THE VALUE OF AI: now and the future (PART 2) AI Failures, Pitfalls, Key Learnings and Success
Litsa Roberts
Principal Consultant | Enterprise Architecture | Enterprise - Digital - AI - Tech - Leadership, Strategy, Architecture, Value, Results | Value-Driven Data & AI Strategy | AI Value | Director | Advisory Boards
This article is second in a 3-part series on THE VALUE OF AI: now and the future — catch up on the first and third in the series.
THE VALUE OF AI: now and the future (3-part series)
(PART 2) AI Failures, Pitfalls, Key Learnings and Success
In this second part of the three-part series on THE VALUE OF AI: now and the future, the article first examines and highlights AI-powered, ML-powered and data-driven transformations and project failure rates and why they fail, then proceeds to provide examples on some epic AI failures and success. It highlights some fundamental AI pitfalls and key learnings for executives, leaders, organisations and industry on the importance of AI strategy, organizational readiness, best practices, quickly adapting and learning from (yours and others) AI mistakes and failures. It explores some AI high performers, their secret sauce and customer, data and AI centricity. It concludes with some crucial key learnings and takeaways for executives and leaders on how to overcome the failure rate, act responsibly and set their organisations on the path towards realising AI value, success,?supercharged growth and profits —?winning with AI —?and what it takes to become an AI high performer —?AI-First, AI-driven —?the AI Factory, Digital Factory.
A lot of enterprises, small to medium-sized businesses, as well as the global technology giants across industries, have invested heavily in adopting AI to their benefit. In fact, if you remove AI from their business, their profitability would instantly collapse. This dependency sheds light on the fact that the growth of artificial intelligence powered technology is rapid and significant.
The road going from data to successful AI-powered, ML-powered and data-driven projects, transformations and enterprise DNA is no straight line. Whilst AI-powered and data science technologies are much improved and advanced now compared to 10 years ago, there is a lot more to improve when it comes to meeting expectations?and real-life implementation of an enterprise AI-powered project. AI operations and processes is one factor but there are many other reasons that lead to failure of AI projects. Following some very bold claims about the business benefits, value and even world-saving?power of Artificial Intelligence (AI), it is not surprising that talk of broken promises and failures is growing. However, has AI failed to deliver, or are we setting our expectations too high?
One might wonder why we do not have a perfect blueprint yet for deploying AI/ML-powered and data driven projects successfully? In reality, there are many variables that go into building effective AI, which makes it difficult to prescribe set steps that will work well every time, for every company. Still, progress is being made in gathering best practices —?mainly through key learnings from success stories and failures —? and with those, common patterns are emerging on what often leads to failure. Companies looking to dive into AI and ML projects are navigating uncharted waters and for this reason, you need the sharpest, experienced but also open-minded minds.
Let's first take a look at some AI failures and success and cover some of the major pitfalls, before we move forward to understanding key learnings, what it takes to become an AI high performer and winning with AI.
AI Failure Rates
Despite massive growth in the data economy and wide acceptance that business leveraging data are more successful than those that do not — AI-powered, ML-powered and data-driven implementations have had a shockingly low success rate. While it’s hard to get a precise number, estimates I have found range from?60%-85% failure rate . A 60–85% failure rate is an unnecessarily dismal record and while there are many reasons why data projects fail — the culprit is often because the project is?spearheaded?by data infrastructure experts who do not have a full understanding of the bigger business goals that the data project serves.
There are just too many AI, ML, big data, data science and data analytics failure examples to cover. Indeed, the failure rates are alarming:
When you take a moment to consider the signs of successful AI-powered examples all around us, these numbers may come as a surprise. We have voice assistants for our phones and homes, optimized online product searches, advanced fraud detection at our banks, and more. Yet as it stands now, you may never see the fruits of the majority of AI endeavours. According to HBR (2019) companies are struggling to scale up their AI efforts. Most have run only ad-hoc projects or applied AI in just a single business process.
Why do AI, ML and data-driven Projects fail?
For many, frustration stretches back much further, with an IBM/MIT Sloan Management Review (2010) study showing?38% of organizations ?lacked the understanding of how to use analytics to make better and faster decisions. Today,?fewer than 25% ?of global organizations have developed an enterprise-wide AI strategy. Yet as companies continue pouring more dollars?into AI tools, it is increasingly important to establish whether investments are worthwhile. According to Lukianoff (2020) as organizations demand ever more from data and the expectations of a world powered by AI / ML comes into focus, the shortcomings of the data systems and processes built during the first phase of this data revolution is the "dirty secret " that no one wants to admit about their overpriced data lake investment. But it is the truth and it’s widespread and it's glaring. One of the main culprits is often because the project is spearheaded by data infrastructure experts who do not have a full understanding of the bigger business goals that the data project serves. Often their credentials are impeccable for building an e-commerce site, or a basic reporting engine but have little understanding of how profoundly different the technical and process requirements are for a data science initiative versus an e-commerce or reporting engine.
Forbes (2019)?recently reported that many executives worldwide have not seen value from AI investments. So with technology promises, some billion-dollar projects are failing to deliver on their promises and I do not want yours to be one of them. AI is tough to get right, but easy to mess up. This does not mean you are doomed and have to pull the plug on your beloved AI after months of hard work. AI doesn’t have to be too complicated or expensive. For most companies that are interested in using AI, there isn’t a clear model to follow. A one size fits all AI?system won’t work for most companies with smaller datasets?than the likes of Facebook, Amazon and Google. The approach to building AI?used by massive internet companies just does not translate — most companies do not have overflowing data they can use to train models.? Many industries and companies need to take a different approach — programming with data, NOT code. Many companies that typically have relatively smaller datasets, face high costs for customizing a system and are scared off by long gaps between pilot and product, otherwise known as Minimum Viable Product (MVP).
Examples of AI Failures
A plethora of technology glitches seems to indicate that AI-powered, ML-powered and data-driven projects are not quite there yet. Although AI, ML and data science is meant to solve problems, as it turns out, it can also create new ones as well. Some of these examples may alarm or amuse consumers but are very embarrassing for the companies involved. Either way, it serves as a reminder of AI vulnerabilities and how the AI technology and organizations still have a long way to go.
Examples of some spectacular AI failures include the following:
AI Pitfalls and Key Learnings
As they journey toward AI, most organizations establish data science teams staffed with people skilled in ML/DL algorithms, frameworks and techniques. Yet, many of those organizations struggle to make their AI/ML-powered and data-driven projects truly relevant to the business, instead failing to get the projects into full production and integrated with existing applications and processes. It is why so many line-of-business stakeholders consider only a small percentage of AI projects to be true successes. Organizations are quickly recognizing that they need a systematic approach to “operationalizing” AI in order to drive AI success. That approach means managing the complete end-to-end AI lifecycle.
