The Mechanical Turk, an 18th Century chess machine - BBC News

The Turk’s legacy: A glimpse into the challenges of AGI

The allure of machines mimicking human intelligence has captivated us for centuries, similar to our modern-day fascination with AGI. In the 18th century, “The Turk”, an elaborate automaton that played chess, toured Europe, astounding audiences with its seemingly mechanical prowess. It even defeated chess masters like Benjamin Franklin and Napoleon Bonaparte. However, The Turk’s secret was far less impressive – a skilled human chess player hidden inside the machine.

355129390 653347520145897 4412066644471166331 n

Get a job in tech. No matter who you are.

87% of our grads get hired in 6 months or less. Most of them have no STEM or IT experience before starting with us.

This story is a stark reminder of the vast gulf between clever machines and Artificial General Intelligence (AGI), a hypothetical AI capable of human-level thinking across various domains. While modern AI can achieve superhuman feats in specific tasks, like playing chess, it lacks the general intelligence and adaptability of The Turk’s hidden operator.

The illusion of automation: Why AGI isn’t quite ready to take over (just yet)

For decades, science fiction has painted a picture of a future dominated by intelligent machines. However, the reality of Artificial Intelligence (AI) is far less dramatic, and a recent example from retail giant Amazon highlights the significant gap between current technology and the dream of Artificial General Intelligence (AGI).

Amazon’s “Just Walk Out” technology promised a futuristic shopping experience, allowing customers to breeze through stores without waiting in checkout lines. The idea was that sophisticated cameras and sensors would track customer movements and automatically bill them for picked-up items. However, the truth behind the magic was far less high-tech.

A report revealed that Amazon relied heavily on a hidden army of human workers in India – over 1,000 according to some estimates – to manually review a significant portion (up to 70%) of these “automated” transactions. This story is eerily similar to the 18th-century chess-playing automaton, “The Turk”, which duped audiences into believing they were witnessing a machine’s brilliance when in reality, a skilled human player was concealed within.

The Amazon example exposes a crucial limitation of current AI: a lack of true general intelligence. AI excels at specific, well-defined tasks where it can be trained on massive datasets. Just Walk Out worked flawlessly for scenarios the AI could anticipate.

However, the system faltered when faced with real-world complexities, like customers picking up items for others or unexpected product placements. This highlights the vast difference between a machine that can play chess at a superhuman level and one that can navigate the messy, unpredictable world like a human.

454608767 912615117567983 7347151526335702200 n

AI Courses | Online Courses for All Levels

AI Courses AI courses delve into the theory and application of algorithms that enable machines to mimic human intelligence and perform tasks autonomously. Build your skills in topics such as generative AI, prompt engineering, deep learning and more.

Here’s why AI hasn’t yet achieved true genius:

AI suffers from narrow focus

Rule-based learning

Modern AI thrives on clearly defined rules and parameters. AlphaGo, for instance, dominated Go because the game operates on a finite board with a set of well-defined moves and winning conditions. The AI was trained on millions of past Go games, allowing it to identify patterns and strategies to achieve victory.

However, this approach crumbles when faced with tasks that lack such rigid structures. Writing a sonnet requires understanding human emotions, cultural references, and poetic metre – concepts outside the realm of rule-based learning. Diagnosing an illness involves interpreting complex medical data, patient history, and subtle physical cues – another area where current AI struggles.

Data dependency

AI’s prowess is heavily reliant on the quality and quantity of data it’s trained on. AlphaGo’s success stemmed from the vast collection of Go games it analysed. But imagine trying to train an AI on all human knowledge – it’s simply not feasible.

Furthermore, data can be biased, leading the AI to perpetuate those biases in its outputs. For instance, an AI trained on medical data from a specific demographic might struggle to diagnose patients from different backgrounds. The Turk, on the other hand, didn’t require vast datasets. Its “intelligence” came from the human player’s ability to adapt and strategise within the established rules of chess.

Transfer learning challenges

Ideally, AI should be able to leverage knowledge gained in one domain and apply it to another. This is called transfer learning. However, current AI often struggles with this concept. An AI trained on chess moves might not be able to translate that knowledge to a completely different game like checkers, even though both involve strategy on a board.

The Turk, however, could apply its understanding of chess strategy to any legal move presented, demonstrating a more fluid and adaptable form of intelligence.

In essence, today’s AI excels at mimicking intelligence within very specific boundaries. It can learn and perform exceptionally well within those constraints but lacks the broader understanding and adaptability needed for true general intelligence. This makes tasks requiring creativity, common sense, or the ability to learn from limited data particularly challenging for current AI systems.

440330817 842846794529301 288663163262839955 n

Why tech is for moms

Dive into the inspiring stories of resilience, growth, and empowerment from four incredible TripleTen graduates. From overcoming career gaps to pivoting to new roles, the transformative power of tech careers in fostering flexibility, balance, and opportunity for working mothers. Explore why tech is the perfect fit for moms seeking meaningful careers with room to thrive.

