AI in Action: 5 Striking Examples of The Capabilities of Artificial Intelligence
Artificial Intelligence Has Arrived, But What's Its Next Destination?
Artificial intelligence (AI) is a rapidly developing field that has the potential to change the way we live and work. As an avid chess player, I first became hyper-aware of AI in 2017 when I came across AlphaZero - Google DeepMind’s chess playing artificial intelligence. I remember being astonished at what I saw.
What made AlphaZero so impressive is that it was able to teach itself how to play chess in just four hours, without any prior knowledge of the game. It was able to do this by using a technique called reinforcement learning, which involves training the algorithm by playing against itself. As it played more games (millions more), it was able to improve its own chess skills and ultimately become an order of magnitude better at chess than anything (or anyone!) that came before it.
AlphaZero was put to the test against the world's best chess engine at the time, Stockfish, developed by expert chess players over decades, in a 100-game match in 2017. The result? 28 Wins, 72 draws, and no losses for AlphaZero. Not only that, but it was able to demonstrate its dominance and superiority with an artistic elegance.
The success of AlphaZero is a testament to the power of AI, and it raised many questions about the future of the field at the time. For example, if an AI algorithm can teach itself how to play chess in just four hours, what other tasks could it be capable of performing? How well would AI adapt to the non-determinism of the real world? Could AI be used to solve complex scientific problems or help us to understand the mysteries of the universe?
The possibilities are endless, and it is exciting to think about what the future holds for AI. But it's also important to consider the ethical implications of such powerful technology and ensure that it's used responsibly. In this article, I am going to dig into four other striking examples of AI in action that have taken hold since, and pose some questions about what the future of AI has in store.
First, the one on everyone’s lips the last couple of months:
The launch of chat GPT, a large language model developed by OpenAI, is another example of the disruptive power of AI. This model has the ability to generate human-like text, making it useful for a wide range of applications such as content creation, idea generation, and language translation.
The outputs from this AI are truly striking, and it’s no surprise that ChatGPT has taken the world by storm, garnering 1M users in just 5 days. To put that into perspective, it took Netflix 40 months to achieve this goal, Twitter over two years, and Spotify 150 days. Even instagram took 2 and a half months to reach this milestone.
For a taste of the kind of things ChatGPT can do, take a look at this motivational letter, for a bit of an insight into ChatGPTs emotional and human understanding:
When people think of the notion that “AI will make all of us redundant”, people typically think of the factory worker, data entry clerk, or similar. What’s becoming clear now is that no-one is safe! Not even CEOs, accountants, or management consultants.
The model's ability to understand and respond to natural language inputs makes it a powerful tool for businesses everywhere when implemented effectively.
Here’s a couple more examples of me bouncing some ideas around. You’ll see immediately that the answers are grounded in clarity and logic, and are immediately actionable.
Even software engineers, will be finding that AI will be playing a bigger role in their life, as Chat GPT is already writing, checking, and optimising code in multiple programming languages.
One of the most exciting potential applications of chat GPT is in the field of education. Imagine having a personal tutor that can understand your questions and provide tailored responses, or a learning platform that can generate personalised practice exercises and quizzes. This technology could also be used to create virtual learning environments, which could make education more accessible to people in remote or under-served areas.
Meanwhile, Microsoft are eyeing up a $10B dollar investment in OpenAI (the developer of ChatGPT), with many speculating they will use the deal to accelerate their development in AI, and potentially revolutionise their search engine, Bing, to create a serious rival to google. Could this be the breakthrough that is needed to break google’s grip on our internet searches?
The deal will see Microsoft take a 75% share of profits after the initial partners receive their investment back. Uniquely though, OpenAI is in fact a non-profit company. So beyond returning an agreed profit cap to investors, eventually profits will go solely into the continued development of OpenAI.
A less visible but equally striking application of AI has been taking a hold of social media over the last decade - the rise of the Recommendation Engine.
Artificial intelligence is being widely used by companies like Netflix, TikTok, Instagram, YouTube, and Shein to improve their recommendation engines. These recommendation engines use a variety of techniques such as collaborative filtering, natural language processing, and machine learning to analyse user data and make personalised recommendations to users.
TikTok uses AI to recommend content to users based on their engagement history. The data is so vast in recent years that it is no longer simply looking at likes, shares, comments, or followers. It’s looking at it all. How many milliseconds you look for, which thumbnails generate the most engagement, pretty much every way you interact with social media content is being recorded and analysed by AI. By using this rich data, TikTok's algorithm can make recommendations that are more likely to be relevant to each viewer - keeping them engaged, and monetised through ads.
Similarly, Instagram, Netflix and YouTube all use AI-powered recommendation engines to show users content that is relevant to their interests.
