Reinforcement learning is a very useful (and currently popular) subtype of machine learning and artificial intelligence. It is based on the principle that agents, when placed in an interactive environment, can learn from their actions via rewards associated with the actions, and improve the time to achieve their goal.
In this article, we’ll explore the fundamental concepts of reinforcement learning and discuss its key components, types, and applications.
Definition of Reinforcement Learning
We can define reinforcement learning as a machine learning technique involving an agent who needs to decide which actions it needs to do to perform a task that has been assigned to it most effectively. For this, rewards are assigned to the different actions that the agent can take at different situations or states of the environment. Initially, the agent has no idea about the best or correct actions. Using reinforcement learning, it explores its action choices via trial and error and figures out the best set of actions for completing its assigned task.
The basic idea behind a reinforcement learning agent is to learn from experience. Just like humans learn lessons from their past successes and mistakes, reinforcement learning agents do the same – when they do something “good” they get a reward, but, if they do something “bad”, they get penalized. The reward reinforces the good actions while the penalty avoids the bad ones.
Reinforcement learning requires several key components:
- Agent – This is the “who” or the subject of the process, which performs different actions to perform a task that has been assigned to it.
- Environment – This is the “where” or a situation in which the agent is placed.
- Actions – This is the “what” or the steps an agent needs to take to reach the goal.
- Rewards – This is the feedback an agent receives after performing an action.
Before we dig deep into the technicalities, let’s warm up with a real-life example. Reinforcement isn’t new, and we’ve used it for different purposes for centuries. One of the most basic examples is dog training.
Let’s say you’re in a park, trying to teach your dog to fetch a ball. In this case, the dog is the agent, and the park is the environment. Once you throw the ball, the dog will run to catch it, and that’s the action part. When he brings the ball back to you and releases it, he’ll get a reward (a treat). Since he got a reward, the dog will understand that his actions were appropriate and will repeat them in the future. If the dog doesn’t bring the ball back, he may get some “punishment” – you may ignore him or say “No!” After a few attempts (or more than a few, depending on how stubborn your dog is), the dog will fetch the ball with ease.
We can say that the reinforcement learning process has three steps:
- Interaction
- Learning
- Decision-making
Types of Reinforcement Learning
There are two types of reinforcement learning: model-based and model-free.
Model-Based Reinforcement Learning
With model-based reinforcement learning (RL), there’s a model that an agent uses to create additional experiences. Think of this model as a mental image that the agent can analyze to assess whether particular strategies could work.
Some of the advantages of this RL type are:
- It doesn’t need a lot of samples.
- It can save time.
- It offers a safe environment for testing and exploration.
The potential drawbacks are:
- Its performance relies on the model. If the model isn’t good, the performance won’t be good either.
- It’s quite complex.
Model-Free Reinforcement Learning
In this case, an agent doesn’t rely on a model. Instead, the basis for its actions lies in direct interactions with the environment. An agent tries different scenarios and tests whether they’re successful. If yes, the agent will keep repeating them. If not, it will try another scenario until it finds the right one.
What are the advantages of model-free reinforcement learning?
- It doesn’t depend on a model’s accuracy.
- It’s not as computationally complex as model-based RL.
- It’s often better for real-life situations.
Some of the drawbacks are:
- It requires more exploration, so it can be more time-consuming.
- It can be dangerous because it relies on real-life interactions.
Model-Based vs. Model-Free Reinforcement Learning: Example
Understanding model-based and model-free RL can be challenging because they often seem too complex and abstract. We’ll try to make the concepts easier to understand through a real-life example.
Let’s say you have two soccer teams that have never played each other before. Therefore, neither of the teams knows what to expect. At the beginning of the match, Team A tries different strategies to see whether they can score a goal. When they find a strategy that works, they’ll keep using it to score more goals. This is model-free reinforcement learning.
On the other hand, Team B came prepared. They spent hours investigating strategies and examining the opponent. The players came up with tactics based on their interpretation of how Team A will play. This is model-based reinforcement learning.
Who will be more successful? There’s no way to tell. Team B may be more successful in the beginning because they have previous knowledge. But Team A can catch up quickly, especially if they use the right tactics from the start.
Reinforcement Learning Algorithms
A reinforcement learning algorithm specifies how an agent learns suitable actions from the rewards. RL algorithms are divided into two categories: value-based and policy gradient-based.
Value-Based Algorithms
Value-based algorithms learn the value at each state of the environment, where the value of a state is given by the expected rewards to complete the task while starting from that state.
Q-Learning
This model-free, off-policy RL algorithm focuses on providing guidelines to the agent on what actions to take and under what circumstances to win the reward. The algorithm uses Q-tables in which it calculates the potential rewards for different state-action pairs in the environment. The table contains Q-values that get updated after each action during the agent’s training. During execution, the agent goes back to this table to see which actions have the best value.
Deep Q-Networks (DQN)
Deep Q-networks, or deep q-learning, operate similarly to q-learning. The main difference is that the algorithm in this case is based on neural networks.
