AI's Common Sense Gap: Bridging Value and Potential

In recent years, artificial intelligence has made significant progress in mimicking human intelligence, particularly in natural language processing. For example, language models like ChatGPT and Google Gemini are now able to hold logical conversations on a wide range of topics, even when the topic is changed suddenly. This ability to understand and respond to context is a sign that AI is developing a elementary form of common sense. However, it is important to note that AI is still in its early stages of development, and it is not yet capable of replicating the full range of human intelligence.

Considering its ongoing evolution, the question lies in how businesses can begin incorporating AI and extracting value from its capabilities.

Imagine AI as a super-powered calculator, amazing at crunching numbers and finding patterns. But like a calculator, it can't understand the real world and or core context. So, businesses can use it to analyse data, automate tasks, and personalise experiences (like recommending the perfect gift). However, AI can't handle "common sense" things like business domain, humour or sarcasm. To avoid mistakes, businesses need to focus on specific tasks and pair AI with human guidance, just like using a calculator with your own brainpower. This way, AI can be a powerful tool for businesses, even though it's not yet a "master of everything.

?Let’s deep dive further into this topic with me. Firstly, we will understand where AI excels in applying common sense and its limitations. Next, we will explore ongoing efforts by AI researchers to enhance its quality. Lastly, we will outline active use cases and immediate steps for businesses to effectively learn and implement AI to their operations and environments.

Some notable strengths of AI today.

• Basic reasoning: AI can handle basic deduction and inference tasks.?For example, if it learns that it rained earlier and the ground is wet, it can infer that the rain caused the wetness and or can observe patterns in consumer behaviour and correlate increased online shopping with economic trends, suggesting potential shifts in market demand to financial analysts.

• Physical world reasoning: Consider an AI-equipped robot in a warehouse: it can navigate the environment, avoid collisions with obstacles, pick up items, and place them in designated locations. This showcases AI's ability to comprehend spatial relationships and execute tasks based on physical constraints.
• ?Social norms and etiquette: AI trained on social media interactions can detect conversational nuances, perceiving between a formal business inquiry and a casual chat among friends. This enables it to adjust its responses appropriately, ensuring a professional tone in formal settings while being more relaxed in casual conversations.
• Context awareness: AI can analyse context within a conversation or situation to identify relevant information and interpret events.?This allows it to understand sarcasm, humour, and double meanings to a certain extent.
• Language Translation: AI's mastery in language translation is remarkable. Platforms like Google Translate utilise AI to translate text across various languages accurately, facilitating global communication.
• Personalisation and Recommendations: AI algorithms power personalised recommendations on platforms like Netflix and Amazon. They analyse user behaviour to suggest movies, products, or content tailored to individual preferences.
• Pattern Recognition: AI demonstrates exceptional proficiency in recognising financial patterns. For instance, in banking, AI algorithms analyse large volumes of transaction data to detect fraudulent activities. They can identify irregular spending patterns or unusual transactions, enhancing security measures and preventing fraudulent actions in real-time.

Let's explore few examples showcasing AI's impact on our lives and businesses.

1. Personal Assistants: Imagine telling your smart speaker, "It's freezing in here, and I'm leaving for work soon." AI, understanding the context of your schedule and temperature preference, might automatically adjust the thermostat to warm up before you leave. This demonstrates not just basic inference but also awareness of time and spatial relationships.

2. Customer Service Chatbots: Chatbots trained on extensive customer data can effectively respond to basic queries. If you ask about return policies, they can not only find the relevant information but also understand your implicit question about the return window based on the specific product you mentioned. This shows AI's ability to grasp contextual clues and adjust responses accordingly.

3. Self-driving Cars: While still under development, autonomous vehicles utilize AI to navigate complex road environments. They perceive traffic signals, identify pedestrians and obstacles, and even adjust speed based on weather conditions. This intricate dance of understanding physical laws, reacting to real-time events, and anticipating future scenarios showcases AI's growing grasp of common sense in action.

4. Medical Diagnosis Systems: AI trained on vast medical datasets can analyse X-rays and scans to detect disease patterns. Beyond pattern recognition, they can consider factors like a patient's age and medical history to suggest potential diagnoses or even flag anomalies for further investigation. This demonstrates AI's ability to handle probabilistic reasoning and combine information with contextual awareness.

