will be finding a school where you can become your best self and invest in the community - and that won't be the same school for every person.

The Meat

Ok, more philosophical and narrative writing aside - here are some very tactical application tips which will probably be much more useful for you.

(1) Live an interesting life.

(2) Develop real mentorships.

(3) Maximize your GPA and LSAT.

(4) Invest in the relationship (with your target school). The applicant pools are small enough that it is still possible to stand out. Just don't be "too much".

12.23.2024

I was recently scrolling through Instagram reels when an interesting podcast interview began to play. My algorithm is built such that I get a decent bit of "hustle" content and although a lot of it is filled with bunk "Here's how to get rich on real estate" videos, occasionally some nuggets of wisdom find their way onto my screen. In this case, I was watching Pillpack founder TJ Parker on the Logan Bartlett Show as he explained how he envisioned the role of CEO: 

“Now when I think about what the job of a CEO is, it's to set the vision and have a really clear perspective on what you're doing. It's to capitalize the business which is incredibly important. Mine is to be the external voice of the business - those things are obvious, I think you have to be world class at those, no one else in the company can do those things. And then internally, like, for me it was finding the right exec team, putting them in place, and making sure they got along, rather than trying to make day-to-day decisions at a granular level or even like a relatively strategic level. It was way more productive to find great people that could do that themselves with a lot of autonomy than it was for me to try to dive in and make those decisions.”

This drove the question into me: what actually are the key responsibilities and attitudes of an effective CEO? I began building out a framework, and then started comparing notes. It just so happened that I was hearing from several CEOs as guest speakers in a variety of my classes at both the law and business schools and so I paid close attention to what they said about leadership. I've shared some of their thoughts here (paraphrased for clarity).

One CEO who was running a startup in the consumer products space had this to share:

• It was critical to shape the vision and strategy, and to bring money into the company. At first I had to do everything because it was a small company, but to be honest, I suck at operations. I needed to hire good people which allowed me to fire myself from the roles I'm bad at. On top of that, it was important for me to identify all the things that make our company successful, segment those roles, and build the roles that solve for each one of the problems that will allow us to be successful. This wasn't arbitrary hiring, but being thoughtful about what are the target milestones and thinking who are the people  I need to onboard to move in that direction.

Another CEO, the founder of a Fortune 100 company turned venture capitalist, built his message around the following:

• Leadership is about staying true to yourself and getting the best out of your people. As a leader, ask yourself the questions: "What do you stand for and why should anyone follow you?" Don't take it for granted that people will follow you because you have a good title, prestigious degrees, or make money. Why should anyone trust you? I want to surround myself with people that are more able than me. Finding spikey people who have amazing talent and joining them together allows you to make 2+2+2=7. One thing that allowed me to succeed has been having the ability to recognizer superb talent around me and get them to work together. Surround yourself with stars. People don't work for you because you're the cleverest. People want to work with you because of what you can do to help them and grow them. It's often not about money. It's about answering the questions for them, "How can you help me grow? How do you help me develop?" You ultimately want to be customer-centric not just in your products, but also when interacting with the people you hire and manage.

In my own life, I have been exploring and applying styles of leadership in several of my organizations, but primarily in my startup and nonprofit. Acting as CEO and President, respectively, I have had to manage teams, coordinate strategy, capitalize / attempt to fundraise, etc. And I have been trying to figure out how to optimize my efforts to empower these entities to achieve their ultimate visions. 

In my non-profit, which is a national organization designed to partner with and support Korean American student associations at colleges around the country, we have three primary departments: External Affairs, Internal Affairs, and Public Affairs. These are supported by an administrative team composed of myself, the Vice President, and Treasurer (think CEO, COO, and CFO). 

The first windfall for the organization was bringing on my Vice President / COO - a West Point alum and current military officer - and someone who supplemented me in areas where I fall short. I'm into vision setting, culture building, and high-level strategy. He's an expert at developing operational infrastructure and implementing on a day-to-day level. I am great at getting people together in a room around a shared mission. He excels at creating the systems through which they can identify and accomplish critical tasks. Working with him has really shown me the power of effective partnerships and proven to me the benefit of bringing on team-members who have comparative advantages (see my Building Teams post below). And it was he who came up with idea to add some structure to the organization through the key departments. We had been operating for two years in kind of a flat system with tasks delegated ad hoc when he came in and asked the simple question: Why don't we have specific roles and responsibilities? When we realized we needed this, we were off to the races. 

