A simple tool to start using AI for decision making.
Hello! Let's start the month with a fairly easy but useful material, the publication of which is timed to the start of the launch of the " Big Data for Managers " course, which starts in mid-April. So, let's begin.
There is a huge amount of authoritative opinions on the subject of the influence of artificial intelligence (AI) on the business of the near future. But much less is said about how companies can start using it. Our study and the book begin with the analysis of AI on its simplest components. We offer a way to implement this first step.
Let's start with a simple idea: the latest developments in the field of AI are aimed at reducing the cost of prediction. AI improves forecasts, makes them faster and cheaper. It has become much easier to predict not only the future (What will the weather be like next week?) But also the present (How is this Spanish website translated into English?). Forecasting is the use of available information to obtain information that you do not have. If you have information (data) that needs to be filtered, compressed, and sorted for ideas that facilitate decision making, forecasting will help to do this. And now the machines can help with this.
')
Improved predictions help make decisions in the face of uncertainty, which is often a business situation. But how to think about the introduction of a machine for prediction in the decision-making process?
We taught this topic to graduates of the Rothman School of Management MBA at the University of Toronto and talked about a simple decision-making tool: AI Canvas. Each element of the canvas contains one of the requirements for making decisions using a computer, starting with the prediction.
AI canvas
Use it to understand how AI will help you make business decisions.
More than 97% of cases of the home alarm system act out to be false. That is, their cause is not an intruder. The security company needs to make a decision: should the police or the security be called? Call the landlord? Ignore? If the company decides to act in more than 90 cases out of 100, it will be in vain. However, taking measures in response to an alarm signaling means that if there really is a danger, the security company will not disregard it.
How to understand if a predictive machine will help you? AI Canvas - a simple tool to systematize the necessary information in seven categories to obtain the necessary solutions. We will understand the example of the burglar alarm.
AI canvas: An example of using AI to improve home security
Prediction no longer consists in “movement = anxiety”, but, for example, “movement + an unknown person = alarm”. More sophisticated predictions reduce the number of false positives, which makes it easier to decide to send a security guard to check, instead of calling the owner.
Prediction cannot be 100% accurate. Therefore, to determine the size of the investment in improving predictions, you need to know the cost of false positives compared to the cost of ignoring the present. It depends on the situation and requires human evaluation. How much is a call back to confirm the situation? How much does it cost to send a guard in response to an alarm? How much does a quick reaction cost? How much will inaction, if the attacker really was in the house? There are many factors to consider; Determining their relative value requires evaluation.
Such an assessment can change the essence of the machine you have deployed to predict. In the case of alarms, cameras throughout the house are one of the best options for detecting the presence of an unknown intruder. But many people may find this uncomfortable.
Some will prefer the confidentiality of reducing the number of false alarms. Assessment sometimes requires the determination of relative values ​​and factors that are difficult to calculate, and therefore compare. The cost of false positives is easy to measure, the price of confidentiality is not.
Then, determine the action that depends on the generated forecast. It can be a simple solution to “respond / not respond”, or something more nuanced. Possible options for action include not only the reaction of someone, but also the immediate inclusion of remote monitoring of who is at home, or some way to contact the owner of the house.
The action leads to the result. For example, the security company responded and sent a security guard to check (action), who found the violator (result). In other words, looking back, we can see whether the right decisions were made at all stages. This knowledge is useful for assessing the need to improve predictions over time. If you do not know what result you want to get, improvements will be difficult, if not impossible.
Part of the canvas - prediction, evaluation, action and result, describe important aspects of the decision. The other part is the three final considerations. They are all associated with the data. To generate a useful prediction, you need to know what happens at the time of the decision - in our case, when the alarm goes off. In the example above, this includes motion sensor data and camera visual data collected in real time. This is the most basic input.
But in order to develop a prediction machine, you first need to train a machine learning model. The training data consists of sensor data for a certain period of time with corresponding results for calibrating the algorithms underlying the prediction machine. In this case, let us present a huge table, where each row is the alarm time, whether the attacker was in fact and some other data, for example, time and place. The richer and more diverse the training data, the better your predictions will be. If there is no data, you will have to start the mediocre prediction engine and wait for its improvement over time.
Improvements will come from feedback. This is the data that you collect while the machine is running in real situations. Feedback data is often generated in a more saturated environment than training. In our example, it is possible to find the connection of the result with the data obtained by the sensors through the windows, which affects how movement is recognized, and how cameras take faces — which is perhaps more realistic than the data used for training. So you can still improve the accuracy of the predictions through continuous training on feedback data. Sometimes such data will be sharpened on a certain house. And in other cases, it may extend to several.