5. Fraud Detection: Financial institutions leverage AI to analyse real-time transactions and identify suspicious activity. They do not just flag unusual spending amounts, but can also consider factors like a user's typical location and purchasing habits to distinguish genuine transactions from potential fraud. This shows AI's ability to understand behavioural patterns and detect anomalies within a broader context.

Additionally; some practical applications of AI across the front, middle, and back office perspectives.

Front Office/Customer facing

• Personalised Marketing: AI enables personalised marketing campaigns by analysing customer behaviour and preferences. It can automate tailored content creation, recommend products based on purchase history, and optimise marketing strategies for individual customers.
• ?Sales Forecasting and Lead Scoring: AI algorithms can analyse sales data, customer interactions, and market trends to predict sales forecasts accurately. Additionally, AI can score leads based on their likelihood to convert, helping sales teams prioritise their efforts.
• ?Dynamic Pricing: In retail and e-commerce, AI can automate dynamic pricing strategies based on factors like demand, competitor pricing, and customer behaviour. This ensures optimal pricing for products or services in real-time.
• ?Appointment Scheduling and Booking: AI-powered systems can automate appointment scheduling and booking processes. They streamline the process for customers, offering available time slots and managing reservations efficiently.

Middle Office

• Risk Management and Analysis: AI sifts through vast datasets to spot potential risks for businesses, using pattern recognition to alert managers, aiding informed decisions despite lacking deep understanding of market fluctuations.
• ?Investment Portfolio Optimisation: Analysing market conditions, AI recommends optimal portfolio allocations, leveraging statistical correlations for maximising returns and managing risk, even without comprehending complex economic theories.
• ?Regulatory Compliance and Reporting: Automating compliance tasks, AI ensures adherence to regulations, generating reports efficiently. While not grasping every intricacy, it understands basic extraction and analysis, saving time and resources.
• Performance Analysis and Forecasting: AI analyses business data, identifying trends and predicting future performance. Though not grasping human factors, it recognises patterns, aiding data-driven decisions for future growth.

Back-Office

• Data Entry and Automation: AI replaces manual data entry by recognizing patterns, automating tasks like invoice processing and customer updates. This reduces errors and manual work significantly.
• Fraud Detection and Prevention: AI spots suspicious patterns in financial transactions, aiding in real-time fraud detection. Despite not understanding human motivations behind fraud, it prevents financial losses and safeguards sensitive data proactively.
• ?Predictive Maintenance: AI analyses equipment sensor data, predicting potential failures. Despite lacking "common sense," it alerts maintenance teams for proactive repairs, reducing downtime and boosting operational efficiency.

Some of AI Limitations today.

• Abstract Concepts and Intentions: AI often grapples with understanding complex concepts like love, trust, or intention. Imagine sharing your dream of becoming a writer with your AI assistant. It might suggest practical steps like taking writing courses, but miss the deeper underlying passion and drive motivating your goal.
• Unforeseen scenarios: AI works best within its training data and struggles with new situations or unexpected events.?It may not understand nuanced context or subtle cues that could dramatically alter the interpretation of a situation.
• Cause-and-effect complexity: While AI can handle simple cause-and-effect relationships, complex, multi-factorial situations present a significant challenge. It may incorrectly attribute blame or misunderstand the web of interactions leading to a specific outcome
• Moral and ethical dilemmas: AI lacks the innate moral compass and ethical framework humans possess.?It may not understand the nuances of ethical implications or make decisions that align with human values, potentially leading to biased or harmful actions.
• Misinterpreting Humour and Sarcasm: Imagine sharing a funny meme with your AI assistant, expecting a laugh. Unfortunately, AI might lack the understanding of humour and sarcasm to grasp the joke's context. It could interpret the content literally, potentially missing the punchline or even misconstruing it as offensive.
• Struggling with Emotional Intelligence: AI may struggle to understand and respond to human emotions. If you entrust in your AI about a stressful day, it might offer generic advice or solutions not tailored to your emotional state. It could miss subtle cues of sadness, frustration, or anger, hindering its ability to provide genuinely helpful support.
• ?Ethical Dilemmas and Moral Biases: AI algorithms can inherit biases present in their training data, leading to unfair or discriminatory outcomes. For example, an AI-powered hiring system trained on biased data might Favor certain demographics over others, perpetuating unfair hiring practices.
• ?Explainability and trust: Transparency in AI decision-making fosters trust. Creating understandable AI systems is vital for public confidence and successful societal integration.
• ?Computational resources: Handling extensive knowledge and continual learning demands substantial computational power. Optimising algorithms and efficient hardware infrastructures are critical for scaling AI with common sense.

