In the discussion of artificial intelligence, AI and generative AI are often understood as synonyms. In discussions of artificial intelligence, one might refer to a generative AI chatbot such as ChatGPT, which represents just a small portion of the broader application of artificial intelligence technologies. Every mobile phone camera photo or route planner route is based on artificial intelligence, even if the user is not necessarily aware of it. The influence of this operational AI, which is non-generative, is currently more significant, but it’s likely that different types of AI will complement each other in the future.
The ambiguity of the terminology is a challenge to the overwhelming discussion. If two people are chatting about AI, they might mean completely different things. Artificial intelligence is a decades-old, rapidly developing and established technology area with many applications. Generative AI is a technology that has been widely adopted in recent years, which is developing at a tremendous pace and is very effective in certain applications, but can be unexpectedly weak in areas that are traditionally strong for computer systems. These two technology areas will probably merge with each other in many ways, as their strengths are almost opposite to each other.
For the sake of more professional discussion, I try to clarify the key application areas in layman’s terms. First, generative AI and operational AI must be separated from each other. Below, we’ll delve into the various subtypes that fall under these categories.
Generative AI
Creative Generation:
- Purpose: Content creation, such as text, image and music generation.
- Example: GPT-4, DALL-E.
- Description: Generates new data by modelling existing data. Creative generation produces data that resembles the output that a human could have made. Therefore, the data can also contain errors, illogicalities, and even outright lies.
Generative Adversarial Networks (GAN):
- Purpose: Realistic data generation, such as images and sounds.
- Example: Deepfake technology, image recognition improvements.
- Description: Two neural networks compete with each other; the generator creates data, and the discriminator evaluates its authenticity, which improves the generator’s performance. With this approach, it is also possible to improve the outcome of creative generation in the future. On the other hand, the opposing actions of the algorithms can also mislead a normally reliable artificial intelligence system to act contrary to its purpose.
Simulative Generation:
- Purpose: Creating simulations and forecasts.
- Example: Physics simulations, weather forecasts.
- Description: Creates simulations of possible future events or phenomena. Unlike predictive artificial intelligence, the purpose of simulation is to predict something that does not yet exist, such as an unbuilt factory or a weather phenomenon that deviates from history.
Interactive Generation:
- Purpose: Creating dynamic and reactive systems.
- Example: Chatbots, game character artificial intelligence.
- Description: Adapts to the user’s actions in real time. For example, customer service chatbots that answer customer questions, and game character artificial intelligence that reacts to the player’s actions and makes the game more realistic challenges. In the near future, entire games with their worlds can also be built by means of interactive generation by talking to a game builder artificial intelligence.
Operational AI
Analytical AI:
- Purpose: Data analysis and prediction.
- Example: Data analytics, machine learning algorithms.
- Description: Uses existing data to create models and forecasts. Analytical artificial intelligence differs from normal computer-assisted analysis in that not all algorithms used for analysis are predefined, but the system learns analysis methods based on the data used.
Optimising AI:
- Purpose: Optimising processes and systems.
- Example: Logistics optimisation, resource management.
- Description: Aims to improve efficiency and reduce costs. Optimising artificial intelligence also learns from its environment. For example, smartphone navigators change the route plan based on real-time traffic data.
Predictive AI:
- Purpose: Predicting future events based on data.
- Example: Weather forecasts, economic forecasts.
- Description: Predicts future events or trends by analysing historical data. Unlike simulative artificial intelligence, the purpose of anticipation is to predict something that has happened at some point and is likely to recur at some time interval.
Prescriptive AI:
- Purpose: Providing action recommendations to support decision making.
- Example: Recommendation systems in online stores, medical recommendations.
- Description: Analyses data and provides recommendations that help make optimal decisions. Typically, prescriptive artificial intelligence is given some goal such as online store conversion, and it optimises its actions such as product advertising to maximise the outcome of this goal.
Autonomous AI:
- Purpose: Self-operating systems.
- Example: Self-driving cars, robotics, factory process automation.
- Description: Makes decisions and performs tasks without human intervention. Autonomous artificial intelligence is the most challenging form of operational artificial intelligence, as it must act appropriately with a very high probability. Physical world automation has been done for a long time, for example in lifts, but when done with artificial intelligence assistance, the challenge is the extreme reliability requirement.
Clear terminology is essential for professional AI discussion. By separating generative and operational AI from each other, we can improve understanding and promote the application of technology effectively. Both areas have their own strengths and weaknesses, which are important to recognise and understand.
Leverage data and artificial intelligence as a practical resource