The second windfall, or key moment that I am currently evaluating, actually took place in the last few weeks. We had been meeting with the Department Directors weekly to review their ongoing tasks, quarterly goals, and standing action items and questions - but I was unhappy with the progress. I didn't feel like the team was taking initiative and that I was needing to micromanage and pull teeth to get stuff done. So I decided to sit down one-on-one with these Directors and collect information about what was going on. When I heard from the Director of External Affairs that his team had mentioned to him they were being overloaded with administrative work that they didn't feel added value and it left them demoralized I was shocked. Of course I didn't want our team-members to be wasting their time on that clerical work - it would be helpful down the line but it wasn't mission critical compared to some of the other tasks that were assigned to the department. The External Affairs Department was tasked with recruiting schools, non-profit partners, and corporate sponsors - without them, we wouldn't have a reason to exist.

I took a pause on the call and thought about how I could align our team on this. After some silence, I asked the Director, "What do you think your job is?" He looked a little confused so I continued with a tired smile, "It's not a trick question, what do you think you are responsible for?" He sat and thought for a while before beginning to list it out some of the tasks that I had assigned to him and his team. This was where I realized I had made a big mistake within our organization. I stopped him - "Those are all things that need to get done in your department. But at the big level - your job is to accomplish the mission of your department with the assets that you have. However you do that, it's up to you. You have total ownership as Director to utilize your team as you see fit. It's not my job to tell your team what to do. But you need to know what you're solving for." As we sat and brainstormed about what exactly the mission of the External Affairs Department was, we came to the sentence: "Building, developing, and maintaining critical partnerships." We both felt good about it, but I knew I needed to recalibrate away from the status quo of over-managing. I left with the final thought: "My job is not to come up with the solution. My job is to remove roadblocks to the solution you identify to accomplish the mission." 

I'm still trying to figure out if that was the right call / framing. A good mentor told me that it's not just the locker-room speech, it's also the day-to-day implementation and support that make implementing that kind of ownership system possible. I know it'll take a lot of work on my part to fill the role in the right way - and I would appreciate any advice you might have on this (so shoot me an email if you have any wisdom to spare).

But circling back to the question that started this whole thing: what actually are the key responsibilities and attitudes of an effective CEO? So far, my thoughts are centered around the following:

• Capitalize the Business: The CEO must ensure that the business has the necessary working capital to continuously operate. This can come from external funding (debt, outside investment, non-dilutive funding), or through the management of operational, financial, or investment cash flows in conjunction with the company's balance sheet.

• Coordinate the Team: Effective CEOs will hire talent aligned with the 5 C's (detailed below) and create operational structures such that the team members are able to effectively identify their target objectives and accomplish them. Being a coordinator here seems similar to an orchestra conductor (which ties into the next point).

• Develop the Company's Competitive Strategy: CEOs must be able to guide the company to survive the quarter, but also build towards a bigger vision within the competitive landscape, leveraging the talents of the team to solve critical problems.

• Effectively Communicate Internally and Externally: As the internal and external face of the company, strong CEOs will be able to express the vision, mission, and values of the company in an understandable and persuasive manner. Internally, they build team alignment and contribute to the company culture. Externally, they build public understanding of the company and shape the company's brand. 

What do you think?

9.19.2023

I've always enjoyed designing processes and systems. At the first meeting of my business school learning team, a semi-mandatory assigned study group, I forced us to sit down and talk about what our goals were. Not in a deep personal sense, but more explicitly what we hoped to get out of the learning team itself. What followed was an intense two-hour meeting where folks openly, and sometimes bluntly, shared what they hoped to get out of their MBA and this study group. For some, our team was an opportunity to make friends and deep relationships, for others they just wanted help with the homework. By level-setting and communicating up front what we hoped to achieve, we were able to establish team norms and develop reasonable expectations along with an efficient and fair work allocation process. Sitting down and creating a space where we could figure out how we could best operate as a team was both a fun and fulfilling experience for me. And the team ultimately came together and did an incredible job at achieving our agreed-upon goals.