Let's look at some fundamental pitfalls and key learnings to consider when adopting and building AI to avoid some of the early and major pitfalls [that often lead to failure] in your journey towards realizing AI potential and value that can positively revolutionize your organisation. Whilst this is not meant to be a definitive list of AI pitfalls and key learnings, these however do cover some of the most common and major pitfalls.
20 Major AI Pitfalls to Avoid and Key Learnings
1 - Lack of [or Incorrect] AI Vision, Strategy, Organizational Readiness and Culture
AI vision, strategy, organizational readiness and culture, along with organizational data and technology governance and policies play a key role in winning with AI. If you want AI transformation, you need to think about the culture of the organization.?It is not just about the use of AI algorithms.?It is not just about hiring AI experts.? It’s the mindset of the company.?It’s the culture.?It’s about embedding AI at the very core of the company.? A lack of OR incorrect AI vision, strategy and lack of organizational readiness, mindset and culture can cause the efficiency drop and let your AI investment go to waste, along with the technology failure. If you want artificial intelligence to work well, it is necessary to consider key factors that are crucial for business success. The processes of yesteryear are not easy to change. APIs and algorithms have been around for a long time as a way to connect systems, but the idea of AI-powered APIs and algorithms as IT’s digital nervous system or the building blocks for new combinations of software?is relatively new. Enterprises undergoing digital transformation must address not only their data and technology stack, but also the organizational culture and IT culture around AI-powered, ML-powered and data-driven transformation.
The key for executives and leaders is to understand the organizational and cultural barriers AI initiatives face and work to lower them. That means shifting workers away from traditional mindsets which often run counter to those needed for AI. Leaders can also set up AI-powered projects for success by conveying their urgency and benefits; investing heavily in AI education and adoption — and accounting for the company’s AI maturity, business complexity and innovation pace when deciding how work should be organized. Executives need to be pragmatic in their approach while implementing AI. They need to understand that it is capital-intensive and that it shows results in the long run. A company needs to set a long-term AI vision, strategy, objectives and adopt AI best practices to ensure success.
"Companies that are succeeding in scaling efforts are more likely than others to apply a core set of practices: they align AI and business strategies, ensure cross-functional collaboration, invest in AI talent and training, empower AI experts with standardized tools, protocols, and methodologies, apply strong data practices, and drive adoption and value"
- HBR (2020)
Frequently, there is little order to an organization’s API efforts, with a dozen or so different business areas all approaching APIs differently. Some business areas embrace “integration-first” ideas, others adopt more modern “ecosystem-first” ideas, and overall, only some of the value housed in an enterprise’s systems ends up maximized.
AI/ML-powered and data-driven transformation also has a talent gap problem, which means companies often have to scramble to recruit team members with the right skillsets for building effective AI. Most organizations are currently ill-designed to support scalable AI ventures, requiring re-orgs, new hiring efforts and leveraging of third-party resources. Scaling calls for embedding multi-disciplinary teams throughout the organisation — with clear sponsorship from the top. For AI high performers, these teams are most often headed by the Chief AI, Data or Analytics Officer, and they comprise data scientists; data modellers; machine learning, data and AI engineers; visualization experts; data quality, training and communications and importantly blended with business and process owners and business translators.
Having the right team and the right mix is critical. AI should be designed collaboratively with business and process owners — not just a siloed AI or data science or data scientists project. AI success no longer hinges on just a group of data scientists anymore — blended teams with business skills become more important for driving business value. The key area of emphasis is about translators: people who can make the connection between the business and the technical side.
Perhaps the bigger issue behind the failure of AI projects is the absence of an enterprise-wide AI strategy. The IDC (2019) findings revealed that only 25% of enterprises have developed these blueprints for success. If you skip the strategy and organizational readiness part,?and you will just become another one of the AI failure statistics.
On organizational readiness — to take advantage of AI and ML, your organization must make sure it has all of the following:
Organizations that focus on building a strong foundation will reap more value from their AI investments in the years ahead. That starts with tidying your data drawer.
2 - Unrealistic Expectations
Setting?unrealistic expectations?will line your organization up for?failure. Know that AI / ML projects often can and do fail - failure rates range from?60%-85% failure rate ; Gartner (2017) estimates that?85% of big data projects fail .
There is no point setting a goal that involves AI-powered algorithms never making mistakes or demonstrating no bias because when a mistake does occur or bias is detected, trust will diminish. By setting achievable expectations, organizations can have an honest debate about what AI-powered success looks. This way, by being realistic about AI’s potential but also its limitations, and by developing an AI strategy and processes to accurately manage expectations, organizations can create a positive AI-powered experience for its stakeholders and build trust in an AI/ML-powered and data-driven algorithmic approach to business — even as they continue to learn from the mistakes being made along the way.
"To help accurately manage expectations and create the desired result, organizations should be realistic about AI’s potential as well as its limitations"
- KPMG (2019)
A one-size-fits-all AI?system will not work for most companies. There is no single cookie cutter method of extracting value from AI. Artificial intelligence is good, but it is not a miracle. So if you do not want to have an epic failure, make sure you do everything possible to avoid the most common mistakes.?
Organizations also need to set realistic timelines for AI projects and ensure the desire to push forward with a popular technology does not overrule realistic drawbacks and planning. The hype itself can be a problem, alongside other logistical and strategic challenges.
“AI projects face unique obstacles due to their scope and popularity, misperceptions about their value, the nature of the data they touch and cultural concerns"
- Gartner (2019)
3 - Lack of Project Planning and Due Diligence
A lack of planning and due diligence with AI projects will lead to failure. Clearly define the purpose, knowledge goals and value of the AI project —?if you can’t summarize the end goal of your AI in one sentence, then it’s not clear enough. Successful data-driven projects start with a clear vision of a specific business problem that needs to be solved and are driven forward by teams that are aligned in that vision from top to bottom.
Planning and asking the right questions upfront for an AI/ML-powered, and data-driven project is key.? Developing an AI product is resource-intensive, and for many organizations, it is challenging to allocate in-house time and talent to work on every facet.? Irrelevant KPIs + small impact —?many organizations get swept up in fun and interesting projects only to find out the business value is centered around vanity or meaningless metrics. Determine relevant KPIs and what would you consider as a goal attained upfront. The MVP, deployment?process and MLOps tools needed to support it are also a central part of the planning.?
