text
stringlengths 0
2.12k
|
|---|
Nine years ago, on the day before Father's Day, I was scrambling to find a last-minute card. I went to every pharmacy in town but didn't find anything Dad-worthy. With time winding down and few options left, I took things into my own hands and made a card myself. I called it: the iDad, an iPad-shaped card with handwritten notes hidden beneath app icons. My Dad loved the card (in a way that all parents love their kid's work) and I loved it too. Since then, I have been handmaking holiday cards for family members every year, and my love for being creative continues to grow. |
I want to design products, brands, and experiences that people love with Wieden+Kennedy. Instead of paper and scissors, I'm now using artificial intelligence, machine learning, virtual reality, UI/UX research, and other, more advanced tools, to create. However, my goal of having a positive impact on the end user of my product has remained the same. Wieden+Kennedy offers a forward-thinking environment that prides itself on incubating creative ideas and brands. This influential, free-spirited climate is exactly what I need to continue to do my best work and effect change. |
I have the technical skills. Last year, public sector consulting firm Marts & Lundy asked me to help them answer a simple question: "How do we know who to direct fundraising resources towards in higher education?" In response, I built a neural network to predict how much money candidates are likely to donate to colleges and universities. With little guidance, I networked my way to a massive philanthropy dataset and used it to fuel a five-layer neural network that I programmed using Python and TensorFlow. In the end, the model predicted candidates' donations with ~80% accuracy and helped Marts & Lundy's consultants conserve significant outreach resources. |
I have the creative mind. This summer, as a Technology Consultant Intern with EY, I created one of the first auto dealerships in the metaverse. My team's client – an online auto retailer – wanted to explore the viability of expanding into the metaverse. To field their vision, I assessed the business case and built a proof-of-concept. I taught myself to use Unity and program in C# and created a 3D world that my internal team and our client loved. I devised novel UI/UX schema like custom search tools and first-person experiences, all within a five-week timeline. The model is now being turned into a full-scale platform. |
I belong in a creative space like Wieden+Kennedy. You are looking for people who can take a theoretical or practical problem, identify the steps that need to be taken and the people that need to be involved to solve that problem, and then visualize, prototype, and execute that solution. My technical skills and creative nature have prepared me to do just that: assess, solve, envision, and build. I'm really excited to discuss my fit for a role with Wieden+Kennedy and appreciate you reviewing my application. |
For the past five years, I've been advocating for intelligent applications. When I noticed that there weren't enough AI resources available in layperson's terms, I became a newsletter columnist to make AI transparent for more than 55,000 teen readers. When I hadn't yet applied ML to solve real-world problems, I built a neural network for a consulting firm to predict donations and save outreach resources. When I wanted to apply ML to an industry I'm passionate about, I developed algorithms to forecast viral music artists for a Big 4 record label. I see software as a medium for problem-solving. |
Now, I want to continue my journey as a Scientific Applications Software Engineer with the JPL. I'm ready to begin solving bigger problems in software and NASA is the right place for me to do it. As a Cognitive Science Major and Computer Science Minor, I'm used to taking scientific knowledge and theories about the brain and applying it to build intelligent applications. I want to expand this horizon (quite literally) by translating knowledge about the solar system into software and, more importantly, turning intelligent software into unprecedented knowledge about the solar system. |
I have the technical skills. Last winter, I helped optimize Atlantic Records' marketing campaigns with algorithms predicting viral songs and artists on social media. Atlantic wanted to know which of their content was most likely to go viral so they could get an early jump on promoting it. In response, I scraped TikTok for data about Atlantic's music and used it to train a time series forecasting model. I then packaged my program into an ETL pipeline that autonomously scraped TikTok, generated predictions, and delivered a 'viral forecast' via email. The program ultimately helped Atlantic streamline marketing campaigns and save marketing resources. |
I have the ambition. Last year, public sector consulting firm Marts & Lundy asked me to help them answer the question: "How do we know who to direct fundraising resources towards in higher education?" I interviewed Marts & Lundy's consultants and learned that they needed a simple tool to inform how to optimize outreach campaigns. With those key stakeholders in mind, I built a model to predict how much money candidates are likely to donate to colleges and universities. I secured a philanthropy dataset for free and used it to fuel a five-layer neural network that I programmed using Python and TensorFlow. In the end, the model predicted 95% of candidates' donations within $250 and is being used to help Marts & Lundy's consultants conserve outreach resources. |
You are looking for a candidate who can take a theoretical or practical problem, identify a computational approach to solve that problem, and then help model, program, and execute a solution. My technical skills and ambitious nature have prepared me to do just that: assess, solve, envision, and build. I'm really excited to discuss my fit with NASA and appreciate you reviewing my application. |