Looking back on this experience in working with my learning team, I was reminded of my first serious internship. I was working directly under the CEO of a non-profit that operated very much like a startup. Over the course of the summer, I had observed as the CEO had managed difficult personalities and leveraged members of his teams to effectively move the ball forward on achieving key objectives. As I sat back and tried to make sense of the decisions he had made, I began imagining myself in his position and wrote in my notebook what I thought were the three most important characteristics a person had to have when considering adding them to a team: care, competence, and comparative advantages.

• 1. Care: Does this person care about their job, the work they do, and the people around them?

• 2. Competence: Is this person qualified to do the job they have been assigned? Do they possess the proper skills needed to succeed (or to learn how to succeed)?

• 3. Comparative Advantages: Does this person bring unique skills and perspectives to the team?

Since that initial list I've added two additional characteristics: curiosity, and compatibility. The first was inspired by a lecture from Doug Lebda, a Darden alum and the founder of LendingTree (whose team criteria were: will, skill, and deliver). The second came from my Leading Teams class taught by Gabe Adams. 

• 4. Curiosity: Is this person a life-long learner? Are they interested in developing themselves and exploring new and interesting concepts and skills?

• 5. Compatibility / Cooperation: Does this person get along well with the existing team? Can they "play well with others" and collaboratively support the team in achieving its overarching objectives? Can they bring the team together around a vision and lead/follow when necessary?

Most people will not meet all five categories. And certain categories are easier to develop than others. For example, a young student or employee might care a lot, have curiosity, and be compatible with a team, but due to their inexperience they may lack task-specific competence or comparative advantages. This can be overcome with training and development support. Noting the difficulty of finding superstar team-members, another guest speaker at Darden had this to say: "Most of the time, you will not be working with a team of A players. The best managers learn how to get the best they can out of their B and C players."

The issue with lists like these is that they always tend to grow (or you forget to include something because it didn't start with the right letter). However, I'm confident that if you find someone that matches all five C's that they would be a great addition to a team. Ultimately, being caring, competent, recognizing one's comparative advantages, having compatibility with one's teammates, and being curious can help foster a positive team dynamic and improve the chances of achieving the team's objectives. 

4.10.2023

"You should write a book." It's not something a lot of 20-year-olds hear - and probably for good reason. Most 20-somethings don't really know enough about anything to be able to create a compelling multi-hundred page piece of writing. But that didn't dissuade my college professor in my sophomore year from standing in front of the class and telling us exactly that. His thesis was pretty simple: even though we weren't exactly experts, we could interview several of them and compile their wisdom in one convenient source. Our value add to the readers was  effective synthesis and analysis of experts' knowledge following our interviews with them. It capitalized on the one resource that we had in spades as students - free time. That and the unique access that students can have to very "important" people - when you email someone and say: "I'm a college kid writing a book, and would like to interview you as an expert in the field," it opens a surprising number of doors. People like to feel like they're helping students, but they also like the idea that the networking call that they're taking is more than just throwing advice at a kid and hoping that they actually find it useful. 

The value to us as writers was two-fold: at the end of the project, you're left with an actual product that you created and a new network of experts that you (hopefully) admire, learned from, and built a connection with.

I've taken that idea and applied it to many different projects in my life. When I knew I wanted to learn more about Asian American political organizing,  I started a weekly discussion group and used it as a tool to invite speakers who were established professionals in the space. That grew into Young Asian Pacific American Leaders and led to me moderating a conversation with two of my heroes - Representatives Young Kim and Michelle Steel. This was also part of the impetus behind the Korean American Student Internships Database which I worked on in the Spring of 2019. I had created this resource originally to help with my own internship hunt, but realized that once I had a job - why not share it? When I sought to expand it, I was able to speak with a ton of great business leaders and mentors who helped me collect more resources to add and ultimately share with Korean college students. This was the seed that then grew into the National Korean Student Alliance - which is a project I'm still working on to this day!

The beauty of project based networking is that you not only get the chance to engage with leaders and experts you might not have had access to otherwise, you're also left with something you created that hopefully brings value into the world. Whether that's a book, a discussion group, a blog, a podcast, or anything else, it feels good to go out and create. In a time where we are obsessed with consumption, including of our mentors' time - to create sets you apart, and provides fulfillment. Additionally, the creation can continue to live on and have value long after you first finished it. I know of friends who still make some money from their books off of Amazon sales, or who used it to gain entry into an industry that became an important part of their career. Folks I met through YAPAL have gone on to become close personal friends. The NKSA is still working to help Korean college students (and giving me more writing fodder as we struggle to build an effective and meaningfully impactful organization).