4 - Lack of Proven Methodology and Best Practices
Follow a proven methodology — AI is not something you want to improvise as you go. Following a tried and tested methodology will?ensure your data science project is on the right path, reliable and successful. The most common methodologies are SEMMA and CRISP-DM. Keep in mind however, one size fits all AI?system won’t work — start with AI Vision, Strategy, Planning, learning goals and organizational readiness, then the data (before moving to software, AI models and technology). Organizations are recognizing that they need a systematic approach to “operationalizing” AI in order to drive AI success. That approach means managing the complete end-to-end lifecycle of AI.
5 - Focusing on Automation
One of the most common mistakes with AI is to focus on automation rather than augmentation of human decision making and interactions.? If organizations focus only on further automation via AI, they miss the hidden opportunities for greater value — personalization and?differentiation. AI can augment humans, as it can classify information and make predictions faster and at higher volumes than humans can accomplish on their own. Gartner (2019) highlights that organizations should look for critical business points?where human interaction or human expertise adds value. They should find examples where such value is manifested in very large amounts of data, especially where the data includes the outcomes that they desire to affect — where customer interactions record whether the customer’s experience was positive, whether a purchaser added an item to a cart, or whether a brake disc was revealed to be worn the predicted amount. They then should consider how AI might augment those efforts to create even more value.
6 - Misalignment — Not knowing the Business Problem
To determine relevant KPIs, you need to know your business — this may sound trivial, but many companies fail as they cannot articulate the business problem they are solving. You must be able to explain it clearly, fast and so that anyone can understand it. Many organizations also choose the wrong problem to solve. They may select something that is too general, resulting in a model that is useless for specific business use cases. They might choose a problem without enough data to support the solution. They also might choose a problem that may be better solved through something other than AI. Any time there is a misalignment with business priorities, problems will occur. It is also important to ensure all stakeholders, from the top down, are clear on the objectives of the project.
Go out and talk to your customers. Find a REAL problem to solve for them. Get leadership agreement on what success means before starting on building data pipelines and designing AI-powered algorithms and models. If you are a leader, then you need to create an environment that fosters cross-department and direct customer engagement throughout the creation of your new data product, AI/ML or analytics service.
Guide yourself through this thought process:
As you are going through this exercise, think of return on investment (ROI). ROI can be determined by looking at how much you stand to make or save if the project succeeds.
7 - Lack of Project and Knowledge Goals
Before being driven by the data, it is important to start with the AI vision, strategy and organizational readiness — followed by program/project and knowledge goals first — then the data — in that order, not the other way around. At the organizational/program/project level, when you start by defining project goals by the knowledge/facts that you are seeking and align the systems and team around the knowledge goals rather than being driven first by the data that you’re putting into it, you will get to an answer faster, cheaper and more persistently.
8 - Too Big, Too Complicated
AI doesn’t have to be too complicated or expensive - start small, win big. Think about where real added value begins. There is definitely the risk to want more from?AI than you actually need. The initial idea of your AI?project should be the overall vision. Take it as a compass for your long-term goal. To guarantee quick and satisfying results, try to start with a tiny project instead of having an overwhelming monster project. The goal is to?create a minimum viable product?(MVP) which can be deployed to production. The more complex, the higher chance of failure!
9 - Lack of Sufficient Quality Data
AI is fundamentally dependent on data. It is the oil that powers the cogs of the machine (learning). Basically, if a system does not have enough data, it cannot draw any inferences (and make good recommendations) until it does. That is a problem and quite a big one. Companies should focus on gathering high quality data, shifting the focus of their data science/engineering corps away from model-centric?approaches and make the deployment?process and MLOps tools needed to support it a central part of the planning?for any AI, ML or data-driven?project. AI is data-driven. Find data from a trusted source. There is no other way around it, to create AI,?machine learning algorithms need data.
Before moving any further, you have to define how much data you need and how you intend on getting it. Most companies typically have relatively smaller datasets. Companies should focus on gathering high quality data — quality data is the key. Asking the right questions when trying to implement an AI or ML-powered project is also key.? Finding the root of goldmine data. To make accurate predictions, AI models need a massive amount of data. These models must be trained to handle any potential use case they will face in production, which means datasets must cover a wide range of use and edge cases. Many companies fail to collect the appropriate amounts of data for their models and have poor data management techniques for accurately labeling that data. This results in poor decision-making by the model.
10 - Complex Area of its Application
It might be a reason that the system under consideration is highly complex and need data that is difficult to obtain. Sometimes, the results obtained should be highly accurate to develop a precise algorithm. For instance, the usage of AI techniques for the medical industry, law, and other complex industries will be complicated. It requires active human minds, efficient workforce and enough information to develop an accurate system.
11 - Expecting the AI to do all the work
One of the great advantages of AI is the flexibility of its processing power. But you cannot just throw in whatever data you want and hope that machine learning will sort it out. In that respect, current-generation AI is much like a human — garbage in, garbage out. Additionally, machine learning-led content can also have a tendency to career off into the extreme.
Organizations now have access to more data in the present time than ever before. However, datasets that are applicable to AI applications to learn are really rare. The most powerful AI-powered machines are those that are trained on supervised learning. This kind of training requires labeled data. Labeled data is organized to make it understandable for machines to learn. One more thing about labeled data is that it has a limit. In future, automated creation of increasingly difficult algorithms will only worsen the problem. Organizations need to do some level of manual data processing before firing up the machine. That may mean having a team of expert curators who are well-versed in their fields tagging quality content based on skills.
12 - Data Noise
Organizations must recognize the importance of core data as the foundation for scaling AI. AI high performers tune out “the noise” surrounding data. They recognize the importance of business-critical data, identifying financial, marketing, consumer and master data as priority domains and are much more likely to wield a larger, more accurate dataset.
13 - Lack of Trust and Governance
Creating the right checks and balances to manage expectations accurately, and so build trust, is a multi-step process of good governance. It involves the operational layer where the data scientists developing the AI are subject to regular peer reviews and risk assessments. It also involves risk management control protocols to establish what level of algorithmic risk is acceptable and what that looks like in regard to business performance and reputation.?
14 - Software-Based Focused or Too Focused On Building the Right AI Model
Organisations should shift their focus away from building the right AI model — a?software-focused?approach — take a different approach: programming with data, NOT code. Focus on getting?good data. Explainability of the machine learning algorithms with the theory is ESSENTIAL — hence programming with data, NOT code (and NOT the other way around).
