Andrew Konya - Remesh Archives - GRBN.ORG

Beyond Intelligence Augmentation: The Next Phase of AI & Market Research

GRBN_News — Mon, 16 Oct 2017 06:20:33 +0000

Advances in artificial intelligence are set to dramatically impact market research in the short term. Specifically, narrow AI will enable the automation of individual research tasks. During this phase of Intelligence Augmentation AI will act to augment the capabilities of researches; enabling one person to achieve what previously would have taken an entire team. The next phase of AI’s impact will happen as AI becomes increasingly generalized; extending automation out to functions which typically are not in the scope of a market researcher – such as content creation. The application of reinforcement learning based approaches to simultaneously solve market research & design challenges will signal the transition into this phase— and potentially the end of market research as we know it.

What is reinforcement learning? And how might it impact market research?

Reinforcement learning (RL) involves an intelligent agent which takes actions and learns from the outcome of those actions in order to become increasingly better at achieving a goal. Formulating a problem into one which can be tackled by an RL Agent (RLA) requires defining three things:

What is the RL’s goal?
What actions can the RL take?
What does it observe in order to learn how its actions impact its goal?

Already, AI researchers have developed RLAs capable of playing video games which require complex strategy and planning at a superhuman level. In this case the RLA’s goal is to achieve the highest score possible, the actions it can take are virtually ‘pushing’ the game controller’s buttons and it observes the video game screen in order to learn how its actions impact the state of the game and ultimately the score. To see how this approach can map to something at the intersection of market research and design, lets consider how an ad is created, tested, & then launched. Specifically, lets consider a simple ad like the ones you see in Google search results, and define the purpose of the ad to drive a person to click it and then sign up for some service. In this case, we can define the RLA’s goal as maximizing the number of people who click the ad and sign up (given a set of constraints, like a budget). The actions it can take are to generate the content of the ad and make ad buys against various demographics. During this process it observes the click through rates & sign up rates of the various ads & targeting profiles. At first, one can imagine it learning to generate ads that get a high rate of click through because it has learned that a certain set of words get people to click & that a person clicking is correlated with them signing up (which is the goal). With more time it may learn that while a certain set of ads generate strong click through rates, the rate of signups after clicking through varies greatly. As it then hones in on the ads that achieve both high click through & sign up rates, it may lean how those rates vary across demographics and optimize its targeting accordingly. With this example we can see how an RLA could learn to produce ads that not only drive click throughs, but specifically those which are likely convert to a sign up, and then learn who to best target those ads at. From a financial perspective, the RLA would learn to continuously reduce the cost per sign up — a clearly quantifiable ROI .

ROI and Morality

While this seems ideal from an economic perspective, morally it does leave many open questions about how such an ROI optimizing RLA might impact society (more on that here). Andrew Konya CEO at Remesh.ai Andrew Konya is the founder and CEO of Remesh. A computational physicist by training, he has spent the past 8 years developing and applying artificial intelligence and machine learning algorithms to problems in material science, bio-sensing, traffic, image analysis and language. His most recent focus is on developing artificial intelligence to engage and understand large crowds of people with Remesh. Find out more at Remesh at remesh.ai

The post Beyond Intelligence Augmentation: The Next Phase of AI & Market Research appeared first on GRBN.ORG.

AI’s Impact on Market Research: The Present

GRBN_News — Mon, 05 Jun 2017 07:36:57 +0000

For decades, researchers have developed advanced methods which allow us to analyze quantitative data (think survey) in sophisticated ways; from clustering & factor models to predictive Bayesian analysis. Perhaps the largest impact AI is already having on market research is enabling these battle-tested quantitative methods to be used on data which is qualitative in nature — namely video, audio, and text.

AI Increasing Accuracy and Speed

While there have been crude methods for quantifying these type of data in the past, a new generation of deep learning algorithms are capable of doing this with dramatically increased accuracy and speed. At that heart of these algorithms is the idea of “encoding” — that is, the act of converting qualitative data into a quantitative vector. A simple way to conceptualize “encoding” is this: Imagine we start with an open-ended question. Now imagine we construct a set of 200 binary ‘quant’ questions which aim to surface the same information as the open-ended question. Given a set of responses to the open-ended question, we can think of “encoding” as doing two things. First, identifying the best 200 quant ‘questions’ to capture all of the qualitative information found in the responses. Then, for each qualitative response, computing what the answers to those 200 quantitative questions would have been. The numerical vector containing these 200 ‘answers’ is the ‘encoded’ response. In this conceptualization, we can think of those open-ended responses as video, audio, or text because encoding is possible agnostic of data type. The main line separating current approaches to encoding is between supervised and unsupervised approaches.