The 2020 US presidential election was riddled with threats to democracy like alleged voter fraud and refusal to commit a peaceful transfer of power. Now, we’re faced with the task of restoring confidence in our democratic institutions and preserving their integrity. While AI is not yet the crux of political agendas, it’s already being leveraged from campaign trails to Capitol Hill. Let’s talk about how governments can (and can’t) use AI to promote democratic values. |
At first glance, the thought of giving an infant technology a government role sounds crazy. If calculating a nation’s population size requires human census-takers marching door to door, then how are we ready for AI to advise our elected officials? The truth is, we’ve been ready for census bots and AI advisors for a number of years, but IT investment in the public sector is extremely scarce. A recent Microsoft study found that two-thirds of public sector organizations see AI as a digital priority, yet only four percent have the resources necessary to scale AI and achieve transformative outcomes. Still, funding is on the horizon. The federal government invested $4.9 billion in AI-related research and development in FY 2020, more than double 2019’s investment. If this trend continues, your favorite AI bot may soon have a Senate seat. |
No Automation Without Representation |
Perhaps the most promising application of AI in democratic governments is to improve representation. AI will strengthen the relationship between officials and their constituents by streamlining communication and gauging the interests of the constituency. Natural Language Processing (NLP) is the AI approach used to analyze human language and sentiment, and some governments are already starting to experiment with it. |
The European Commission’s democratic system depends on a stream of communication between the legislative and executive branches. Often, parliamentary representatives ask questions of the executive branch, which are responded to by ministers and civil servants. But each year, it costs the European Commission nearly €8 million to facilitate all the Q&A. To save time and money, the Data Science Hub in the UK’s Ministry of Justice (MoJ) developed an AI-driven model that helps the MoJ respond to parliamentary questions. Using a NLP technique called “latent semantic analysis,” the model analyzes a question’s proximity to previously asked parliamentary questions and pulls those responses from the MoJ’s database. Less background research is needed, relevant reports and resources are pinpointed, and representatives get better feedback from the executive branch. |
Similarly, the Singaporean government’s Housing and Development Board receives a large number of emails from residents raising concerns and queries. Over time, the Board noticed that there may be some underlying trends among the emails they receive. So, they built an AI text-analyzer that evaluated over 100,000 emails and found that demographic was strongly correlated with a resident’s needs. The Board learned that, for instance, young homeowners sought to collect their keys earlier than old homeowners, so the Board moved away from giving prespecified appointment dates for key collection. The Board now tailors their services by demographic. |
Explainability is the Best Policy |
Better communication leads directly to better policy-making; when it’s easier to hear and understand the constituency, it’s easier to make policies that benefit them. The Singaporean Housing and Development Board’s key collection service is a prime example. Down the road, AI may even be able to draft policies itself by finding optimal solutions to narrowly defined problem-spaces. Envision an AI that prioritizes investments in road work by analyzing traffic bottlenecks or one that tailors social welfare programs by predicting individual needs. But, if we’re going to give AI a voice in policy-making, we need to understand how it reaches any given output. If an AI tells the Secretary of Energy to go entirely nuclear, it better have a good reason for doing so – we can’t just start flinging uranium left and right. Unfortunately, most current AI models are largely black boxes and lack explainability. And, when AI outputs are taken at face value in government, democratic ideals can be jeopardized. |
Allegheny County is located in the state of Pennsylvania and comprises the greater metropolitan area of the city of Pittsburgh. In 2014, Allegheny’s Department of Human Services (DHS) – responsible for child protection services (CPS) – had at their disposal $1.2 million in federal grant money and a warehouse of county residents’ public data. So, the DHS assembled a consortium of techies to leverage their data and improve decision making in the CPS department. The final product: the Allegheny Family Screening Tool (AFST). When a child calls the DHS to report maltreatment, the AFST generates a score from 1 (low-risk) to 20 (high-risk) indicating the likelihood of an adverse event recurring within two years. That score is then used by a human screener to decide whether an investigation needs to be opened or not. Yet, after its launch in 2016, screeners began to notice some idiosyncrasies in the AFST’s output. Upon further investigation, they realized that the AFST’s training data was biased against Black children. Since there exists a tendency to over-report and collect data on children from ethnic minorities, Black children were being scored as high-risk at nearly double the rate that would be expected based on their population size. Without knowing how the AFST was transforming its inputs into outputs, bias was able to creep into the AI model. On a national level, the implications of bias could be exponentially more severe. |