Many data scientists unfortunately learn machine learning algorithms using a software-based approach (they applied algorithms first, then explored how they work through the articles and open-source documentation and reverse engineering) — major pitfall and can lead to many failures. Data Scientists need to understand the theory and math behind the algorithms.? It is also important for business leaders to also take note (at a high level) the difference between algorithms and APIs. An algorithm is a way to do something. Algorithms can give you the exact result for any input values (eg, exact result for the sum of 1/3 and 1/7). APIs are a sequence of instructions; the result from an API is limited to declaration of variables and will be an approximation of this result.
15 - Focused on AI Technology and/or Infrastructure
As mentioned previously, one of the main culprits for AI failures is often because technology is the first consideration (without an AI organizational strategy, proper planning, project and learning goals, without due diligence) and/or the project is spearheaded by data infrastructure experts who do not have a full understanding of the bigger business goals that the data project serves. Building an AI infrastructure is a strategic decision where you have to consider things like data storage, computing resources, budget and time.
"Many companies that have spent years developing AI technologies are facing the stark reality that successfully scaling AI requires more than just deploying AI technology"
- HBR (2020)
Here are some questions that you also need to answer before selecting any technology and infrastructure that will properly support your AI projects:
Answer these questions now so you do not have to deal with switching infrastructures later.?
16 - Bad Engineering
It is tough to spot a particular issue while detecting the reasons for failure in the AI system. However, faulty engineering leads to wrong neural network settings, even when the data is accurate. But the many AI failure examples are about highly responsible companies; they can afford the best engineers.
17 - Bias
A big problem with AI systems is that their level of goodness or badness depends on the much data they are trained on. Be honest about AI bias. Most companies do not set out intending to create biased models, but do so accidentally by failing to include diverse perspectives and data in their processes. AI insights are only as good as the quality of data algorithms have to draw on. Currently, a great deal of the data AI-powered algorithms depend on come from human activity. As a result, that data will have an inherent element of human bias built in and could compromise the supposed agnostic analysis that AI promises.
AI needs to be trained before it can be useful. Test and validate your model and data. This means running your AI application through a training dataset so it can create a model, then running it again on an entirely new set to test the accuracy of results. Sounds simple in theory, but there are dangers such as data bias which results in bad functionality (and bad press). As mentioned, "it makes business sense to know what decisions you're making and why". In future, such biases will be more highlighted as many AI systems will continue to be trained to utilize bad data. Hence, the urgent need in front of organizations working on AI is to train these systems with unbiased data and create algorithms that can be easily explained.
Starting small makes the project more manageable, especially since you have to measure, correct and improve the AI model over time. A minimum viable product (MVP) increases the return on investment (ROI).
18 - Model Drift and Inaccuracies
The accuracy of your AI should be decided upon a calculation that compares the cost of better accuracy with the expected benefit. Constantly monitor and retrain your model — once you have an AI model that is finally trained and validated, it can be tempting to lean back and call it a day. But?the reality — what your model monitors is dynamic, which means your model should be too. Of the projects that are deployed successfully, many face challenges with model drift — or changing external conditions — that lower the model’s accuracy or even make it obsolete. Models must consistently be retrained with new, relevant data to overcome this hurdle.
19 - Bad Training Data and Results
Quality of training data is everything — the training data for your AI model consists of two parts: data and labels. Training data represents the complete truth from which an AI model learns, thus it determines the maximum quality of the resulting model. Always remember: good training data does not guarantee good results, but bad training data guarantees bad results.
20 - Unknowns
BEWARE — there will always?be “known unknowns” and?“unknown unknowns”?in the AI/ML powered and data-driven world. Even if AI and data scientists and experts spent decades in this field — there will be “known unknowns” and “unknown unknowns” since the field itself is constantly evolving and its applications widen. The key to reducing these is to understand the theory behind the machine learning algorithms.
Source: Towards Data Science (2020)
AI Success - Learning from Mistakes and Failures
Despite many incomplete AI promises and failures which are irritating, it is essential to think that all failures are not wrong in reality. We are going through ongoing learning and evolving on AI from our mistakes and failures.
“Finding how not to do something might be a success. It’s relevant in the world of AI and data; so, we need to be careful in broad-brushing failures”
- Wayne Butterfield, Director Cognitive Automation and Innovation, ISG
With all of these failures, major pitfalls and key learning factors in mind and the many additional aspects that are not highlighted, it may be clearer now how difficult it is to deploy (and maintain) AI-powered, ML-powered and data-driven projects and enterprise successfully. All projects come with a risk of failure, but with the right team, knowledge, planning (purpose, value and learning objectives) and strategies (and importantly), by continuously learning from best practices, as well as yours and others mistakes/failures and quickly adapting, you will set your organization on the right path towards success. Overcoming the failure rate will be challenging for companies large and small, but the future of AI is looking brighter for a number of reasons.?
If you are not making mistakes, then you are not really trying. Failure is so important for mastery because it reminds you of your own willingness to go past your own capacity. It lets you know that you are reaching and not just trying to arrive, this is the journey and evolution of AI mastery — it is about the reach — about the constant reach.
AI Success - Learning from AI High Performers
As highlighted in PART 1 of this three-part series, there is a significant gap between AI high performers and others (the rest) on realizing and maximizing value from AI.
A 2020 report from Grand View Research (2021) suggests that the AI industry ?will see a compounded annual growth rate of 42.2% between 2020 and?2027.?
Some further statistics from Grand View Research (2021):
These statistics speak loud about how AI is leading us into the future with the market expected to grow at a CAGR of 36.62% till 2025. Currently, the race to develop the best AI tools is one of the highlights of the tech world, from the battle between web browsers to fighting terminal illnesses.
"The gap between AI winners and losers continues to grow in nearly every sector. For those already behind, it's time to adopt the critical strategic, organizational and leadership behaviours of today's AI winners"
- Shervin Khodabandeh,?Senior Partner and Managing Director, BCG
Three of the biggest tech giants that are AI high performers making waves right now are Amazon, Google and Microsoft who have their respective visions for artificial intelligence. Let's take a closer look at their AI strategy, projects, products and their market share to understand why they are AI high performers, what their secret sauce is and who is leading the AI race.
While there are no clear winners, these three companies are far ahead of many companies when it comes to leveraging AI. While Google, Amazon, and Microsoft focus on providing tools and solutions for enterprises and their business model, Microsoft also has an eye on using AI for environmental purposes for a better future. It will be interesting and exciting to see how AI will be utilized in the future by these companies and other AI high performers, with an aim to unleash the full potential of AI.
Here’s how AI high performers Amazon, Alphabet + Google and Microsoft are using artificial intelligence to create better products and services, as well as supercharge their market share and profits.?