Supervised vs. Unsupervised AI Approaches

Supervised approaches start by having a human decide what the ‘questions’ are ahead of time, and then assemble a ‘training set’ which has example pairs of qualitative data and human-specified answers to the corresponding ‘questions.’ A good example of this is encoding facial expressions into the emotions they express. In this case, humans identify the ‘questions’ of which emotions a person’s face might express, create a ‘training’ set which contained pairs of faces & the emotions they expressed, then use this to train a model which encodes a picture of a face into the emotions it expressed. Supervised encoding models have the advantage that they are relatively easy to build, however, they require human labor to develop training sets and are limited by their ability to only encode in the way they were trained. Unsupervised models do not require a human to identify the ‘questions’ or tag a training set — the only data they require to learn is the ‘qualitative’ data which they aim to encode. Unsupervised models simultaneously learn both the ‘questions’ which best capture the qualitative information & what the answers to those ‘questions’ should be for a given input (like a video or sentence). While crude, unsupervised models have been around for some time (like LDA for topic analysis & k-means for simple clustering), deep-learning based approaches (like auto-encoders) are enabling meaningful encoding of qualitative data previously thought impossible. These new capabilities, which enable the deep quantification of qualitative data, are already enabling researchers to bring the methodological rigor of quantitative methods into the world of qualitative research but I am confident this is only the beginning. Andrew Konya, Remesh Andrew Konya is the founder and CEO of Remesh. A computational physicist by training, he has spent the past 8 years developing and applying artificial intelligence and machine learning algorithms to problems in material science, bio-sensing, traffic, image analysis and language. His most recent focus is on developing artificial intelligence to engage and understand large crowds of people with Remesh. Find out more at Remesh at remesh.ai

The post AI’s Impact on Market Research: The Present appeared first on GRBN.ORG.

The Morality of Market Research in the Age of Artificial Intelligence

GRBN_News — Mon, 27 Feb 2017 09:00:16 +0000

As I plunged into the MR world, after a decade of working on computational physics and artificial intelligence, the first thing that struck me was the tremendous amount of technological overhang. Meaning, the solutions dominating the market were laden with inefficiencies that current technology already had solutions to address. So, with a focus on artificial intelligence, I set out to break down the primary functions of market research and plot the trajectory that artificial intelligence (AI) fueled disruption was likely to take.

Automation

First, advances in natural language processing (NLP) and machine learning models will lead to the automation of data analysis (and already are!). Second, advances in natural language generation and other deep learning techniques will drive the automation of data collection via conversation with humans and passive data digestion. Third, with data automatically collected and analyzed, AI will learn by example how humans turn these insights into reports; leading to automation of report generation. Lastly, building on the layers of automation described above, AI can begin to generate and understand the landscape of possible actions to take. Then, using the insights derived from the data available, it will map those actions to the most likely outcomes. After weighing those outcomes against the desired goal, the AI will recommend the action, or course of actions, which yields the highest probability of achieving the desired goal. Along this path it is important to understand the role a ‘goal’ plays — AI can be thought of as a machine which optimizes variables to achieve a specific goal. The power of AI is growing and its ability to more perfectly achieve goals is increasing. Thus, we may use this paradigm as a lens to consider the morality of specific goals by considering what would happen if those goals were achieved to perfection.

Market Research AI Goals

For market researchers three broad categories of goals stand out: the first is to better understand the needs of consumers so that a product can be developed to better meet those needs; the second is to better understand how to accurately communicate a product offering to consumers so that they know how well it meets their needs; the third is to understand the psychology of consumers so that they may be manipulated into buying products that they do not necessarily need. If we extrapolate these first two goals we arrive at a world where companies produce products which perfectly meet the needs of consumers and where consumers perfectly understand which products best meet their needs. In this world, products are produced when a need is identified, and these products quickly evolve along with consumers’ needs. In contrast, if we extrapolate the third goal we get a world where people’s core beliefs about themselves and the world they inhabit are systematically distorted in order to maximize the amount of products they buy (or who they vote for). In this world, the focus is not on products which meet needs, but rather on how to manipulate people’s psychology to believe they need to buy a product which a company already produces. To me it is clear that building towards one of these worlds is morally permissible and one is not. One makes human lives better, while the other maximizes profits. With this in mind, I believe it is important that we as an industry of market researchers are aware of the consequences of the goals we work to achieve. We hold a key role in shaping the future world our children and grandchildren will inhabit and it is important we take this responsibility seriously. Andrew Konya, CEO Remesh.ai Andrew Konya is the founder and CEO of Remesh. A computational physicist by training, he has spent the past 8 years developing and applying artificial intelligence and machine learning algorithms to problems in material science, bio-sensing, traffic, image analysis and language. His most recent focus is on developing artificial intelligence to engage and understand large crowds of people with Remesh. Find out more at Remesh at remesh.ai

The post The Morality of Market Research in the Age of Artificial Intelligence appeared first on GRBN.ORG.