SARSA
The acronym stands for state-action-reward-state-action. SARSA is an on-policy RL algorithm that uses the current action from the current policy to learn the value.
Policy-Based Algorithms
These algorithms directly update the policy to maximize the reward. There are different policy gradient-based algorithms: REINFORCE, proximal policy optimization, trust region policy optimization, actor-critic algorithms, advantage actor-critic, deep deterministic policy gradient (DDPG), and twin-delayed DDPG.
Examples of Reinforcement Learning Applications
The advantages of reinforcement learning have been recognized in many spheres. Here are several concrete applications of RL.
Robotics and Automation
With RL, robotic arms can be trained to perform human-like tasks. Robotic arms can give you a hand in warehouse management, packaging, quality testing, defect inspection, and many other aspects.
Another notable role of RL lies in automation, and self-driving cars are an excellent example. They’re introduced to different situations through which they learn how to behave in specific circumstances and offer better performance.
Gaming and Entertainment
Gaming and entertainment industries certainly benefit from RL in many ways. From AlphaGo (the first program that has beaten a human in the board game Go) to video games AI, RL offers limitless possibilities.
Finance and Trading
RL can optimize and improve trading strategies, help with portfolio management, minimize risks that come with running a business, and maximize profit.
Healthcare and Medicine
RL can help healthcare workers customize the best treatment plan for their patients, focusing on personalization. It can also play a major role in drug discovery and testing, allowing the entire sector to get one step closer to curing patients quickly and efficiently.
Basics for Implementing Reinforcement Learning
The success of reinforcement learning in a specific area depends on many factors.
First, you need to analyze a specific situation and see which RL algorithm suits it. Your job doesn’t end there; now you need to define the environment and the agent and figure out the right reward system. Without them, RL doesn’t exist. Next, allow the agent to put its detective cap on and explore new features, but ensure it uses the existing knowledge adequately (strike the right balance between exploration and exploitation). Since RL changes rapidly, you want to keep your model updated. Examine it every now and then to see what you can tweak to keep your model in top shape.
Explore the World of Possibilities With Reinforcement Learning
Reinforcement learning goes hand-in-hand with the development and modernization of many industries. We’ve been witnesses to the incredible things RL can achieve when used correctly, and the future looks even better. Hop in on the RL train and immerse yourself in this fascinating world.
-
From the same author
Read the articleReinforcement learning is a very useful (and currently popular) subtype of machine learning and artificial intelligence. It is based on the principle that agents, when placed in an interactive environment, can learn from their actions via rewards associated with the actions, and improve the time to achieve their goal.
In this article, we’ll explore the fundamental concepts of reinforcement learning and discuss its key components, types, and applications.
Definition of Reinforcement Learning
We can define reinforcement learning as a machine learning technique involving an agent who needs to decide which actions it needs to do to perform a task that has been assigned to it most effectively. For this, rewards are assigned to the different actions that the agent can take at different situations or states of the environment. Initially, the agent has no idea about the best or correct actions. Using reinforcement learning, it explores its action choices via trial and error and figures out the best set of actions for completing its assigned task.
The basic idea behind a reinforcement learning agent is to learn from experience. Just like humans learn lessons from their past successes and mistakes, reinforcement learning agents do the same – when they do something “good” they get a reward, but, if they do something “bad”, they get penalized. The reward reinforces the good actions while the penalty avoids the bad ones.
Reinforcement learning requires several key components:
- Agent – This is the “who” or the subject of the process, which performs different actions to perform a task that has been assigned to it.
- Environment – This is the “where” or a situation in which the agent is placed.
- Actions – This is the “what” or the steps an agent needs to take to reach the goal.
- Rewards – This is the feedback an agent receives after performing an action.
Before we dig deep into the technicalities, let’s warm up with a real-life example. Reinforcement isn’t new, and we’ve used it for different purposes for centuries. One of the most basic examples is dog training.
Let’s say you’re in a park, trying to teach your dog to fetch a ball. In this case, the dog is the agent, and the park is the environment. Once you throw the ball, the dog will run to catch it, and that’s the action part. When he brings the ball back to you and releases it, he’ll get a reward (a treat). Since he got a reward, the dog will understand that his actions were appropriate and will repeat them in the future. If the dog doesn’t bring the ball back, he may get some “punishment” – you may ignore him or say “No!” After a few attempts (or more than a few, depending on how stubborn your dog is), the dog will fetch the ball with ease.
We can say that the reinforcement learning process has three steps:
- Interaction
- Learning
- Decision-making
Types of Reinforcement Learning
There are two types of reinforcement learning: model-based and model-free.
Model-Based Reinforcement Learning
With model-based reinforcement learning (RL), there’s a model that an agent uses to create additional experiences. Think of this model as a mental image that the agent can analyze to assess whether particular strategies could work.
Some of the advantages of this RL type are:
- It doesn’t need a lot of samples.