Amazon and Artificial Intelligence
Amazon is a trillion-dollar company, thanks to artificial intelligence. Jeff Bezos and Amazon was one of the first companies to build their business around AI and machine learning. Amazon has always had a significant competitive advantage. Amazon significantly invested, adopted and embedded narrow AI, ML and robotics early on across its entire business and business model and has not stopped its growth, innovation and sky-rocketing profits and market domination ever since - all powered by AI. Amazon is a company that has reorganized and restructured itself to leverage artificial intelligence in every part of the organization — integrating AI from Top to Bottom. Amazon’s recommendation engines are now driving 35% of total sales. Not only has it been using AI to enhance its customer experience but has been heavily focused internally. From using AI to predict the number of customers willing to buy a new product to running a cashier-less grocery store, Amazon's AI capabilities are?designed to provide customized recommendations to its customers.
"I would say, a lot of the value that we’re getting from machine learning is actually happening beneath the surface. It is things like improved search results. Improved product recommendations for customers. Improved forecasting for inventory management. Literally hundreds of other things beneath the surface"
- Jeff Bezos, Amazon Founder and Executive Chairman?
So, what is Amazon’s secret sauce, besides being an early adopter and being powered with AI across the entire business? How have they integrated artificial intelligence into their business so successfully? Here is the key: Amazon uses a “flywheel” approach to artificial intelligence. “Flywheel” is an engineering term that describes the way companies can conserve energy and keep up momentum. Flywheels are similar to potter’s wheels — we can press a pedal and keep them running consistently by adding a tiny amount of energy at a time. Like a flywheel, using artificial intelligence takes a lot of energy to get started — but once the wheel begins turning, it is far easier to keep it going by giving it continuous smaller boosts. Artificial intelligence has its own momentum and Amazon created a plan to keep up that momentum within that organization, so their efforts never lagged. This is a long-term strategy that ensures maximum benefit from AI efforts.
Amazon’s entire organization is constantly humming with artificial intelligence, and founder Jeff Bezos mandated that data is shared across the organization, not hoarded in one department or process. Datasets are always connected to other data in the organization, to make sure they can be externalized from the ground up. The flywheel of continuous data and AI keeps different parts of the Amazon engine going, and innovations that take place in one department or team can be transferred to other parts of the organization.
Amazon also has an impressive 97.26% grant rate of its patents. Amazon’s biggest patent share can be found in Logistics and Artificial Intelligence.?
"Amazon has filed 14,316 patent applications at USPTO so far (Excluding Design and PCT applications). Out of these 12,764 have been granted leading to the grant rate of 97.26%. Not only this, the application Abandonment/Rejection rate of Amazon is meager at 2.47% and only 299 applications have faced abandonment or rejection".
Amazon's AI Products?
1. “Anticipatory shipping” is amazon’s patented feature that allows products to reach your nearest possible location before you actually place the order to purchase the product. This forecasting is powered by AI which is also the underlying technology for its Prime Now service that facilitates one-hour deliveries. Amazon calculates the number of drivers required to make deliveries using an app called Flex that depends on AI. The app considers many factors like the number of packages in the same locality, the weight of the package, etc to make every minute count.?
2. Amazon also uses an AI-powered dynamic pricing algorithm to get an edge over competitors. The algorithm uses AI to enable optimal sales and revenue automatically by decreasing the prices of the products to increase sales when it’s needed and vice-versa.?
3. Amazon’s AI-backed recommendation engines generate 35% of the total revenues. This algorithm uses data from customer’s previous purchases, preferences, browsing history, search history to create a list of personalized products that a customer will likely buy.
4. The company’s AI-powered sampling strategy uses infrastructure and product purchase data to identify products that each customer is likely to buy. The company then sends samples of new products to customers that chosen by machine learning models. This has been implemented on Prime subscribers.?
5. Amazon’s check-out free physical stores have AI cameras and sensors that charge a customer automatically when they walk out of the store with the product using the Amazon Go App.?
6. Alexa is also powered by AI and has helped many companies add value to their customer service.?
Jeff Bezos net worth in 2009: US $6,800,000,000 (US $6.8 Billion)
Jeff Bezos net worth in 2021: US $191,300,000,000 (US $191.3 Billion)
Alphabet + Google and Artificial Intelligence
Alphabet is a conglomerate including Google which mainly consists of?search, maps, YouTube, Chrome, Cloud, AdWords, AdSense and the Android mobile phone platform. Beyond Google, Alphabet consists of mostly moonshots and investments: Google X, Calico, Nest, Ventures, Fiber, and Capital. AI is integral to Alphabet and Google. As Alphabet’s executive chairman Eric Schmidt?summarizes , he was not optimistic about the potential of?AI in the early 2000s.?However after hiring leading AI researchers including Andrew Ng and acquiring?Deepmind in?2014 ?Google is currently one of the leaders in the field of AI.
Google was relentless in its pursuit of artificial intelligence even before the current wave of AI commercialization took off. Google did two things well earlier on that are paying dividends today: (1) threw a ton of resources into the problem — both computing power and money — and; (2) it scooped up top researchers in the artificial intelligence field.
“There were inklings, early on, that Google was an AI company pretending to be a search company”
Google gradually went from a search engine to an US $800 billion AI powerhouse. Let's take a closer look at the sectors where Google is forging ahead.
Alphabet's AI Centricity
Alphabet wants to be mile-wide and mile-deep in artificial intelligence.
领英推荐
“Across the company, machine learning and artificial intelligence (AI) are increasingly driving many of our latest innovations, from YouTube recommendations to driverless cars to healthcare diagnostics”
— Alphabet 10K filing, 2018
To understand current state of Alphabet’s AI strategy one needs to look at:
Most of Alphabet's companies serve one of large industries like Healthcare that are?experiencing an AI?revolution. Another large group of companies are active as AI enablers.?Like Deepmind or api.ai they aim to build the building blocks of AI including systems, libraries and APIs.
Google’s own ventures are specifically focused on AI enablers, transport/logistics and healthcare.