Finally, Shein, a Chinese-owned eCommerce clothing platform, uses AI to recommend products to their customers based on their browsing history, purchase history, and other data. Shein are famous for using data and AI to generate new clothing ideas, and release 1000s of these new items at scale each day.
By analysing data from a huge volume of people, and thousands of product lines, Shein's algorithm can make personalised product recommendations that are tailored to each customer's preferences and interests - again, engaging and monetising them more. In this case the narrow domain expertise of recommending clothing helps the algorithm excel at the task even more.
Two things to note that recommendation engines rely on heavily.
They need a huge amount of data: To become the addictive, all knowing, powers that be today on the big social media platforms, these algorithms feed and live off of data. To maximise their output, the data needs to be big, and high quality. Big Data is a vital developing industry for the enablement of AI. Along with companies that provide the cloud infrastructure and computing power to store, access, and analyse this data.
Diversity of content and recommendations: They also need a large and varied universe of things to recommend in the first place, to truly generate the depth and richness in the machine learning that recommendation engines go through.
This is how social media giants like TikTok and Instagram have created such powerful recommendation engines - they have more data than anyone else from their apps, and their user generated content is in endless supply, giving the machine learning algorithms all they need to discover what to show you next.
AI is also helping social media companies in another way. With AI enabled ad targeting, the social media companies are able to better target advertisements. With the effectiveness of ads increasing, companies are increasingly willing to pay more for ads, driving up advertising earnings across big tech companies like Meta and Google over the last decade.
This is being achieved through the use of advanced machine learning algorithms that can analyse large amounts of data to identify patterns and make predictions about consumer behaviour. By using this data, companies can target their ads to specific segments of the population, increasing the chances of conversions and reducing the cost of advertising. In recent years though, the advertising industry has moved beyond this, into a state where, after a sufficient testing period, AI can improve or even outperform the targeting that is being originally set by the advertisers.
One of the key companies involved in this disruption is Google. Google's AI-powered ad platform, Google Ads, uses machine learning to help businesses target their ads to specific audiences. It uses data from a user's search history, browsing history, and location to show ads that are more likely to be relevant to them. Google's AI also helps to optimise ad campaigns by adjusting bids, placements, and targeting in real-time to maximise returns on ad spend.
Some would argue that more relevant ads for users is a good thing, but others would argue that its influence and power make it something to be weary of - in fact over the last several years there has been major privacy concerns and pushback from the general public.
AI is now being used for image recognition in a wide range of applications, including object detection, facial recognition, and image classification. These applications make use of machine learning algorithms that can analyse images and extract useful information from them.
One of the most popular applications of image recognition is object detection, which involves identifying specific objects within an image or video. This technology is used in a variety of fields, including self-driving cars, surveillance, and retail. Companies like NVIDIA, Waymo (an Alphabet subsidiary), and Tesla use object detection technology in their self-driving cars to identify and respond to objects in the environment, such as pedestrians, other vehicles, and traffic signals.
Another popular application of image recognition is facial recognition, which involves identifying people in an image or video by analysing facial features. This technology is used in a variety of fields, including security, marketing, and entertainment. Companies like Face++, Microsoft (Face API), and Amazon (Amazon Rekognition) are using facial recognition technology in their products and services to improve security and personalisation.
With the further development of sensing and vision systems, it’s easy to imagine a world where surveillance becomes so powerful that security cameras can identify individuals and continually monitor their movements and activities.
Robotics and AI are closely related fields, as AI provides the intelligence and decision-making capabilities that robots need to function autonomously.
One of the most popular applications of AI in robotics is in autonomous vehicles, such as self-driving cars and drones. These vehicles use AI-powered navigation systems to navigate their environment, avoid obstacles, and make decisions about how to reach their destination. Companies like Waymo, Tesla, and Baidu are at the forefront of developing self-driving cars and drones, using AI and machine learning algorithms to enable them to navigate safely and efficiently.
Another popular application of AI in robotics is in industrial automation. In this field, robots are being used to perform tasks that are dangerous or difficult for humans, such as welding, painting, and assembly. These robots use AI-powered systems to analyse their environment and make decisions about how to perform their tasks.
AI is also being used in service robots, such as robots that are used in retail, hospitality, and healthcare. These robots use AI-powered systems to understand and respond to natural language inputs, making them more capable of interacting with humans - and even capable of entertaining them.
As with several of the other AI applications I have shown here, Robotics also has a dark-side. Having seen the way that AI has excelled at thought-tasks, it’s easy to imagine a scenario where robotics and autonomous air vehicles become so advanced that they can dominate any military force which isn’t enabled by AI. If AI does indeed become this powerful, and ends up in the control of the wrong hands - dystopian scenarios are bound to happen.
Where do we go from here?