- It can save time.
- It offers a safe environment for testing and exploration.
The potential drawbacks are:
- Its performance relies on the model. If the model isn’t good, the performance won’t be good either.
- It’s quite complex.
Model-Free Reinforcement Learning
In this case, an agent doesn’t rely on a model. Instead, the basis for its actions lies in direct interactions with the environment. An agent tries different scenarios and tests whether they’re successful. If yes, the agent will keep repeating them. If not, it will try another scenario until it finds the right one.
What are the advantages of model-free reinforcement learning?
- It doesn’t depend on a model’s accuracy.
- It’s not as computationally complex as model-based RL.
- It’s often better for real-life situations.
Some of the drawbacks are:
- It requires more exploration, so it can be more time-consuming.
- It can be dangerous because it relies on real-life interactions.
Model-Based vs. Model-Free Reinforcement Learning: Example
Understanding model-based and model-free RL can be challenging because they often seem too complex and abstract. We’ll try to make the concepts easier to understand through a real-life example.
Let’s say you have two soccer teams that have never played each other before. Therefore, neither of the teams knows what to expect. At the beginning of the match, Team A tries different strategies to see whether they can score a goal. When they find a strategy that works, they’ll keep using it to score more goals. This is model-free reinforcement learning.
On the other hand, Team B came prepared. They spent hours investigating strategies and examining the opponent. The players came up with tactics based on their interpretation of how Team A will play. This is model-based reinforcement learning.
Who will be more successful? There’s no way to tell. Team B may be more successful in the beginning because they have previous knowledge. But Team A can catch up quickly, especially if they use the right tactics from the start.
Reinforcement Learning Algorithms
A reinforcement learning algorithm specifies how an agent learns suitable actions from the rewards. RL algorithms are divided into two categories: value-based and policy gradient-based.
Value-Based Algorithms
Value-based algorithms learn the value at each state of the environment, where the value of a state is given by the expected rewards to complete the task while starting from that state.
Q-Learning
This model-free, off-policy RL algorithm focuses on providing guidelines to the agent on what actions to take and under what circumstances to win the reward. The algorithm uses Q-tables in which it calculates the potential rewards for different state-action pairs in the environment. The table contains Q-values that get updated after each action during the agent’s training. During execution, the agent goes back to this table to see which actions have the best value.
Deep Q-Networks (DQN)
Deep Q-networks, or deep q-learning, operate similarly to q-learning. The main difference is that the algorithm in this case is based on neural networks.
SARSA
The acronym stands for state-action-reward-state-action. SARSA is an on-policy RL algorithm that uses the current action from the current policy to learn the value.
Policy-Based Algorithms
These algorithms directly update the policy to maximize the reward. There are different policy gradient-based algorithms: REINFORCE, proximal policy optimization, trust region policy optimization, actor-critic algorithms, advantage actor-critic, deep deterministic policy gradient (DDPG), and twin-delayed DDPG.
Examples of Reinforcement Learning Applications
The advantages of reinforcement learning have been recognized in many spheres. Here are several concrete applications of RL.
Robotics and Automation
With RL, robotic arms can be trained to perform human-like tasks. Robotic arms can give you a hand in warehouse management, packaging, quality testing, defect inspection, and many other aspects.
Another notable role of RL lies in automation, and self-driving cars are an excellent example. They’re introduced to different situations through which they learn how to behave in specific circumstances and offer better performance.
Gaming and Entertainment
Gaming and entertainment industries certainly benefit from RL in many ways. From AlphaGo (the first program that has beaten a human in the board game Go) to video games AI, RL offers limitless possibilities.
Finance and Trading
RL can optimize and improve trading strategies, help with portfolio management, minimize risks that come with running a business, and maximize profit.
Healthcare and Medicine
RL can help healthcare workers customize the best treatment plan for their patients, focusing on personalization. It can also play a major role in drug discovery and testing, allowing the entire sector to get one step closer to curing patients quickly and efficiently.
Basics for Implementing Reinforcement Learning
The success of reinforcement learning in a specific area depends on many factors.
First, you need to analyze a specific situation and see which RL algorithm suits it. Your job doesn’t end there; now you need to define the environment and the agent and figure out the right reward system. Without them, RL doesn’t exist. Next, allow the agent to put its detective cap on and explore new features, but ensure it uses the existing knowledge adequately (strike the right balance between exploration and exploitation). Since RL changes rapidly, you want to keep your model updated. Examine it every now and then to see what you can tweak to keep your model in top shape.
Explore the World of Possibilities With Reinforcement Learning
Reinforcement learning goes hand-in-hand with the development and modernization of many industries. We’ve been witnesses to the incredible things RL can achieve when used correctly, and the future looks even better. Hop in on the RL train and immerse yourself in this fascinating world.