Instead of diving into all of Alphabet’s bets in the field of AI, let's look at Deepmind, Waymo and Nest which are Alphabet’s leading investments leveraging AI. Deepmind is at the heart of Google operations: Google collaborates with Deepmind to solve some of their engineering challenges like energy efficiency. Their collaboration model is explained?here :
Alphabet’s AI Investments in 38 Companies (2018)
AI is embedded in Google's core. AI is starting to power all Google products. Central to Alphabet and Google's strategy is shifting from mobile-first to AI-first. Right from Google's search engines, Google Maps, Google Photos, to YouTube and Gmail Smart Reply, AI is integrated inside all of the Google apps we use on a daily basis. Google's AI strategy is?to create strong capabilities and patent positions in core data science and computer technology areas. Google's AI technology and patents will have significant implications across a wide range of technologies and business. Google has intensified significantly it’s AI innovation and patents effort in the last few years. Google ranks currently among three largest AI patentees. Google’s AI innovation ecosystem is heavily concentrated in the Silicon Valley, but includes hubs also in the UK, Switzerland and China.
From smartphone assistants to image recognition and translation, a myriad of AI functionality hide within Google apps that you daily use. Here it is mapped leveraging Smart Faktory’s?Google Strategy Framework .
Google continues to innovate in AI.
Microsoft and Artificial Intelligence
Microsoft is among the companies that have been investing heavily into AI. Microsoft's vision for AI can be summarized as follows: “Our vision for the?enterprise is to enable every company to transform by bringing AI to every application, every business process and every employee”. ... Developing new “AI-first” business areas through cutting-edge research and development.
"We are now witnessing a new shift in computing: the move from a mobile-first to an AI-first world"
- Sundar Pichai, CEO, Alphabet
Microsoft Research AI, which is an organization, founded by Microsoft and focused on AI research and development. Because the company already deploys AI in its processes (Skype chatbots, data analysis, interaction with Cortana, etc.), no wonder Microsoft plans to grow in this direction and expand its use of AI.
Apart from that, Microsoft has been launching its AI-driven tools through Azure cloud computing service and working on AI implementation into Office 365.
Microsoft is gradually entering partnerships with the more successful startups. And when the time is right, it will acquire the company that gives it the best leverage in the market. By 2018, Microsoft acquired five AI tech companies, which included LinkedIn acquisition and also?XOXCO : a software product design and development studio. We can also see this exact cycle with Microsoft's most recent acquisition earlier this year, Nuance. Microsoft recently announced its $19.7 billion acquisition of?Nuance, a company that provides speech recognition and conversational AI services. Nuance is best known for its deep learning voice transcription service, which is very popular in the healthcare sector. As Microsoft evolved from being Nuance’s cloud provider to its partner to its owner. And this evolution tells us a lot about Microsoft’s AI strategy, which I think is very smart, given how fast things can change in the AI industry.
The two companies had already been?working together closely before the acquisition . Nuance had built several of its products on top of Microsoft’s Azure cloud. Microsoft had been using Nuance’s Dragon service in its Cloud for Healthcare solution, which launched last year in the midst of the pandemic. The acquisition is Microsoft’s biggest since the $26 billion purchase of LinkedIn. And it tells a lot about Microsoft’s AI strategy. “Nuance provides the AI layer at the health care point of delivery and is a pioneer in the real-world application of enterprise AI,”?Microsoft CEO Satya Nadella ?said. “AI is technology’s most important priority, and health care is its most urgent application”.
Microsoft developed the AI 360° model after conducting a vast study in partnership with consultants EY, to better understand where and how AI will be used—and is being used—by businesses. Together, the duo obtained responses from nearly 300 organizations across seven sectors, creating a model that companies can follow to reap the benefits of embracing AI. The model comes from the identification of three key and interlinked elements of AI business transformation: AI’s key benefits, its functional use cases, and the capabilities required to enable AI within an organisation.
These have been visualized as three connected circles – emphasizing the 360° approach to planning an AI strategy. The model allows organisations to consider their AI maturity within different areas of their infrastructure and lays out the key technological and human factors that need to be considered in any program of innovation or change.
In short, the model can be used:
All in all seems like Microsoft takes the Artificial Intelligence technology beyond serious and plans to offer real value for the Microsoft product users.
Microsoft has launched many AI solutions across industries like healthcare, retail, education, banking, etc. One of the oldest contenders, here is Microsoft’s AI products and plans.?
1. Microsoft uses AI to fight against cybercriminals by learning from the data of every company that uses its services. The company’s?Azure security team ?customizes security to a client’s online behavior. By this, they were successful at bringing its false-positive rate down from 2.8% to 0.001% by tracking down fake logins.?
2. With its AI initiative project Hanover, Microsoft is bringing a change to the healthcare industry by helping to find the best cancer treatment. A group of researchers have created algorithms to understand how cancer develops and predict the best drug combinations to fight it.?
3. Maritime ships transport 90% of goods across the sea which produces at least 3% of the global carbon emissions. As many companies don’t know how much fuel the ships actually consume, Microsoft’s AI solutions help in determining fuel efficiency with frequent data on climate and vessel speed to reduce fuel charges.?
4. With an aim to help farmers, Microsoft partnered with ICRISAT to AI in increasing the crop yields by 10 to 30% by predicting the best sowing date for crops.?
5. Carlsberg, a leading beer and beverages company, uses Microsoft’s AI solutions to detect various aromas and flavors of beer, and also improve the quality of existing beers.?
Successful AI Startups and Artificial Intelligence
From Silicon Valley to London to Shanghai AI startups are in abundance.? But like any gold-rush there are a lot of prospectors who are making grandiose claims. As London?MMC Ventures noted in a recent AI report over 40% of?2,830 so called AI startups ?in Europe are not really using AI. Of the AI startups that are doing well - what makes these AI startups successful? What is it that has allowed these companies to gain momentum and raise such large sums of money?
AI startups as those that either (1) would not exist if there were no AI modern technologies, such as deep neural networks — it is?core?to their?existence?or; (2) provide AI infrastructure and tools such as AI specialist hardware, cloud services for AI applications, or tools to accelerate the implementation of AI solutions.
The AI startups that are doing well generally understand the nature of AI technology and the opportunity in the enterprise. But more than that AI startups that are starting to scale have all crossed the commercial divide from a technical world to the enterprise.
According to Towards Data Science (2019) successful AI startups have learnt:
It is clear that many AI startups are providing?valuable?point solutions to enterprises and are succeeding as they have access to (1) large and proprietary?data training sets, (2) domain knowledge?that gives them deep insights into the opportunities within a sector, and (3) a deep pool of?talent?around applied AI.
Successful AI startups have really understood the nature of AI technology and how it is bought and consumed. AI is an?enabling?technology — a set of tools, technologies and methods that can be applied to solve a myriad of use cases.