I never ended up publishing my manuscript - it wasn't of a quality that I felt comfortable publicly releasing and attaching my name to, especially because it was political in nature (and probably a little immature to be honest). My college friends to this day like to jokingly ask "Has your book come out yet?" I'm happy to grin and bear it, because while I didn't end that class with a book, I did leave with a tool that I still use to create and to learn from and engage with important people in my life.

5.1.2022

We all have role models in our lives. For many it’s an older family member, former boss, or professor. These role models often have proven their credibility because of the success they’ve found in their personal or professional lives. But, due to the nature of having proven their success, they are often much older. They were in school or applied for their first job years, if not decades, ago. 

We should embrace looking for peer role models.  In the traditional mentorship model, we are often told to look for someone who has found success and is in their mid- to late-career and to ask them to take us under their wing. One issue with this framework is that these workers might not know the best tips and tricks to help you navigate the unique intricacies of the modern job market. Peer mentors can serve as a great supplement to mentor relationships you have already developed.

I remember at one point in my sophomore year I noticed that some of my friends were finding great professional and academic success while I was struggling. I was networking, pursuing internships, and working hard at school, but these folks seemed to be gliding through all of these and effortlessly crushing the game. Obviously, part of this is the duck effect, where someone appears calm and composed on the surface but in reality is treading incredibly hard underwater to keep moving forward. But, another part was that these friends had positive habits and professional strategies that helped them be competent, reliable, and hard-working. I decided that these peers were at a place I wanted to be and so I began looking at the habits that seemed to help, and began copying them. 

The clearest example that comes to mind is one friend who I had great respect for. While the rest of us looked like partying 20-year-olds (which we were), he was helping run a company and took great care and interest in his family. I noticed he carried around a notebook with him everywhere, and often jotted down thoughts and notes as we were in club meetings or conversations together. My thought process was simple - this guy is killing it, it seems that carrying a notebook is a habit that helps him pursue success, I want to be more like him, I should get a notebook. From that point on I carried a notebook with me and began noting things down and keeping track of my to-dos and thoughts. It helped me get more organized and put me on a path towards building a more efficient and healthy lifestyle. It goes to show how sometimes you have no idea how a little habit can have big impacts.

There is a famous saying: “You become the average of the 5 people you spend the most time with.” As someone who has been blessed with incredible friends, many of whom have found success in both their personal and professional lives, I’d like to think that this is true. I was lucky that within the first week of college I met a great group of guys who joined the "International Relations Club". These guys have landed positions at elite consulting and banking firms, prestigious roles in the public sector and nonprofit spaces, with another heading off to law school, all immediately after graduating. Another group of friends have found great success in the political space, with several holding government appointments or founding rapidly growing nonprofits. We met at conferences, leadership programs, and informal networking events over the years, but were drawn together by our shared professional interests and the simple fact that we got along as friends.

Each person within these groups of friends has provided invaluable insight to me at one point or another, whether that was teaching me how to send out a cold email, helping edit my grad school applications, or showing me the ropes of starting a nonprofit. Unlike a more traditional mentor where I was concerned about wasting their time with simplistic questions, my buddy and I could edit each others emails, resumes, or essays for hours without having to worry about taking up the other’s time - if we weren’t working we’d probably end up watching a TV show or going out for drinks together. Additionally, my friends could provide greater insight as to what professor to take, or which job applications to keep an eye out for, because they were going through the same process I was. While I could turn to a more traditional mentor for broader career advice and insight, it was my friends who could give me real, actionable, and tactical advice. That's not to say a traditional mentor couldn't do that, but I was particularly helped and influence by my peers.

Too often we leave our friendships totally up to chance. Surround yourself with hard-working, intelligent, driven people and you will find yourself becoming better in all that you do. Iron sharpens iron. And don't be afraid to ask your friends for help - they might share a little habit that could have a big impact on your life.

12.23.2020

In my fourth quarter at Darden I gathered a small crew of "willing" classmates for weekly meetings where we would brainstorm and gut-check each other with ideas about our careers and support each other as we began our professional careers. We called ourselves the "Business Accountability Group" aka BAG.