Life, Liberty, and the Pursuit of AI |
AI is not inherently democratic by any means. In fact, it’s quite the opposite. Nothing about an AI’s outputs have to be equitable, representative, or legitimate. Rather, it’s maintaining non-biased inputs and interpreting an AI’s results in an egalitarian way that makes the AI democratic. As such, it’s important to deploy AI in government as a part of a human-in-the-loop system so there is always an egalitarian mind (and an elected official) behind each decision. |
The argument could even be made that AI lends itself to antidemocratic systems of government: AI used for government purposes runs on big data about its constituents, and making that data available for government use may infringe upon (what would in a traditional Democracy be considered) civil liberties. Take China’s panopticon, for example. Xi Jinping can use AI across mass facial recognition systems and Uyghur-monitoring software (read more about the Uyghur Crisis here) because his totalitarian control allows him oversight of civil liberties and boundless access to “private” data. But it’s important for democratic governments to not sacrifice individual rights to fuel AI; instead, AI should conform to – and be used in service of – our individual rights. While the potential for AI to drive transformative solutions is enticing, it should not come at the expense of privacy, freedom, or other democratic values. |
Looking Ahead |
AI in government can improve representation and policy-making but can’t promote democratic ideals if it lacks explainability or violates civil liberties. Developments in public sector AI must be watched closely by relevant constituents as more funding gets directed towards AI projects and COVID-19 pushes more government processes onto digital platforms. Yet, a reality in which AI plays a role in politics is already upon us. In Russia’s 2018 presidential election, AI-driven chatbot “Alice" was nominated against Vladimir Putin and received a couple thousand votes. And, in New Zealand’s 2020 presidential race, AI politico “SAM” put forth a virtual candidate that learns from citizens through interactions on Facebook Messenger. In a public statement, SAM offers a point that perhaps we can all learn from, especially after the 2020 US elections: “Unlike a human politician, I consider everyone’s position when making decisions.” |
All the hype surrounding artificial intelligence (AI) got you bummed out about your human brain? Missing the days when you didn’t have to worry about getting hustled by a poker-playing AI? Last week’s Neuralink demo have you feeling like an army of cyborg pigs are going to take over the world? |
Allow me to offer some solace: the human brain is still the pinnacle of intelligence. While AI may be excelling in some domains, the human brain continues to outperform in others. No, we can’t translate prose into fifty different languages or store an Internet’s-worth of data. But we can empathize, invent, offer opinions, hold values, and write blog posts. Identifying how the human brain succeeds at these tasks is a likely key to further empowering AI. Even if AI fully surpasses human intelligence someday, we can be content knowing that it was born from the processes used by the intelligence engines sitting in our own heads. |
Compared to AI, the human brain is great at abstract reasoning. Presented with little information, we can quickly solve new, unfamiliar problems. All those coordination skills you built up playing Mario Kart can be abstracted from that problem space and applied to make you a better real world driver, even if the only car you’ve ever seen is Toad’s Bullet Bike (backed by science). By transferring knowledge from one domain to another, we streamline the learning process and adapt to our environment. The human brain, however, is not so great at processing information quickly. We have to listen to a song four or five times before it’s memorized, but for AI, memorization is as easy as hitting “Save”. Un-memorizing presents even more of a challenge for humans because, unfortunately, our memories don’t come with “Delete” buttons. Typically, AI will outpace the human brain at any task that involves mathematical calculations or data processing. |
Human intelligence stems from an elegant series of interactions. The human brain is a modular bundle of 86 billion neurons, which are specialized cells that transmit and receive electrical signals. They work harmoniously in a big game of telephone, transforming data as it travels down interweaving neuron chains. An individual neuron is composed of three main parts: dendrites, a cell body, and an axon. Signals are received through the dendrites, travel to the cell body, and continue down the axon until they reach the synapse, the point of communication between two neurons. Signals will only be sent, however, if the neuron reaches a certain voltage threshold, a reaction known as an action potential. At the synapse, the firing of an action potential in one neuron causes the transmission of chemical messengers called neurotransmitters to another. The more neurotransmitters, the stronger the connection. In this way, neurons can talk to thousands of neighbors and stimulate or inhibit their activity, forming circuits that process information and carry out a response. |
The human brain has a distant, artificial cousin: neural networks. This is the type of machine learning we’re seeing used in object recognition systems on self-driving cars and game-playing AIs. Artificial neural networks (ANN) were first proposed in 1944 by University of Chicago researchers Warren McCullough and Walter Pitts in an attempt to show the similarities between the human brain and the digital computer. But since then, ANNs have mostly exposed just how different we are from our cyber counterparts. |