Explaining these seven factors for each important decision of your organization will help determine whether the AI ​​can reduce costs or improve performance. Here we have discussed a solution related to a specific situation. To get started with AI, your task is to identify key decisions in your organization where the outcome depends on uncertainties. Completing the AI ​​Holst will not be able to say whether you need your own AI or you can buy ready-made from a supplier, but will be able to explain how the AI ​​will contribute (prediction), how it will interact with people (assessment), how it will affect decisions (action), how the success (result) will be assessed, and what types of data are needed for learning, operating and improving AI.
The potential is huge. For example, triggering an alarm sends a prediction to a remote agent. One of the reasons for this approach is a huge number of false positives. But think about it - if the predictive machines become so smart that there are no more false positives, will the response and sending the guard be the right decision? One can only imagine alternative solutions, for example, a system of capturing an attacker in place (as in cartoons!), Which could exist with more accurate and high-quality predictions. In general, improving predictions creates more opportunities for the emergence of new approaches to security, or even a potential prediction of an attacker's intentions before they penetrate.
If you find the material useful, we will be grateful for your advantages. And to learn more about the program of the course right now, you can sign up for a free open webinar , which our teacher, Artyom Prosvetov , will conduct on April 3.
There is a huge amount of authoritative opinions on the subject of the influence of artificial intelligence (AI) on the business of the near future. But much less is said about how companies can start using it. Our study and the book begin with the analysis of AI on its simplest components. We offer a way to implement this first step.
Let's start with a simple idea: the latest developments in the field of AI are aimed at reducing the cost of prediction. AI improves forecasts, makes them faster and cheaper. It has become much easier to predict not only the future (What will the weather be like next week?) But also the present (How is this Spanish website translated into English?). Forecasting is the use of available information to obtain information that you do not have. If you have information (data) that needs to be filtered, compressed, and sorted for ideas that facilitate decision making, forecasting will help to do this. And now the machines can help with this.
')
Improved predictions help make decisions in the face of uncertainty, which is often a business situation. But how to think about the introduction of a machine for prediction in the decision-making process?
We taught this topic to graduates of the Rothman School of Management MBA at the University of Toronto and talked about a simple decision-making tool: AI Canvas. Each element of the canvas contains one of the requirements for making decisions using a computer, starting with the prediction.
AI canvas
Use it to understand how AI will help you make business decisions.
PREDICTIONTo explain the work of canvas AI, we use an example that was devised at one of the workshops on Craig Campbell’s AI strategy, Craig Campbell, CEO of Peloton Innovations, an organization that is engaged in introducing AI into the security industry. (This is a real example based on a product called RSPNDR.ai that sells Peloton.)
What you need to know to make a decision?
EVALUATION
How are the various results and errors evaluated?
ACT
What are you trying to do?
RESULT
What metrics are used to measure success?
INPUT DATA
What data is needed to run the predictive algorithm?
TRAINING
What data is needed to train the predictive algorithm?
FEEDBACK
How can the results be used to improve the algorithm?
More than 97% of cases of the home alarm system act out to be false. That is, their cause is not an intruder. The security company needs to make a decision: should the police or the security be called? Call the landlord? Ignore? If the company decides to act in more than 90 cases out of 100, it will be in vain. However, taking measures in response to an alarm signaling means that if there really is a danger, the security company will not disregard it.
How to understand if a predictive machine will help you? AI Canvas - a simple tool to systematize the necessary information in seven categories to obtain the necessary solutions. We will understand the example of the burglar alarm.
AI canvas: An example of using AI to improve home security
PREDICTIONFirst, clarify what needs to be predicted. In the case of an alarm, you need to find out whether it was caused by an unknown person or not (a false alarm or not). A predictive machine can potentially report this — after all, an alarm with a simple motion sensor is to some extent a predictive machine. Machine learning allows you to use a wider range of sensor data to determine exactly what you want to predict: whether the movement was caused by an unknown person. With the right sensors, such as a camera that recognizes faces of people and pets, or a door lock that recognizes when someone is near the door, modern AI technologies provide more detailed predictions.
Predict whether the alarm has triggered an unknown person or something else (i.e. true or false).
EVALUATION
Compare the cost of reacting to a false alarm with the cost of inaction in the event of a real response.
ACT
Respond or not in the event of a signal.
RESULT
Whether the right decision was made when the alarm was triggered.