Read the articleAlgorithms are the backbone behind technology that have helped establish some of the world’s most famous companies. Software giants like Google, beverage giants Coca Cola and many other organizations utilize proprietary algorithms to improve their services and enhance customer experience. Algorithms are an inseparable part of the technology behind organization as they help improve security, product or service recommendations, and increase sales.
Knowing the benefits of algorithms is useful, but you might also be interested to know what makes them so advantageous. As such, you’re probably asking: “What is an algorithm?” Here’s the most common algorithm definition: an algorithm is a set of procedures and rules a computer follows to solve a problem.
In addition to the meaning of the word “algorithm,” this article will also cover the key types and characteristics of algorithms, as well as their applications.
Types of Algorithms and Design Techniques
One of the main reasons people rely on algorithms is that they offer a principled and structured means to represent a problem on a computer.
Recursive Algorithms
Recursive algorithms are critical for solving many problems. The core idea behind recursive algorithms is to use functions that call themselves on smaller chunks of the problem.
Divide and Conquer Algorithms
Divide and conquer algorithms are similar to recursive algorithms. They divide a large problem into smaller units. Algorithms solve each smaller component before combining them to tackle the original, large problem.
Greedy Algorithms
A greedy algorithm looks for solutions based on benefits. More specifically, it resolves problems in sections by determining how many benefits it can extract by analyzing a certain section. The more benefits it has, the more likely it is to solve a problem, hence the term greedy.
Dynamic Programming Algorithms
Dynamic programming algorithms follow a similar approach to recursive and divide and conquer algorithms. First, they break down a complex problem into smaller pieces. Next, it solves each smaller piece once and saves the solution for later use instead of computing it.
Backtracking Algorithms
After dividing a problem, an algorithm may have trouble moving forward to find a solution. If that’s the case, a backtracking algorithm can return to parts of the problem it has already solved until it determines a way forward that can overcome the setback.
Brute Force Algorithms
Brute force algorithms try every possible solution until they determine the best one. Brute force algorithms are simpler, but the solution they find might not be as good or elegant as those found by the other types of algorithms.
Algorithm Analysis and Optimization
Digital transformation remains one of the biggest challenges for businesses in 2023. Algorithms can facilitate the transition through careful analysis and optimization.
Time Complexity
The time complexity of an algorithm refers to how long you need to execute a certain algorithm. A number of factors determine time complexity, but the algorithm’s input length is the most important consideration.
Space Complexity
Before you can run an algorithm, you need to make sure your device has enough memory. The amount of memory required for executing an algorithm is known as space complexity.
Trade-Offs
Solving a problem with an algorithm in C or any other programming language is about making compromises. In other words, the system often makes trade-offs between the time and space available.
For example, an algorithm can use less space, but this extends the time it takes to solve a problem. Alternatively, it can take up a lot of space to address an issue faster.
Optimization Techniques
Algorithms generally work great out of the box, but they sometimes fail to deliver the desired results. In these cases, you can implement a slew of optimization techniques to make them more effective.
Memorization
You generally use memorization if you wish to elevate the efficacy of a recursive algorithm. The technique rewrites algorithms and stores them in arrays. The main reason memorization is so powerful is that it eliminates the need to calculate results multiple times.
Parallelization
As the name suggests, parallelization is the ability of algorithms to perform operations simultaneously. This accelerates task completion and is normally utilized when you have a lot of memory on your device.
Heuristics
Heuristic algorithms (a.k.a. heuristics) are algorithms used to speed up problem-solving. They generally target non-deterministic polynomial-time (NP) problems.
Approximation Algorithms
Another way to solve a problem if you’re short on time is to incorporate an approximation algorithm. Rather than provide a 100% optimal solution and risk taking longer, you use this algorithm to get approximate solutions. From there, you can calculate how far away they are from the optimal solution.
Pruning
Algorithms sometimes analyze unnecessary data, slowing down your task completion. A great way to expedite the process is to utilize pruning. This compression method removes unwanted information by shrinking algorithm decision trees.
Algorithm Applications and Challenges
Thanks to this introduction to algorithm, you’ll no longer wonder: “What is an algorithm, and what are the different types?” Now it’s time to go through the most significant applications and challenges of algorithms.
Sorting Algorithms
Sorting algorithms arrange elements in a series to help solve complex issues faster. There are different types of sorting, including linear, insertion, and bubble sorting. They’re generally used for exploring databases and virtual search spaces.
Searching Algorithms
An algorithm in C or other programming languages can be used as a searching algorithm. They allow you to identify a small item in a large group of related elements.
Graph Algorithms
Graph algorithms are just as practical, if not more practical, than other types. Graphs consist of nodes and edges, where each edge connects two nodes.
There are numerous real-life applications of graph algorithms. For instance, you might have wondered how engineers solve problems regarding wireless networks or city traffic. The answer lies in using graph algorithms.
The same goes for social media sites, such as Facebook. Algorithms on such platforms contain nodes, which represent key information, like names and genders and edges that represent the relationships or dependencies between them.