In the 1980s the introduction of SQL databases allowed structured data to be stored and queried in a tabular format. This enabled millions of applications giving rising to the ERP and CRM billion dollar industries. Similarly AI will?enable?a huge number of use cases. Vision algorithms are being used, for example, in nearly every industry from spotting manufacturing defects to recognizing shoplifters to helping autonomous vehicles navigate the streets. Natural language processing can be used everywhere from reviewing customer sentiment in social media to reviewing legal documents for contract completeness to screening resumes and CVs to identifying potential opportunities for financial trading. And speech to text technology is being used in medical transcription algorithms.
As AI is such an enabler many AI startups will be helping to solve problems for corporations and organisations.
AI Startups and the Holy Grail - The Enterprise Flywheel
AI is generally a?scale?game. The more and higher quality the data, the better the AI algorithms which leads to deeper insights, greater productivity, enhanced products and services, and a better customer service. This can lead to more enterprise customers that leads to the collection of more data which leads to deeper insights and so on. That leads to more customers and more financing. There is a real potential for a?flywheel?effect and the creation of defensibility. Successful AI startups have figured this out and have been able to push the accelerator and get further and further ahead of their competition.
AI Value Chain
The playing field for AI is wide and long consisting of AI?makers?— those who make AI technologies — and the?takers?— those who take AI technologies and extract value. The AI value chain — introduces a seven layer value chain — for who will make money in AI. The companies noted are representative of larger players in each category but in no way is this list intended to be comprehensive or predictive.
Source: Towards Data Science (2019)
Let's take a closer look at the seven layers of the AI Value Chain — there are:
(1) AI?chip?and?hardware?makers who are looking to power all the AI applications that will be woven into the fabric of organisations big and small globally;
(2) the?cloud?platform?and?infrastructure?providers who will host the AI applications;
(3) the AI?algorithmic?and?cognitive?services?building?block?makers who provide the vision recognition, speech and deep machine learning predictive models to power AI applications;
(4)?enterprise?solution?providers whose software is used in customer, marketing, HR and asset management and planning applications;
(5)?industry?vertical?solution?providers who are looking to use AI to power companies across sectors such as healthcare to finance;
(6)?corporate?takers of AI who are looking to increase revenues, drive efficiencies and deepen their insights; and finally
(7)?nation states?who are looking to embed AI into their national strategies and become AI enabled countries.
While AI startups are looking to offer new chips, cloud services and algorithms, this area of the AI value chain is dominated by deep pocketed technology giants such as Google, Microsoft and Amazon. They have become the picks and shovels of this gold rush. Whatever AI company is digging for gold the giants want to make sure they are powering that organisation with their AI hardware, cloud and algorithmic solutions.
WINNING WITH AI - Turn AI into ROI
Across?multiple?industries,?organizations?are?starting?to integrate AI into their businesses, making use of of machine learning and predictive analytics to take on tasks that have typically required human intelligence. As highlighted by BCG (2019) but so far, the majority are seeing little to no return on their investments. A survey of more than 2,500 senior executives and other managers from 29 industries in 97 countries conducted by MIT Sloan Management Review and BCG (2019), sheds further light on the challenges of AI:
AI Positive Business Impact Across Industries
AI is creating business impact across industries. Despite the obstacles, many companies report positive business impact from AI across multiple industries.
According to BCG (2019) the industry respondents reporting business impact from AI in the past three years, the results from the survey are as depicted in the diagram directly below.
INDUSTRIES REPORTING BUSINESS IMPACT FROM AI (PAST 3 YEARS)
Source: BCG (2019)
AI Maturity and AI High Performers
BCG (2019) confirms that there is an increasing gap in AI maturity, while many companies are struggling, a significant minority are realizing value from AI and continuing to invest more. The BCG (2019) survey identified four distinct segments of AI users.
Source: BCG (2019)
Gartner (2020) has provided an AI Maturity Model to help organizations to accelerate and optimize their AI strategy and implementations to achieve the best value from AI. Gartner's AI Maturity Model defines five —?within the business —?adoption levels.
THE HOW — Winning with AI — Implementing AI Across The Enterprise
Companies and countries around the globe increasingly view development of strong AI capabilities as imperative to staying competitive. Although AI failure statistics and examples seem troublesome, most enterprises at least understand the potential value of AI. Two-thirds of enterprises are attempting to build an 'AI First' culture and half see the technology as a necessity.
BCG (2021) has identified the following six step plan to drive business value through Artificial Intelligence. The experience of AI pioneers suggests six moves that separate the winners from the rest.
Source: BCG (2021)
THE HOW - AI Roadmap and End-to-End Lifecycle
There are reams of information on the?“what”?of AI, but less about the?“how”. Organizations are starting to recognize the importance of adopting a more systematic approach to “operationalizing” AI in order to drive AI success. That approach means managing the complete end-to-end lifecycle of AI. ?For practical ideas, Accenture (2019) provides an?AI roadmap ?or “route to live”— the steps to take AI projects from?Proof of Concept to production , effectively and expediently. This should be used as a guideline for best practices.
Nail It — Then Scale It. Scaling the exponential power of AI across the enterprise is a journey. Those who scale well stand to re-define, accelerate and multiply value.
Provided below are two proven methods of the "HOW" — an Accenture (2019) AI Roadmap and the IBM (2019) End-to-End AI Lifecycle which can be leveraged as some best practices to help companies to adopt a more systematic approach to operationalize AI.
Source: Accenture (2019)
Source: IBM (2019)
AI-Driven, AI-First Companies and the AI Factory
The AI Factory operationalizes the end-to-end AI Lifecycle to achieve AI at scale and at speed. An AI Factory?combines data, people, process, product and platform to move beyond science experiments and deliver AI that drives business value.
The?virtual?circle?of?data,?algorithms,?service?usage?is?what?we?see?happening?in?all?AI-first?companies.? The age of AI is being ushered in by the emergence of this new kind of firm —?AI-First. HBR (2020) highlights some examples of AI-First companies such as Ant Financial Services Group (spinoff from Alibaba) and include giants like Google, Facebook, Alibaba, and Tencent, and many smaller, rapidly growing firms, from Zebra Medical Vision and Wayfair to Indigo Ag and Ocado. Every time we use a service from one of these companies, the same remarkable thing happens: Rather than relying on traditional business processes operated by workers, managers, process engineers, supervisors, or customer service representatives, the value we get is served up by algorithms. Microsoft’s CEO, Satya Nadella, refers to AI as the new “runtime” of the firm.
The essence of the AI-First company is really the ability to industrialize and scale-up the process which you accumulate, integrate and deploy data —?at the core of that is the AI factory. The AI Factory feeds data and models systematically into a software enabled operating core of the firm.
Source: HBR (2020)
The AI Factory builds on the principles of the AI Ladder, which describes the importance of creating a solid information architecture for sustained AI success.
