A.I. flavours

Since the success of ChatGPT, we might almost think that Generative A.I. is the same as A.I. in general. Nothing could be further from the truth. There is an important difference between generative AI and Narrow AI (also known as Weak AI). Time for a brief clarification, we think at The House of Coaching.

Narrow AI (Weak AI)

Narrow AI refers to AI systems that are designed and trained to perform a specific task. These systems are "narrow" or "weak" because their intelligence and learning capacity are limited to the specific domain for which they are programmed. They cannot operate outside that framework or learn new tasks without additional programming or training.

Examples: Chatbots designed for customer service, facial recognition systems, and recommendation systems like those used by Netflix or Amazon. There are numerous other applications that we use almost daily without realising that "Narrow A.I." is behind them. Just think of:
Speech Recognition in Smartphones: Systems like Siri on iPhones, Google Assistant on Android devices, and Bixby on Samsung smartphones. These AIs can understand speech and respond based on pre-programmed commands.
Recommendation Systems: Used by online platforms like Netflix, Spotify, and Amazon to recommend products, films, or music based on your previous preferences and behaviour.
Email Spam Filters: AI in email services like Gmail or Outlook that automatically filter spam and unwanted emails based on certain characteristics.
Customer Service Chatbots: Many companies use chatbots on their websites or in customer service apps to answer customer questions and provide support.
Facial Recognition: Applied in smartphones for unlocking or by social media platforms like Facebook for tagging photos.
Narrow AI in Navigation Apps: Such as Google Maps and Waze, which analyse traffic patterns and suggest the fastest route to your destination.
Banking and Fraud Detection: AI systems used by banks to recognise unusual activities on your account and detect potentially fraudulent transactions.
Smart Assistants in Household Appliances: For example in smart thermostats like Nest, which learn from your preferences and automatically adjust the temperature in the house.
Automatic Translation Services: Such as Google Translate, which translates texts between different languages.
Automatic Content Moderation: Used by social media platforms like Instagram and YouTube to scan and filter content for inappropriate or harmful material.
Medical Imaging: AI that compares ideal X-ray images with those of patients and thus supports doctors in their decision-making.
Characteristics of Narrow AI are:
Specialisation: The AI is trained for a very specific purpose, for example analysing medical images.
Based on Training Data: The ability of the AI to be deployed depends on the quality and diversity of the training data on which the system is trained.
No General Intelligence: These systems cannot function outside their training. For example, they cannot be used for other types of tasks without additional training.
Decision Support: In medical imaging, for example, Narrow AI systems are often used as supporting tools to help doctors make decisions, but they generally do not replace the clinical assessment of a specialised doctor.

Generative AI

Generative AI, on the other hand, refers to AI systems that can generate new content that is comparable to human-generated content. This can include text, images, music, speech or even code. Generative AI models learn from a large dataset of existing content and use this knowledge to create new, original content.

Examples: GPT-3 or GPT-4 (for text generation), DALL-E (for image creation), and AI composers that can generate new pieces of music.
Characteristics:
Can be creative and generate new content that does not explicitly appear in the training data.
Flexible in application across different domains.
Has the ability to learn and adapt to new information or guidelines.

Just as is common with Narrow AI, we will also have to remember with generative AI that it is supportive and that human intelligence remains necessary for the evaluation and use of the final result.

In essence, the biggest difference between the two is the scope and flexibility. Narrow AI is highly specialised and limited to specific tasks, while generative AI has a broader range of possibilities and is capable of producing new, original output that is comparable to human creativity.

By Hans Demeyer