The meetings were structured such that each week a new member would be responsible for determining the topic and format of the discussion of the day. As the weeks went on we shared about our long-term visions for our careers, analyzed our strengths and weaknesses, and traded tips and tricks about how to best manage both personal and professional relationships. I am writing this after being inspired by one of those meetings where the structure, set by a classmate named Nick, was as follows: each person was to present their five- and ten-year goals and the other members were supposed to try to poke holes in the plan in order to find weaknesses or areas for development. I watched and participated as we went around the room - impressed by the goals my friends had set for themselves and happy to try to help them figure out if they had left any blind spots that would be helpful to address.

When it was my turn I had to say, "Sorry guys, but I don't really have any set goals for the next five to ten years." The words felt strange coming out of my mouth, because if you know me at all, you know that for a long time I was obsessed with setting goals and using those as a tool to promote my personal development. As I was thinking about the prompt though, I realized that while my brain had been filled with goals for a long time -- for the first time in my life, it was fairly empty. And I was comfortable with that. 

Instead of prioritizing setting goals (which are still an important tool), my new theory is that it is more important to focus on process implementation. Every time a speaker has come into class at Darden this year, my question from the audience was the same, "What daily habits do you have that you think contributed to your success?" I firmly believe that how we spend our time and who we spend it with shapes who we are. The small habits that make up your daily routine will add up over the course of your life to build you into the person you become. And so instead of saying "I want job x" or "I want to have y thing" I think it would be interesting to think about what kind of life we want to live, and what habits we need to implement in our daily life to achieve that state of being. The fruits will follow if you plant well.

When I first developed my personal mindfulness exercise it began by having you list out your goals in ten key categories: physical health, mental health, spiritual health, intellectual curiosity, professional development, social relationships, community engagement, financial stability, creativity, and adventure. I'm sure there are more categories that would be good to include, or that some of these categories could be bundled up. But I'm also confident that if you can be growing in all ten of these categories that life is going pretty well. Obviously, in life there are seasons where some categories will take precedence over others. But I posit that a balance across these categories can leave a person with fulfillment. Now, instead of just writing down goals for each of these categories, I want to also implement some processes that allow me to sustainably improve in each of the ten areas ("Sustainable Improvement Processes" - see what I did there?) .  To provide an example, a goal is: I want to lose weight. A SIP is: I will go on a 30-minute walk five times a week. Processes dive deeper into actual implementation of behavior that achieves goals.

What really drove this home for me was sitting in that room with my classmates and sharing these thoughts when it hit me - what I'm doing right now is a SIP. By gathering a group of my friends who I respected and wanted to develop a relationship with in a room every week, I had created a process through which I could improve social relationships that were important to me, provide energy and attention to my professional development, and engage in conversations that inspired further intellectual curiosity.  And I want to be thoughtful about the processes I am building in my life - because frankly, whether you want to or not, you are building out the processes which will shape your approach to life. I personally like the idea of building out processes because that's just how my brain interacts with the world. I also want to be cautious of over-analyzing and structuring things to death. The world also needs to exist in creative, unplanned spaces. But even though I like to identify and build out frameworks for all the aspects of my life, it might not be that way for you and that's ok. 

That night, I came across an interesting video from Mark Manson (language warning), who wrote The Subtle Art of Not Giving a F***, where he talks about the concept of self-help. Here are some key excerpts:

"There's nothing new in self-help... Most of the advice that we're consuming when we buy these books, whether it's on how to be more compassionate, less anxious, more humble and grounded, more productive, more honest, and vulnerable. These things have been covered for thousands of  f****** years from Buddha and Jesus, to Plato and Seneca, to Adam Smith and Benjamin f***** Franklin. Pretty much nothing you find on a self-help shelf is new. And that's fine because what does change, and what does matter is the packaging... Self help ideas are simple but difficult. I.e. they are easy to understand, but really f***** difficult to actually go do... This is where the packaging comes in. The way ideas are packaged can do a lot to alleviate the emotional difficulty of these problems in the short run."

So for me, the way I'm packaging and then internalizing (and hopefully externalizing) the wisdom and lessons that I've learned through my faith, my family, and my education, is through these Sustainable Improvement Processes. Over the course of this summer, I hope to explore each of category I identified in my first mindfulness exercise and develop SIPs in each area. And hopefully this will help me lead a good life.

5.4.2023

• Growing through Pivots

• Social / Civic Entrepreneurship

• The Theory of the Conscious Guess

• Wanting to be Heard

• Consumer Surplus Pricing

under construction...