Source: Cloud Factory (2021)
To maximize AI value, enterprises must evolve from the standalone model to a model factory or AI Factory. For leaders of?traditional firms,?competing?with?AI-first, AI-Driven?digital rivals?involves more than deploying enterprise software or even building data pipelines, understanding algorithms, and experimenting.?It requires rearchitecting the firm’s organization and operating model.
Source: TowardsDataScience (2020)
How AI-First and AI-Driven companies can outstrip traditional firms? The value that scale delivers eventually tapers off in traditional operating models, but in AI-First and AI-Driven digital operating models, it can climb much higher. AI-First companies look and operate very differently than traditional companies and can generate value much faster.
Source: HBR (2020)
BOTTOM LINE
As AI becomes increasingly?embedded deeper?into the fabric of business,?industry,?society, humanity?and?our everyday lives, here are some crucial key learnings and takeaways on AI failures, pitfalls and success for executives, leaders and their organizations and industries —?towards realizing AI value and winning with AI —?and the journey towards becoming an AI high performer.
Key Takeaways for Business and Industry
A Lasting Thought
"AI is not just about technology. It's about people, process, culture and strategy — and, ultimately, redefining the relationship between human and machine"
- BCG Gamma
Share your thoughts on AI|ML-powered and data-driven failures, pitfalls, key learnings and success. Please click below, join the discussion.
SOURCES:
Accenture (2019) - visiontovalue.economist.com/ai-built-to-scale/
AI Multiple (2018) - research.aimultiple.com/alphabet-ai/
AI Multiple (2021) - research.aimultiple.com/ai-is-already-at-the-heart-of-google/
BBC (2018) - www.bbc.com/news/technology-46357004
BCG (2021) - www.bcg.com/en-au/featured-insights/how-to/roi-of-ai
BCG (2019)?- image-src.bcg.com/Images/Final-Final-Report-Winning-With-AI-R_tcm9-231660.pdf
Business Insider Australia (2018) - www.businessinsider.com.au/google-photos-ski-photo-fail-2018-1
CNBC (2017) - www.cnbc.com/2017/08/01/facebook-ai-experiment-did-not-end-because-bots-invented-own-language.html
Data Science Process Alliance (2021) - www.datascience-pm.com/project-failures/
Forbes (2021) - www.forbes.com/sites/forbestechcouncil/2021/08/09/ai-failed-promise-or-a-case-of-unrealistic-expectations/?sh=222054f0312d
Forbes (2019) - www.forbes.com/sites/gilpress/2019/10/17/ai-stats-news-65-of-companies-have-not-seen-business-gains-from-their-ai-investments/?sh=68c681319f47
Gartner (2019) - blogs.gartner.com/andrew_white/2019/01/03/our-top-data-and-analytics-predicts-for-2019/
Gartner (2019) - https://www.gartner.com/smarterwithgartner/the-cios-guide-to-artificial-intelligence/
Grand View Research (2021) - www.grandviewresearch.com/industry-analysis/artificial-intelligence-ai-market
GreyB (2020) - insights.greyb.com/amazon-patents-grant-rate/
HBR (2020) - hbr.org/2020/01/competing-in-the-age-of-ai
HBR (2020) - hbr.org/2020/03/how-high-performing-companies-develop-and-scale-ai
HBR (2019) - hbr.org/2019/07/building-the-ai-powered-organization
IBM (2019) - medium.com/inside-machine-learning/ai-ops-managing-the-end-to-end-lifecycle-of-ai-3606a59591b0
IBM/MIT Sloan Management Review Study (2010) - www.ibm.com/downloads/cas/PQ2ANXL3
IDC (2019) - venturebeat.com/2019/07/08/idc-for-1-in-4-companies-half-of-all-ai-projects-fail/
Kaspersky (2017) - usa.kaspersky.com/blog/voice-recognition-threats/10855/
KPMG (2019) - home.kpmg/xx/en/home/insights/2019/08/managing-expectations-of-an-ai-utopia.html
Lukianoff (2020) - towardsdatascience.com/the-success-resounding-failure-of-big-data-50b3f17756f1
Medium (2019) - medium.com/hackernoon/how-facebook-apple-microsoft-google-and-amazon-are-investing-in-ai-f58b5706e34a
MIT Technology Review (2016) - www.technologyreview.com/2016/01/12/163910/could-ai-solve-the-worlds-biggest-problems/
NetworkWorld (2010) - www.networkworld.com/article/2194368/biggest-barriers-to-business-analytics-adoption--people.html
Politico (2021) - www.politico.eu/newsletter/ai-decoded/politico-ai-decoded-how-cambridge-analytica-used-ai-no-google-didnt-call-for-a-ban-on-face-recognition-restricting-ai-exports/
Reuters (2018) - www.reuters.com/article/amazon-com-jobs-automation/insight-amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idINKCN1MK0AH?edition-redirect=in
TechWalls (2018) - www.youtube.com/watch?v=ZZUtZdJawCs
The Guardian (2018) - www.theguardian.com/technology/2018/apr/16/cambridge-analytica-scandal-highlights-need-for-ai-regulation
TheVerge (2016) - www.theverge.com/2016/3/24/11297050/tay-microsoft-chatbot-racist
ThinkML (2020) - thinkml.ai/five-biggest-failures-of-ai-projects-reason-to-fail/
Towards Data Science (2020) - towardsdatascience.com/3-key-takeaways-from-a-machine-learning-course-4a36030960d5
Towards Data Science (2020) - towardsdatascience.com/model-evolution-from-standalone-models-to-model-factory-5a8e01fa03cb
Towards Data Science (2019) - towardsdatascience.com/the-secrets-of-successful-ai-startups-whos-making-money-in-ai-part-ii-207fea92a8d5
VentureBeat (2019) - venturebeat.com/2019/07/19/why-do-87-of-data-science-projects-never-make-it-into-production/
VentureBeat (2021) - venturebeat.com/2021/04/17/why-microsofts-new-ai-acquisition-is-a-big-deal/#:~:text=Microsoft's%20recent%20shopping%20spree%20reached,recognition%20and%20conversational%20AI%20services.
Wired (2018) - wired.me/technology/artificial-intelligence/microsoft-ai-middle-east/
ZDNet (2020) - www.zdnet.com/article/ai-for-business-whats-going-wrong-and-how-to-get-it-right/
Design Leader
2 年It was a privilege to read this article: Going through the series and trying to fathom the fascinating work you have done.
Principal Product Manager of Data Intelligence Storage
3 年Very interesting. Thx