The future of AI looks promising and has the potential to bring about significant changes in many different fields. But as has been echoed throughout this article, the potential for abuse in the wrong hands is huge. The ethics and morality of the use of AI will continue to become an emerging and important issue of the day, as this technology and its applications evolve.
There are several moral and ethical considerations, including:
Bias: AI systems can perpetuate and even amplify societal biases if the data used to train them is biased. This can lead to unfair or discriminatory decisions.
Privacy: The collection and use of personal data by AI systems raise concerns about privacy and the protection of personal information.
Autonomy: As AI systems become more autonomous, there are questions about how responsible and accountable they should be for their actions.
Transparency: As AI systems become more complex, it can be difficult to understand and explain their decision-making process. This can make it difficult to trust the decisions made by the AI system.
… and several dystopian scenarios that could develop if AI got into the wrong hands or became an unfair advantage that was only available to the wealthy or elite:
Widening of the wealth gap: AI-powered automation may lead to increased economic inequality, as the wealthy gain access to advanced AI technologies that can automate their businesses and create new opportunities, while the less fortunate lack access to these technologies.
Job displacement: AI-powered automation may lead to widespread job displacement, as machines can perform many tasks more efficiently than humans. This could lead to mass unemployment and social unrest.
Loss of privacy: AI systems that are controlled by the wrong people may be used to collect and exploit personal data on a large scale, leading to a loss of privacy for the general population.
Weaponisation: AI systems may be weaponised to target specific groups or individuals, leading to large-scale human rights violations.
Social manipulation: AI systems that are controlled by the wealthy or elite may be used to manipulate public opinion and influence elections, leading to loss of democracy
All these possibilities need to be considered by governments, regulators, and responsible AI developers, to ensure that AI brings about positive change in society, and not destruction.
For AI to thrive, it needs the following conditions:
High-quality data: AI systems need high-quality and diverse data to train on. This data should be free from bias and accurately represent the problem that the AI system is trying to solve.
Robust algorithms: AI systems need robust algorithms that can handle complex data and make accurate predictions.
Strong computational power: AI systems require significant computational power to operate. This includes not only the computational power of the system but also the infrastructure to support it.
Human oversight: AI systems should be designed with human oversight in mind, to ensure that the decisions made by the AI system align with human values and ethics.
Transparency: AI systems should be transparent and explainable, so that their decisions can be understood and trusted.
But even with these conditions, there are still some big limitations on AI.
Limited domain expertise: AI systems are typically designed to excel in a specific domain or task. They may not perform well in tasks outside of their domain expertise.
Computational power: AI systems require significant computational power to operate. This can be costly and may limit the deployment of AI in certain environments.
Lack of common sense: AI systems do not possess the common sense that humans have, which can limit their ability to understand and respond to certain situations. Take this riddle for example. All that computing power and intelligence, yet it can’t answer a simple riddle. It was painful.
While it’s easy to think of dystopian scenarios … there are also an incredible number and variety of scenarios that could bring tremendous good to the world.
Although it is difficult to make definitive predictions about AI in the next 10 to 20 years, there are some possible developments that we can be pretty confident and excited about:
Advancements in natural language processing: AI systems are likely to become more advanced in their ability to understand and generate human-like language, allowing for more natural and seamless human-computer interaction.
Development of more advanced autonomous systems: AI systems are likely to become more autonomous and capable of performing complex tasks with minimal human supervision. This could include self-driving cars, drones, and robots.
Advancements in healthcare: AI systems are likely to become more sophisticated in their ability to analyse medical data and make accurate diagnoses, leading to improved patient outcomes.
Development of AI-powered virtual assistants: AI-powered virtual assistants are likely to become more advanced in their ability to understand and respond to human input, leading to more natural and efficient human-computer interaction.
Advancements in the field of robotics: AI will likely play a crucial role in the development of advanced robots that can perform a wide range of tasks, from industrial automation to service robots.
Advancements in security: AI will likely play a crucial role in the development of more advanced security systems that can detect and respond to threats in real-time.
The overall hope for AI, is that it helps the human race to be more productive and economically successful, improving the quality of life for large segments of the population, and bringing people out of poverty.
So what do I think?
I believe that AI will bring huge productivity gains and quality of life improvements to the human race in the not too distant future, but it would be naive to think that it will never be abused. I see AI abuse as an unfortunate certainty which the world will need to deal with at certain points in the future with global collaboration and regulation of the dangerous aspects of this technology. I also firmly believe that AI will create or strengthen some of the most valuable companies of the future.
Does that mean that AI startups will all be smash hits? Of course not, like any other emerging technology, there will be winners and losers. The challenge of finding the best is one we can leave for venture capitalists (or AI?) to solve.
This Is Not Financial Advice is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.