Cryptography Algorithms
When creating an account on some websites, the platform can generate a random password for you. It’s usually stronger than custom-made codes, thanks to cryptography algorithms. They can scramble digital text and turn it into an unreadable string. Many organizations use this method to protect their data and prevent unauthorized access.
Machine Learning Algorithms
Over 70% of enterprises prioritize machine learning applications. To implement their ideas, they rely on machine learning algorithms. They’re particularly useful for financial institutions because they can predict future trends.
Famous Algorithm Challenges
Many organizations struggle to adopt algorithms, be it an algorithm in data structure or computer science. The reason being, algorithms present several challenges:
- Opacity – You can’t take a closer look at the inside of an algorithm. Only the end result is visible, which is why it’s difficult to understand an algorithm.
- Heterogeneity – Most algorithms are heterogeneous, behaving differently from one another. This makes them even more complex.
- Dependency – Each algorithm comes with the abovementioned time and space restrictions.
Algorithm Ethics, Fairness, and Social Impact
When discussing critical characteristics of algorithms, it’s important to highlight the main concerns surrounding this technology.
Bias in Algorithms
Algorithms aren’t intrinsically biased unless the developer injects their personal biases into the design. If so, getting impartial results from an algorithm is highly unlikely.
Transparency and Explainability
Knowing only the consequences of algorithms prevents us from explaining them in detail. A transparent algorithm enables a user to view and understand its different operations. In contrast, explainability of an algorithm relates to its ability to provide reasons for the decisions it makes.
Privacy and Security
Some algorithms require end users to share private information. If cyber criminals hack the system, they can easily steal the data.
Algorithm Accessibility and Inclusivity
Limited explainability hinders access to algorithms. Likewise, it’s hard to include different viewpoints and characteristics in an algorithm, especially if it is biased.
Algorithm Trust and Confidence
No algorithm is omnipotent. Claiming otherwise makes it untrustworthy – the best way to prevent this is for the algorithm to state its limitations.
Algorithm Social Impact
Algorithms impact almost every area of life including politics, economic and healthcare decisions, marketing, transportation, social media and Internet, and society and culture in general.
Algorithm Sustainability and Environmental Impact
Contrary to popular belief, algorithms aren’t very sustainable. The extraction of materials to make computers that power algorithms is a major polluter.
Future of Algorithms
Algorithms are already advanced, but what does the future hold for this technology? Here are a few potential applications and types of future algorithms:
- Quantum Algorithms – Quantum algorithms are expected to run on quantum computers to achieve unprecedented speeds and efficiency.
- Artificial Intelligence and Machine Learning – AI and machine learning algorithms can help a computer develop human-like cognitive qualities via learning from its environment and experiences.
- Algorithmic Fairness and Ethics – Considering the aforementioned challenges of algorithms, developers are expected to improve the technology. It may become more ethical with fewer privacy violations and accessibility issues.
Smart, Ethical Implementation Is the Difference-Maker
Understanding algorithms is crucial if you want to implement them correctly and ethically. They’re powerful, but can also have unpleasant consequences if you’re not careful during the development stage. Responsible use is paramount because it can improve many areas, including healthcare, economics, social media, and communication.
If you wish to learn more about algorithms, accredited courses might be your best option. AI and machine learning-based modules cover some of the most widely-used algorithms to help expand your knowledge about this topic.
-
Go to all authors
Related posts
Source:
- IE University – Insights, Published on October 15th, 2024.
By Francesco Derchi
Purpose is a strategic tool for driving innovation, competitive advantage, and addressing AI challenges, writes Francesco Derchi.
Since the early 2000s, technology has dominated discussions among scholars and professionals about global development and economic trends, with the first wave of research regarding the internet’s impact on firms and society focusing on the enabling potential of technologies. The concept of “digital revolution,” as popularized by Nicholas Negroponte, became the new paradigm for broader considerations about the development of the firm’s macro environment, and how businesses could leverage it as an asset for creating competitive advantage. The following wave focused on the convergence of different technologies, such as manufacturing, and included the dynamics of coexistence between humans and machines. From the management side, the major challenges are related to defining effective digital transformation practices that could help to migrate organizations and exploit this new paradigm.
The current technological focus builds on these previous trends, particularly on artificial intelligence and more recently on the emergence of generative AI. The Age of AI is characterized by technology’s power to reshape business and society on a variety of levels. While AI’s pervasive impact is not new for firms, the mainstream adoption of ChatGPT for business purposes and the response to this ready adoption from big tech players like Microsoft, Google, and more recently Apple, shows how AI is reshaping and influencing companies’ strategic priorities.
From a research perspective, AI’s societal impact is inspiring new studies in the field of ethics. Luciano Floridi, now of Yale University, has identified several challenges for AI, characterizing them by global magnitudes like its environmental impact and has identified several challenges for AI security, including intellectual property, privacy, transparency, and accountability. In his work, Floridi underlines the importance of philosophy in defining problems and designing solutions – but it is equally important to consider how these challenges can be addressed at the firm level. What are the tools for managers?