The common sense chasm: Why AI needs more than just data

One of the biggest hurdles on the path to Artificial General Intelligence (AGI) is the lack of common sense in current AI systems. While AI can process information and identify patterns with impressive accuracy, it often struggles with the intuitive understanding of the physical world and human behaviour that we take for granted.

This “common sense deficit” creates significant challenges when applying AI to real-world scenarios with their inherent messiness and unpredictability. Let’s delve deeper into this critical gap:

Understanding physical reality

Our ability to navigate the physical world relies heavily on common sense. We grasp basic principles like gravity, object permanence, and cause-and-effect without needing them explicitly programmed. AI, however, often lacks this intuitive understanding.

For instance, an AI tasked with cleaning a room might struggle to differentiate between a pile of dirty laundry and a strategically placed beanbag chair. It might not understand the concept of “fragile” and accidentally damage delicate objects. This lack of grounding in the physical world makes real-world applications, like robotics, challenging.

Reasoning with incomplete information

Humans excel at making sound judgements even with limited data. We can fill in gaps based on common sense and past experiences. An AI, however, might struggle with situations where all the information isn’t readily available.

Imagine an AI tasked with composing an email. It can analyse past emails and identify patterns, but understanding the recipient’s emotional state or the underlying context of the message requires a grasp of social cues and unspoken rules – areas where common sense plays a crucial role.

Social intelligence deficit

Human interaction is a complex dance of verbal and nonverbal cues, cultural norms, and unspoken expectations. AI often struggles to understand these nuances. For instance, an AI chatbot designed for customer service might misinterpret sarcasm or miss subtle emotional cues in a customer’s voice, leading to awkward or frustrating interactions.

Bachelor (B.Sc) Applied Artificial Intelligence

Online Study Model Degree Bachelor of Science (B.Sc.) 180 ECTS credits Language English Tuition Fees & Discounts Tuition Fees Starting from 130 € (131 $)* per month *based on the exchange rate from 01.07.2023 09:05 AM (CEST) Discounts Discounts and promotions are available on a seasonal basis Online application No application fees Study Duration Study start any time Duration36 – 72 months (depending on time model)

The frame problem

One of the core challenges in AI is the frame problem – deciding which information is relevant to a particular situation. Humans excel at filtering out irrelevant details and focusing on what matters.

AI, however, might get overwhelmed by the sheer amount of data it can perceive, struggling to separate the background noise from the crucial elements needed to make a decision.

Overcoming the common sense chasm

Researchers are actively exploring ways to bridge this common sense gap. Some approaches involve incorporating knowledge graphs and symbolic reasoning into AI systems. Others focus on developing AI that can learn through interaction and experience in the real world. Ultimately, achieving true general intelligence will likely require a combination of these approaches, allowing AI to develop a more human-like understanding of the physical and social world.

AGI’s long road ahead

While AI is making significant strides, achieving true AGI – machines with human-level thinking across various domains – remains a distant goal. This means the current job market still heavily values human data scientists who can bridge the gap between specialised AI tools and real-world applications.

Data science skills in high demand

Even without AGI, the demand for skilled data scientists is booming. Businesses across industries rely on data analysis to make informed decisions, identify trends, and develop innovative products. A Data Science Bootcamp can equip you with the necessary skills to:

Wrangle and analyse data: You’ll learn how to collect, clean, and manipulate large datasets, a crucial first step in any data-driven project.
Master programming languages: Languages like Python and R are essential for data scientists, and bootcamps provide a solid foundation in these tools.
Statistical analysis and modelling: You’ll gain expertise in statistical techniques and machine learning algorithms used to extract insights from data.
Data visualisation: Effectively communicating insights through clear and compelling visualisations is a key skill for data scientists.

TripleTen Bootcamp’s advantages

Focus on practical skills: TripleTen’s programme emphasises project-based learning, ensuring you develop hands-on experience applying data science techniques to real-world scenarios.
Part-time and full-time options: The flexibility of part-time and full-time programmes allows you to fit data science training into your existing schedule, making it accessible to a wider range of learners.
Career support: TripleTen, offers career services to help you with resume writing, interview preparation, and job placement after graduation.

AGI as a collaborator, not a competitor

Instead of fearing AGI as a replacement, view it as a potential future collaborator. As a data scientist, you’ll be well-positioned to work alongside these advanced AI systems, interpreting their outputs and ensuring they are used effectively.

Investing in yourself

Regardless of the pace of AGI development, data science skills are a valuable asset in today’s job market. Enrolling in a bootcamp is an investment in yourself, allowing you to gain the knowledge and experience to thrive in the ever-growing data-driven world.

By taking advantage of the current gap between AI and AGI, you can position yourself for a successful career in data science. Remember, the skills you learn will be valuable for years to come, regardless of the future trajectory of AI.

The Turk may have been a clever illusion, but it embodied a key aspect of intelligence – the ability to adapt and respond to novel situations. Modern AI is getting closer, but we’re still in the era of well-trained specialists, not generalists. The dream of a machine that can truly think and learn like a human remains on the horizon, a testament to the complexity of human intelligence and the long road ahead for AI.

_______________________________________