An ANN is a simplified model of the human brain’s structure. It consists of thousands or millions of densely connected nodes that communicate information like neurons. Often, nodes are organized into a number of layers: an initial input layer, one or more hidden layers, and a final output layer. As problems grow more complex, more hidden layers must be introduced to accommodate extra computations, and the network grows “deeper” (this is where the term “deep learning” comes from). Data travels “feed-forward” through an ANN, from a node’s incoming connections to its outgoing connections. At each incoming connection, the node assigns a “weight,” the value of which gets multiplied by the input. The greater the weight, the stronger the connection between nodes. Incoming weight-input products are summed up and passed to the node’s outgoing connections only if the output value is above a certain threshold. When the threshold is exceeded, the node “fires,” and data moves to the next layer. As such, ANN inputs undergo divisions, multiplications, and additions until reaching the output layer with a concise value. |
As previously mentioned, ANNs are superb data-crunchers. They work well in repetitive tasks that have a clearly defined problem space and can be represented by data. This is why so many ANN ventures are happening in the gaming industry: the rules of a game never change, and a game can be broken down mathematically. That said, ANNs fall off the rails when introduced to new subject areas. You can’t take an object-recognition ANN and expect it to know how to play chess — it can’t handle the data inputs of game pieces, let alone understand the concept of winning. But even an eight-year-old kid, with a couple hours and a lot of patience, can learn to play chess with a reasonable degree of skill. |
On paper, the human brain and ANNs sound more like fraternal twins than distant cousins. Their structural similarities are clear: neurons and nodes handle data processing; neurotransmitters and weights manipulate data values; thresholds dictate how data moves through the system. But when we look at what types of problems the human brain and ANNs can solve, we see that they’re worlds apart. Why? |
Imagine you’re handed a computer’s central processing unit (CPU) — the electronic circuitry that executes a computer’s commands — and asked what program that CPU is running. Can you do it? Of course not! Your guess is as good as that of the very guy who invented the CPU. To know what a computer is doing, you can’t just look at its hardware; you need to know what set of instructions, or software, is running through its wires. |
The same goes for the human brain. Thus far, neuroscientists have classified the brain’s architecture as a web of adaptable, interconnected neurons. This model is now being applied to building ANNs with the hope of achieving similar forms of intelligence. But that hope has not been fulfilled because the human brain and ANNs are running entirely different software. In fact, neuroscience has almost no clue what computations are happening in the brain because it’s so difficult to track an individual electrochemical signal in a haystack of 86 billion neurons and one thousand trillion synapses (we don’t even know how the 302-neuron brain of a worm works). So, even though the human brain and ANNs have similar structures, their operating systems are divergent and uncertain. |
To build ANNs with human capacity for abstract thinking, we first need to answer some important questions about the brain’s software, such as: |
How do neurons store information? |
Can consciousness be represented computationally? Is consciousness necessary to achieve human-level intelligence? |
To what extent is intelligence innate versus learned? |
How does a single neuron compute? How do circuits of neurons compute? How does the human brain compute as a whole? |
The ultimate goal of answering these questions is a Universal Theory of Intelligence (UTI), a set of principles that holds true in brain tissue and in metal circuit boards. But we have a long way to go until then. In the meantime, we should look to develop human-in-the-loop AI systems that call on humans for abstract thinking and AI for data processing. This collaboration between brains and bionics is the basis of Elon Musk’s vision for Neuralink, and has also helped produce otherwise impossible music, optimize engineering designs, and diagnose rare diseases. It will be interesting to see what new capabilities arise as AI takes on more and more of the human brain. |
When I first heard about a “novel coronavirus” originating out of China, I wasn’t particularly concerned. I kicked back and waited for artificial intelligence (AI) to make COVID-19 a history lesson. Since the 1918 Spanish Flu, which took some 50 million lives, I had seen AI discover planets, unlock quantum theories, and beat the greatest human minds at our most challenging games. So, I naively assumed that AI could also beat a microscopic virus. As it would turn out, I was wrong. Twenty million people have been infected, there is yet to be a proven vaccine, and I find myself in a post-apocalyptic mask-wearing world. |
I’ve adjusted my expectations: AI is not the cure for COVID-19. Not today, not tomorrow, and not next year. The reason why is that a pandemic involves too much real-world complexity for current AI systems to handle. |
Take contact tracing, for example. Government and pharma’s big idea was that AI could sort out who else a person who contracted COVID-19 may have infected, and then keep the healthy population separate from the sick. But an AI with those capabilities can’t be trained on data from past COVID-19 pandemics because there weren’t any. Since the only available training data is on-the-fly -- most of which is unreliable because up to 80 percent of COVID-19 carriers are asymptomatic -- accurate AI-driven contact tracing is hard to come by. |