Part of the answer may lie in the increasing and recent focus of management studies around “corporate purpose” and “brand purpose.” This trend represents an important attempt to deepen our understanding of “why to act” (purpose framing) and “how to act” (purpose formalizing and internalizing), while technology management studies address the “what to act” (purpose impacting) question. Furthermore, studies show that corporate purpose is critical for both digital native firms as well as traditional companies undergoing a digital transformation, serving as an important growth engine through purpose-driven innovation. It is therefore fair to ask: can purpose help in addressing any of the AI challenges previously mentioned?
Purpose concepts are not exclusively “cause-related” like CSR and environmental impact. Other types have emerged, such as “competence” (the function of the product) and “culture” (the intent that drives the business). This broadens the consideration of impact types that can help address specific challenges in the age of AI.
Purpose-driven organizations are not new. Take Tesla’s direction “to accelerate the world’s transition to sustainable energy” – it explicitly addresses environmental challenges while defining a business direction that requires constant innovation and leverages multiple converging technologies. The key is to have the purpose formalized and internalized within the company as a concrete drive for growth.
Due to its characteristics, the MTP plays a key role in digital transformation. This necessarily ambitious and long-term vision or goal – the Massive Transformative Purpose – requires firms, particularly those focused on exponential growth, to address emerging accelerating technologies with a purpose-first transformation logic. P&G’s Global Business Services division was able to improve market leadership and gain a competitive advantage over various start-ups and potential disruptors through its “Free up the employee, for free” MTP. This served as a north star for every employee, encouraging them to contribute ideas and best practices to overcome bulky processes and limitations.
My research on MTPs in AI-era firms explores their role in driving innovation to address specific challenges. Results show that the MTP impacts the organization across four dimensions, requiring commitment and synergy from management. Let’s consider these four dimensions by looking at Airbnb:
- Internal Impact: The MTP acts as the organization’s genetic code and guiding philosophy. It is key for leveraging employee motivation, with a strong relationship between purpose, organizational culture, and firm values. Airbnb’s culture of belonging highlights this, with its various purpose-shaping practices, starting with culture-fit interviews delivered during the recruitment process.
- Brand and Market Influence: The MTP contributes directly to building a strong brand and influencing the market. It allows firms to extend beyond functional and symbolic benefits to make the impact of the company on society visible. This involves addressing market demand coherently and consistently. Airbnb’s “Bélo” symbol visually represents this concept of belonging while their MTP features in campaigns like “Wall and Chain: A Story of Breaking Down Walls.”
- Competitive Advantage and Growth: The MTP drives innovation and can lead to superior stock market performance. In digital firms, it’s key in the creation of ecosystems that aggregate leveraged assets and third parties for value creation. The company’s “belong anywhere transformation journey” is a strategic initiative that formalized and interiorized the MTP through various touchpoints for all the different ecosystem members. As Leigh Gallagher details in her 2016 Fortune feature about the company, “When travellers leave their homes, they feel alone. They reach their Airbnb, and they feel accepted and taken care of by their host. They then feel safe to be the same kind of person they are when they’re at home.”
- Core Organization Identity: The MTP is considered part of the core dimension of the organization. More than a goal or business strategy, it is a strategic issue that generates a sense of direction and purpose that affects every part of the organization: internal, external, personality, and expression. This dimension also involves the role of the founder(s) and their personality in shaping the business. At Airbnb, the MTP is often used as a shortcut to explain the firm’s mission and vision. The founders’ approach is pragmatic, and instead of debating differences, time should be spent on execution. At the same time, the personalities of the three founders, Chesky, Gebbia, and Blecharcyzk, are the identity of the firm. They were the first hosts for the platform. Their credibility is key for making Airbnb a trustworthy and coherent proposal in a crowded market.
Executives and leaders of business in the current AI era should embrace three key principles. Be true: Purpose is an essential strategic tool that enables firms to identify and connect with their original selves, decoding their reason for being and embedding it into their identity. Be ambitious: The MTP allows for global impact, confronting major challenges by synthesizing business values and guiding innovation paths to address AI-related issues. Be generous: Purpose allows firms to explicitly address environmental and social issues, taking action on values-based challenges such as transparency, respect for intellectual property, and accountability. By following these principles, organizations and their leaders can maintain their direction and continue to advance in the AI era.
Read the full article below:
Source:
- Authority Magazine Medium, Published on September 15th, 2024.
Gaining hands-on experience through projects, internships, and collaborations is vital for understanding how to apply AI in various industries and domains. Use Kaggle or get a free cloud account and start experimenting. You will have projects to discuss at your next interviews.
By David Leichner, CMO at Cybellum
14 min read
Artificial Intelligence is now the leading edge of technology, driving unprecedented advancements across sectors. From healthcare to finance, education to environment, the AI industry is witnessing a skyrocketing demand for professionals. However, the path to creating a successful career in AI is multifaceted and constantly evolving. What does it take and what does one need in order to create a highly successful career in AI?
