Come pensa un’intelligenza artificiale? | Giulio Deangeli | TEDxBari

Traditional AI (Discriminative AI)

• Traditional AI involves techniques like deep learning and neural networks.
• Deep learning neural networks perform classification (assigning labels to data) and regression (estimating numerical values).
• Neurons, the basic units of neural networks, process electrical signals and generate output signals.
• The complexity of life and the brain arises from the non-linear nature of neurons and their massive interconnectedness.
• Training neural networks enables them to perform classification and regression tasks, resulting in pattern recognition and prediction capabilities.
• Traditional AI has applications in medicine, such as drug discovery and development.

Generative AI

• Generative AI, like the Transformer architecture, uses attention mechanisms to filter important information and generate new text.
• The scalability of generative AI models like GPT has exceeded expectations, with increased computational power leading to significantly improved performance.
• Unlike traditional AI models, generative AI models do not necessarily perform better with increased scale.
• GPT (Generative Pre-trained Transformer) generates human-like text by predicting the next word in a sequence based on preceding words.
• GPT introduces a probabilistic element, making its outputs more varied and human-like, mimicking the non-deterministic nature of the human brain.
• GPT has applications in medicine, such as analyzing medical literature, creating standardized clinical datasets, and assisting doctors in decision-making.

Societal Impact of AI

• AI is already integrated into healthcare, with patients interacting with AI-powered systems for triage and other tasks.
• AI has a significant societal impact and is an integral part of our lives, presenting both challenges and opportunities.
• AI techniques can significantly increase productivity, and tools like CoPilot enhance programmers' work.
• The widespread adoption of AI has societal implications, potentially leading to rapid job displacement similar to the Industrial Revolution but on a shorter timescale.
• While some jobs will become obsolete, AI can democratize access to certain professions by reducing the need for extensive memorization and technical knowledge.
• The focus of human work will shift towards decision-making and evaluation, while routine tasks can be automated.
• There is a risk of social inequality if access to AI tools is limited, so governments must ensure universal access.
• AI is an integral part of the world and can be used by others to gain a competitive advantage.