Researchers are actively exploring various avenues to improve AI

1. Explainable AI (XAI): Making AI models more transparent and understandable is crucial for identifying biases, understanding decision-making processes, and ultimately building trust in AI systems. Research in XAI focuses on providing insights into how AI models arrive at their conclusions, allowing humans to evaluate their reasoning and address potential issues.
2. Commonsense Knowledge Bases: Building vast databases of commonsense knowledge about the world, including social norms, cultural nuances, physical laws, and even basic human psychology, is another critical step. These knowledge bases would serve as a reference point for AI systems, providing them with context and background information to inform their reasoning and decision-making.
3. Continual Learning: Traditional AI models are often "trained and deployed," lacking the ability to adapt and learn from new experiences. Research in continual learning aims to develop AI systems that can continuously learn and update their knowledge base as they interact with the world, allowing them to refine their understanding of common sense and adapt to ever-changing situations.
4. Multimodal Learning: Integrating information from multiple senses, such as sight, sound, and touch, can significantly enhance AI's understanding of the world. By combining visual data with auditory cues and spatial awareness, AI can better grasp context and infer meaning, leading to more robust and flexible common sense capabilities.
5. Embodied AI: Allowing AI to interact with the physical world through robots or other physical interfaces can provide another valuable source of learning. By experiencing the world directly and making mistakes, AI can learn from its interactions and develop a deeper understanding of physical laws, cause-and-effect relationships, and the nuances of human-environment interaction.

Some important steps businesses can take to design strategic approach to leverage AI

1. Holistic Integration: Build an AI Strategy

• Think like a futurist. How can AI revolutionise your industry? Design a pilot project to explore that future
• Move from isolated use cases to an integrated AI strategy across the organisation.
• Align AI initiatives with overall business objectives to ensure maximum impact.
• Conduct internal audits to evaluate potential AI applications within specific processes.
• Prioritise areas where AI can offer significant efficiency, cost reduction, or innovative advantage.
• Partner with AI experts:?"Don't do it alone. Collaborate with AI specialists to unlock capabilities hidden within your data and operations.
• Set realistic and measurable goals for AI implementation.

?2.Foster a Culture of Data-Driven Decision Making

• Invest in data infrastructure and quality management practices.
• Train employees on data analysis and interpretation skills.
• Encourage a culture of data exploration and experimentation.

?3.Upskill and Reskill the Workforce

• Invest in training programs to equip employees with the skills needed to work alongside AI.
• Focus on skills like critical thinking, communication, and problem-solving, complementary to AI capabilities.
• Foster collaboration and create cross-functional teams with AI expertise.

?4.Experiment and Pilot AI Solutions

• Start small with well-defined pilot projects to test the feasibility & impact of AI.
• Partner with AI solution providers or consultants for guidance and expertise.
• Continuously iterate and refine based on feedback and results.

?5. Build an Ethical Framework for AI Development and Use

• Establish clear guidelines and policies on data privacy, bias mitigation, and responsible AI use.
• Ensure transparency in AI decision-making processes.
• Foster open communication and engage stakeholders in AI discussions.

?6. Research and Learn Continuously

• Stay updated with AI advancements and how they impact various industries. Invest in ongoing education and training for employees to understand AI's capabilities and limitations.

Additional Opportunities and Considerations.

• Explore cloud-based AI solutions for cost-effective access and scalability.
• Consider citizen development platforms to empower non-technical employees to build simple AI applications.
• Collaborate with universities and research institutions to tap into cutting-edge AI advancements.

Closing note:

AI progresses in context understanding but lacks complete common sense. There is a lot of work happening and this area is evolving rapidly, would be really interesting to watch and learn more in to it. Businesses can capitalise on its strengths in data analysis and automation and many are already. To navigate limitations, pair AI with human expertise for effective integration and decision-making.

Views expressed are personal.

Would be great to hear your views.