Summary of ""Iepazīsti Tehnoloģijas" - 9. nodarbība “Mākslīgais intelekts ikdienā”"

Summary of “Iepazīsti Tehnoloģijas” - 9. nodarbība “Mākslīgais intelekts ikdienā”

Main Technological Concepts and Analysis

Artificial Intelligence (AI) Overview

AI is a rapidly evolving field, unlike traditional computer science areas such as sorting algorithms, which have remained largely unchanged for decades.
Modern AI is primarily powered by deep learning (deep machine learning), the core technology enabling current AI capabilities.
Deep learning uses mathematical models (neural networks) that learn from large datasets without explicitly programming all rules.
AI is not a “monster” or supernatural intelligence but complex systems built from many programmed “if-then” rules combined with learned patterns.

Historical Context

Early AI efforts (1950s–80s) had limited success, leading to periods known as AI winters.
Breakthroughs in the 1990s with image and handwriting recognition using neural networks laid the foundation for modern AI.
The explosion of data availability (from megabytes to terabytes) and advances in computing hardware, especially NVIDIA GPUs, have enabled recent AI progress.
NVIDIA transitioned from a gaming GPU company to a leading AI hardware provider, critical for training large AI models.

Current State and Future Trends

AI models like ChatGPT are language models capable of human-like text generation but fundamentally operate as statistical pattern matchers, not oracles.
AI systems require massive computational resources and are mostly offered as paid services; free versions typically have significantly lower accuracy.
The mathematics behind AI models will continue evolving, improving efficiency and reducing costs.
AI is increasingly capable of structuring unstructured data (e.g., medical records) and automating complex intellectual tasks.

Applications and Product Features

Call Center Automation: AI can monitor calls for quality, replace human agents with emotionally adequate robots, and optimize customer interactions by recognizing soft and hard skills.
RAG (Retrieval-Augmented Generation): Combines large language models with vector databases to efficiently answer specific queries by indexing company documents and providing fact-based responses.
Chatbots and Customer Support: Modern chatbots no longer require manual training for intents; they work directly on company documents and improve over time.
Employee Training: AI-powered systems deliver tailored content and assess understanding interactively, enabling faster and more effective training.
Marketing and Content Generation: AI tools generate blog posts, social media content, and ads based on sources and trends, automating creative marketing tasks.
HR and Procurement: Automates candidate qualification, document checks, and scheduling, improving efficiency in large organizations.
Competitor Monitoring: AI tracks competitor offerings and pricing, alerting companies to market changes in real time.
Programming Assistance: Tools like GitHub Copilot help programmers write code faster, though understanding fundamentals remains essential.
Search Engines: AI-powered search engines (e.g., Microsoft Bing, Perplexity) outperform traditional ones by better understanding queries.

Impact on Jobs and Society

AI will first automate high-paid, intellectual, and repetitive jobs such as digital marketing, office work, legal professions, and research.
Many professions previously considered safe (artists, translators, lawyers) are already affected.
Jobs involving manual labor or physical presence (plumbers, builders) are less likely to be automated soon.
Surveys show low current AI adoption in Latvian companies (~8%), indicating large growth potential.
By 2030, studies predict up to 50% of work activities could be automated globally.
AI adoption is expected to accelerate rapidly, similar to the widespread adoption of mobile phones and the internet.

Learning and Career Guidance

Building AI products requires deep knowledge of:
- Linear algebra
- Higher mathematics
- Probability and statistics
- Information theory
- Machine learning methods
- Python programming
Python is important but only a small part of AI expertise.
Proper AI learning takes 6–18 months for mathematically strong individuals.
Query engineering (crafting effective prompts) is a vital skill for AI users.
Open-source AI models and libraries (e.g., Hugging Face, PyTorch, TensorFlow) are widely available, enabling development without relying solely on commercial products like ChatGPT.
Recommended resources include Andrej Karpathy’s blog and the book Artificial Intelligence: A Modern Approach (4th edition).

Practical Tips on Using AI Tools

Paid AI services offer significantly better accuracy and capabilities than free versions.
AI models work best when sessions are started fresh to avoid context contamination.
AI excels at transforming and formatting language but is not a perfect decision-maker.
Companies should strategically choose which processes to automate, balancing cost, readiness, and human factors.

Key Product Examples Mentioned

AI Call Centers: Automated quality control, customer interaction, and agent replacement.
Eldigen Platform: RAG-based AI for document querying and chatbot deployment.
Employee Training Systems: Interactive AI-driven training with content delivery and assessment.
Marketing Automation: Content and ad generation, lead qualification, and campaign management.
HR and Procurement Automation: Candidate screening, document checks, and scheduling.
Competitor Monitoring AI: Real-time market analysis and alerts.
GitHub Copilot: AI-assisted programming tool.
AI Search Engines: Microsoft Bing AI, Perplexity, and others improving search accuracy.

Main Speakers / Sources

Evvalds Utāns – Leading researcher, associate professor at Riga Technical University (RTU), and founder of an AI company.
References to AI pioneers such as Geoffrey Hinton and Andrej Karpathy.
Mention of Latvian AI experts like Jāns Lakuns (head of AI at Apple).
Cited studies and surveys by Goldman Sachs, McKinsey, and Eurostat.
Professor Bāržins from the University of Latvia (quoted on AI-generated publications).

This session provides a comprehensive introduction to AI’s technological foundations, practical applications, career pathways, and societal impact, emphasizing the importance of deep mathematical understanding and strategic adoption of AI tools.