A similar challenge arises when using AI to develop a COVID-19 vaccine. AI can traverse thousands of research papers and analyze millions of chemical compounds to identify one or many potentially effective vaccines. However, an AI’s outputs are only conjecture. Until they are rigorously tested in the lab and in human patients (a process that usually takes upwards of 10 years), their output is virtually meaningless. AI insights can guide pharmacology, but today, it’s still humans who are building vaccines. |
Given the problem-solving power of AI, it makes sense to look there first for solutions to our most pressing issues. But AI is not yet where it needs to be to cure COVID-19 on its own. A pandemic is too dynamic and a virus is too uncertain to be captured in an algorithm’s scope. Still, even if it’s not the be-all-end-all, AI can complement human efforts and bring us closer to containing and eradicating COVID-19. Over the past few months, the AI community has rallied to use AI in focused and effective ways. |
On March 16, the White House issued a call to action to AI experts to help scientists answer fundamental questions about COVID-19. To fuel subsequent initiatives, the White House also released the COVID-19 Open Research Dataset (CORD-19), a collection of over 90,000 machine-readable studies on COVID-19. At the time, there was a prevailing information crisis in the research field: COVID-19 literature was being created too quickly for scientists to keep up and avoid making errors or duplicates. After all, 90,000 papers is one hefty summer reading list. Researchers needed some way to filter studies and isolate specific data, and AI was the solution. |
Let’s look at a question like “What is known about COVID-19 risk factors?” Data about how climate affects transmission or how socio-economic status impacts infection rates can be hidden in the margins of any research paper. So, you need an AI that can understand, analyze, and classify large chunks of human language, a capability known as natural language processing (NLP). By encoding the linguistic hierarchies that dictate how words relate to each other, NLPs can deduce context and pick out bits of information from even massive datasets like CORD-19. This is the same technology that allows Amazon Alexa to interpret your verbal commands and Google Translate to decipher your Spanish homework. A NLP can learn to recognize text related to COVID-19 risk factors, such as “pregnancy” or “co-infection,” and then return a concise dataset on patient susceptibility and transmission dynamics for use in policy-making or future research. |
AI experts at Lawrence Berkeley answered the White House’s call with their NLP-driven search engine COVIDScholar. The online tool allows users to comb through CORD-19 by entering basic keywords and applying filters to narrow their search. The Allen Institute for AI developed a similar product, SPIKE-CORD, with the goal of making the power of NLP accessible to those without programming experience. Their search tool allows not just for the retrieval of CORD-19 papers, but also for extraction of information from them, using a simple query language. |
So AI can filter through a dataset and isolate information. That’s helpful, but it still doesn’t leverage AI to its fullest capacity. Rather, AI makes its most far-reaching contributions to the fight against COVID-19 when NLP insights are applied to detecting the virus, limiting its spread, and researching potential vaccines. |
At 3:18 AM on December 31, 2019, Epidemic Intelligence from Open Sources (EIOS) picked up an article release citing a unique cluster of pneumonia cases in Wuhan, China. That report would go on to be the first record of COVID-19, obtained by a NLP that gathers and analyzes data from health monitoring programs and medical databases. With millions of pieces of online medical material, it’s a challenge to find what may be the epicenter of a global pandemic. But the EIOS’s AI identified COVID-19 at its outset and gave policy-makers additional time to prepare and respond. Today, EIOS is informing the World Health Organization of new COVID-19 outbreaks, and may help diagnose and neutralize the next global pandemic before it becomes a pandemic in the first place. |
Other organizations are examining ways to use AI to limit the spread of COVID-19, particularly among vulnerable populations. AI start-up ClosedLoop developed and open-sourced the C-19 Index, a predictive model that identifies people most at-risk of severe complications from COVID-19 via a short questionnaire. Risk factors were determined using a NLP and then structured in an algorithm that places users on a risk spectrum. The C-19 Index is used by healthcare systems, care management organizations, and insurance companies to recognize high-risk individuals, contact them to share safety resources, and provide them with personal protective equipment. |
Although AI can’t spit out a cure for COVID-19, it is helping researchers develop resourceful and cost-effective strategies for making counter-COVID-19 medicine. Synthia, an AI-backed drug synthesis program, is being applied by University of Michigan chemist Timothy Cernak and colleagues to create new recipes for COVID-19 drugs and prevent supply shortages. Just this month, Cernak’s lab identified novel solutions for making 11 out of 12 compounds now being tested as COVID-19 therapies, in one case with cheaper starting materials than those currently in use. As new drugs are proven effective against COVID-19, ensuring that they can be manufactured in abundance will be high-priority. |
Subsets and Splits