In this interview series, we are talking to successful AI professionals, AI founders, AI CEOs, educators in the field, AI researchers, HR managers in tech companies, and anyone who holds authority in the realm of Artificial Intelligence to inspire and guide those who are eager to embark on this exciting career path.
As part of this series, we had the pleasure of interviewing Zorina Alliata.
Zorina Alliata is an expert in AI, with over 20 years of experience in tech, and over 10 years in AI itself. As an educator, Zorina Alliata is passionate about learning, access to education and about creating the career you want. She implores us to learn more about ethics in AI, and not to fear AI, but to embrace it.
Thank you so much for joining us in this interview series! Before we dive in, our readers would like to learn a bit about your origin story. Can you share with us a bit about your childhood and how you grew up?
I was born in Romania, and grew up during communism, a very dark period in our history. I was a curious child and my parents, both teachers, encouraged me to learn new things all the time. Unfortunately, in communism, there was not a lot to do for a kid who wanted to learn: there was no TV, very few books and only ones that were approved by the state, and generally very few activities outside of school. Being an “intellectual” was a bad thing in the eyes of the government. They preferred people who did not read or think too much. I found great relief in writing, I have been writing stories and poetry since I was about ten years old. I was published with my first poem at 16 years old, in a national literature magazine.
Can you share with us the ‘backstory’ of how you decided to pursue a career path in AI?
I studied Computer Science at university. By then, communism had fallen and we actually had received brand new PCs at the university, and learned several programming languages. The last year, the fifth year of study, was equivalent with a Master’s degree, and was spent preparing your thesis. That’s when I learned about neural networks. We had a tiny, 5-node neural network and we spent the year trying to teach it to recognize the written letter “A”.
We had only a few computers in the lab running Windows NT, so really the technology was not there for such an ambitious project. We did not achieve a lot that year, but I was fascinated by the idea of a neural network learning by itself, without any programming. When I graduated, there were no jobs in AI at all, it was what we now call “the AI winter”. So I went and worked as a programmer, then moved into management and project management. You can imagine my happiness when, about ten years ago, AI came back to life in the form of Machine Learning (ML).
I immediately went and took every class possible to learn about it. I spent that Christmas holiday coding. The paradigm had changed from when I was in college, when we were trying to replicate the entire human brain. ML was focused on solving one specific problem, optimizing one specific output, and that’s where businesses everywhere saw a benefit. I then joined a Data Science team at GEICO, moved to Capital One as a Delivery lead for their Center for Machine Learning, and then went to Amazon in their AI/ML team.
Can you tell our readers about the most interesting projects you are working on now?
While I can’t discuss work projects due to confidentiality, there are some things I can mention! In the last five years, I worked with global companies to establish an AI strategy and to introduce AI and ML in their organizations. Some of my customers included large farming associations, who used ML to predict when to plant their crops for optimal results; water management companies who used ML for predictive maintenance to maintain their underground pipes; construction companies that used AI for visual inspections of their buildings, and to identify any possible defects and hospitals who used Digital Twins technology to improve patient outcomes and health. It is amazing to see how much AI and ML are already part of our everyday lives, and to recognize some of it in the mundane around us.
None of us are able to achieve success without some help along the way. Is there a particular person who you are grateful for who helped get you to where you are? Can you share a story about that?
When you are young, there are so many people who step up and help you along the way. I have had great luck with several professors who have encouraged me in school, and an uncle who worked in computers who would take me to his office and let me play around with his machines. I now try to give back and mentor several young people, especially women who are trying to get into the field. I volunteer with AnitaB and Zonta, as well as taking on mentees where I work.
As with any career path, the AI industry comes with its own set of challenges. Could you elaborate on some of the significant challenges you faced in your AI career and how you managed to overcome them?
I think one major challenge in AI is the speed of change. I remember after spending my Christmas holiday learning and coding in R, when I joined the Data Science team at GEICO, I realized the world had moved on and everyone was now coding in Python. So, I had to learn Python very fast, in order to understand what was going on.
It’s the same with research — I try to work on one subject, and four new papers are published every week that move the goal posts. It is very challenging to keep up, but you just have to adapt to continuously learn and let go of what becomes obsolete.
Ok, let’s now move to the main part of our interview about AI. What are the 3 things that most excite you about the AI industry now? Why?
1. Creativity
Generative AI brought us the ability to create amazing images based on simple text descriptions. Entire videos are now possible, and soon, maybe entire movies. I have been working in AI for several years and I never thought creative jobs will be the first to be achieved by AI. I am amazed at the capacity of an algorithms to create images, and to observe the artificial creativity we now see for the first time.