INPUT DATA
The data of motion sensors, heat, cameras for each moment during the alarm. This data will control the AI.
TRAINING
Sensory data for a certain period of time and the corresponding data of the results of the operation (a real attacker or a false response); This data is used to teach the AI ​​before launching it.
FEEDBACK
Sensor data and the corresponding response results (confirmed by the attacker or confirmed false positives); This data is used to update the model while the AI ​​is running.
Prediction no longer consists in “movement = anxiety”, but, for example, “movement + an unknown person = alarm”. More sophisticated predictions reduce the number of false positives, which makes it easier to decide to send a security guard to check, instead of calling the owner.
Prediction cannot be 100% accurate. Therefore, to determine the size of the investment in improving predictions, you need to know the cost of false positives compared to the cost of ignoring the present. It depends on the situation and requires human evaluation. How much is a call back to confirm the situation? How much does it cost to send a guard in response to an alarm? How much does a quick reaction cost? How much will inaction, if the attacker really was in the house? There are many factors to consider; Determining their relative value requires evaluation.
Such an assessment can change the essence of the machine you have deployed to predict. In the case of alarms, cameras throughout the house are one of the best options for detecting the presence of an unknown intruder. But many people may find this uncomfortable.
Some will prefer the confidentiality of reducing the number of false alarms. Assessment sometimes requires the determination of relative values ​​and factors that are difficult to calculate, and therefore compare. The cost of false positives is easy to measure, the price of confidentiality is not.
Then, determine the action that depends on the generated forecast. It can be a simple solution to “respond / not respond”, or something more nuanced. Possible options for action include not only the reaction of someone, but also the immediate inclusion of remote monitoring of who is at home, or some way to contact the owner of the house.
The action leads to the result. For example, the security company responded and sent a security guard to check (action), who found the violator (result). In other words, looking back, we can see whether the right decisions were made at all stages. This knowledge is useful for assessing the need to improve predictions over time. If you do not know what result you want to get, improvements will be difficult, if not impossible.
Part of the canvas - prediction, evaluation, action and result, describe important aspects of the decision. The other part is the three final considerations. They are all associated with the data. To generate a useful prediction, you need to know what happens at the time of the decision - in our case, when the alarm goes off. In the example above, this includes motion sensor data and camera visual data collected in real time. This is the most basic input.
But in order to develop a prediction machine, you first need to train a machine learning model. The training data consists of sensor data for a certain period of time with corresponding results for calibrating the algorithms underlying the prediction machine. In this case, let us present a huge table, where each row is the alarm time, whether the attacker was in fact and some other data, for example, time and place. The richer and more diverse the training data, the better your predictions will be. If there is no data, you will have to start the mediocre prediction engine and wait for its improvement over time.
Improvements will come from feedback. This is the data that you collect while the machine is running in real situations. Feedback data is often generated in a more saturated environment than training. In our example, it is possible to find the connection of the result with the data obtained by the sensors through the windows, which affects how movement is recognized, and how cameras take faces — which is perhaps more realistic than the data used for training. So you can still improve the accuracy of the predictions through continuous training on feedback data. Sometimes such data will be sharpened on a certain house. And in other cases, it may extend to several.
Explaining these seven factors for each important decision of your organization will help determine whether the AI ​​can reduce costs or improve performance. Here we have discussed a solution related to a specific situation. To get started with AI, your task is to identify key decisions in your organization where the outcome depends on uncertainties. Completing the AI ​​Holst will not be able to say whether you need your own AI or you can buy ready-made from a supplier, but will be able to explain how the AI ​​will contribute (prediction), how it will interact with people (assessment), how it will affect decisions (action), how the success (result) will be assessed, and what types of data are needed for learning, operating and improving AI.
The potential is huge. For example, triggering an alarm sends a prediction to a remote agent. One of the reasons for this approach is a huge number of false positives. But think about it - if the predictive machines become so smart that there are no more false positives, will the response and sending the guard be the right decision? One can only imagine alternative solutions, for example, a system of capturing an attacker in place (as in cartoons!), Which could exist with more accurate and high-quality predictions. In general, improving predictions creates more opportunities for the emergence of new approaches to security, or even a potential prediction of an attacker's intentions before they penetrate.
If you find the material useful, we will be grateful for your advantages. And to learn more about the program of the course right now, you can sign up for a free open webinar , which our teacher, Artyom Prosvetov , will conduct on April 3.
Source: https://habr.com/ru/post/446298/
All Articles