2. Abstraction
I think with the success and immediate mainstream adoption of Generative AI, we saw the great appetite out there for automation and abstraction. No one wants to do boring work and summarizing documents; no one wants to read long websites, they just want the gist of it. If I drive a car, I don’t need to know how the engine works and every equation that the engineers used to build it — I just want my car to drive. The same level of abstraction is now expected in AI. There is a lot of opportunity here in creating these abstractions for the future.
3. Opportunity
I like that we are in the beginning of AI, so there is a lot of opportunity to jump in. Most people who are passionate about it can learn all about AI fully online, in places like Open Institute of Technology. Or they can get experience working on small projects, and then they can apply for jobs. It is great because it gives people access to good jobs and stability in the future.
What are the 3 things that concern you about the AI industry? Why? What should be done to address and alleviate those concerns?
1. Fairness
The large companies that build LLMs spend a lot of energy and money into making them fair. But it is not easy. Us, as humans, are often not fair ourselves. We even have problems agreeing what fairness even means. So, how can we teach the machines to be fair? I think the responsibility stays with us. We can’t simply say “AI did this bad thing.”
2. Regulation
There are some regulations popping up but most are not coordinated or discussed widely. There is controversy, such as regarding the new California bill SB1047, where scientists take different sides of the debate. We need to find better ways to regulate the use and creation of AI, working together as a society, not just in small groups of politicians.
3. Awareness
I wish everyone understood the basics of AI. There is denial, fear, hatred that is created by doomsday misinformation. I wish AI was taught from a young age, through appropriate means, so everyone gets the fundamental principles and understands how to use this great tool in their lives.
For a young person who would like to eventually make a career in AI, which skills and subjects do they need to learn?
I think maybe the right question is: what are you passionate about? Do that, and see how you can use AI to make your job better and more exciting! I think AI will work alongside people in most jobs, as it develops and matures.
But for those who are looking to work in AI, they can choose from a variety of roles as well. We have technical roles like data scientist or machine learning engineer, which require very specialized knowledge and degrees. They learn computing, software engineering, programming, data analysis, data engineering. There are also business roles, for people who understand the technology well but are not writing code. Instead, they define strategies, design solutions for companies, or write implementation plans for AI products and services. There is also a robust AI research domain, where lots of scientists are measuring and analyzing new technology developments.
With Generative AI, new roles appeared, such as Prompt Engineer. We can now talk with the machines in natural language, so speaking good English is all that’s required to find the right conversation.
With these many possible roles, I think if you work in AI, some basic subjects where you can start are:
- Analytics — understand data and how it is stored and governed, and how we get insights from it.
- Logic — understand both mathematical and philosophical logic.
- Fundamentals of AI — read about the history and philosophy of AI, models of thinking, and major developments.
As you know, there are not that many women in the AI industry. Can you advise what is needed to engage more women in the AI industry?
Engaging more women in the AI industry is absolutely crucial if you want to build any successful AI products. In my twenty years career, I have seen changes in the tech industry to address this gender discrepancy. For example, we do well in school with STEM programs and similar efforts that encourage girls to code. We also created mentorship organizations such as AnitaB.org who allow women to connect and collaborate. One place where I think we still lag behind is in the workplace. When I came to the US in my twenties, I was the only woman programmer in my team. Now, I see more women at work, but still not enough. We say we create inclusive work environments, but we still have a long way to go to encourage more women to stay in tech. Policies that support flexible hours and parental leave are necessary, and other adjustments that account for the different lives that women have compared to men. Bias training and challenging stereotypes are also necessary, and many times these are implemented shoddily in organizations.
Ethical AI development is a pressing concern in the industry. How do you approach the ethical implications of AI, and what steps do you believe individuals and organizations should take to ensure responsible and fair AI practices?
Machine Learning and AI learn from data. Unfortunately, lot of our historical data shows strong biases. For example, for a long time, it was perfectly legal to only offer mortgages to white people. The data shows that. If we use this data to train a new model to enhance the mortgage application process, then the model will learn that mortgages should only be offered to white men. That is a bias that we had in the past, but we do not want to learn and amplify in the future.
Generative AI has introduced a new set of fresh risks, the most famous being the “hallucinations.” Generative AI will create new content based on chunks of text it finds in its training data, without an understanding of what the content means. It could repeat something it learned from one Reddit user ten years ago, that could be factually incorrect. Is that piece of information unbiased and fair?
There are many ways we fight for fairness in AI. There are technical tools we can use to offer interpretability and explainability of the actual models used. There are business constraints we can create, such as guardrails or knowledge bases, where we can lead the AI towards ethical answers. We also advise anyone who build AI to use a diverse team of builders. If you look around the table and you see the same type of guys who went to the schools, you will get exactly one original idea from them. If you add different genders, different ages, different tenures, different backgrounds, then you will get ten innovative ideas for your product, and you will have addressed biases you’ve never even thought of.
Read the full article below:
Have questions?
Visit our FAQ page or get in touch with us!
Write us at +39 335 576 0263
Get in touch at hello@opit.com
Talk to one of our Study Advisors
We are international
We can speak in:
