> Markdown version of [/videos/1000-the-shadows-of-reasoning-new-design-paradigms-for-a-gen-ai-world](https://www.wearedevelopers.com/videos/1000-the-shadows-of-reasoning-new-design-paradigms-for-a-gen-ai-world). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# The shadows of reasoning – new design paradigms for a gen AI world

Generative AI merely mimics true understanding, leaving enterprise apps vulnerable to logical traps. Learn a new design paradigm that makes large language models fully transparent and auditable.

- **Speakers:** [Jonas Andrulis](https://www.wearedevelopers.com/@jonas-andrulis)
- **Event:** World Congress 2024
- **Published:** August 20, 2024
- **Duration:** 25:14
- **URL:** https://www.wearedevelopers.com/videos/1000-the-shadows-of-reasoning-new-design-paradigms-for-a-gen-ai-world

## Summary

**The Limits of AI Reasoning.** Generative AI and modern language models represent a profound shift in software development, progressing from expertly crafted algorithms to deep learning models capable of complex pattern recognition. Despite impressive capabilities in text and video generation, these systems lack true conceptual understanding of the physical world. They merely observe the "shadows of reasoning," memorizing intricate structural patterns but confidently failing when confronted with logical traps that demand genuine comprehension over statistical mimicry. **Visualizing Knowledge Flow.** To build reliable enterprise applications, developers must move beyond basic chatbot interactions and address inherent limitations like AI hallucinations. A new design paradigm centers on making the inner workings of large language models auditable. By visualizing positive and negative token correlations between prompts and sources, systems can trace the flow of factual knowledge and highlight exactly why a model reached specific conclusions. **Human-in-the-Loop Integration.** Applying deep transparency enables robust solutions, such as cross-referencing financial claims against quarterly earnings transcripts or detecting contradictions in text-generation tasks. Embedding audit trails directly into software architectures allows developers to construct intelligent workflows that seamlessly loop human decision-making into the process. Ultimately, this approach shifts the focus from fine-tuning brittle chat experiences toward orchestrating transparent, accountable systems where implicit model patterns are made explicit and verifiable.

**Keywords:** generative AI design paradigms, LLM pattern recognition, AI hallucination debugging, model knowledge tracking, explainable AI systems, RAG workflow orchestration, human-in-the-loop AI, deep learning limitations, convolutional neural networks, token correlation analysis, enterprise AI transparency, software system auditability, generative AI logic traps

## Chapters

1. **From crafted algorithms to deep learning and pattern recognition** (00:04) — How computer vision evolved from human-designed filters to deep learning models that independently identify complex patterns.
1. **Evaluating generative AI capabilities and physical reasoning limitations** (03:51) — Why large language models plateau in advanced reasoning and fail to independently comprehend rules governing the physical world.
1. **Testing large language models with classic logic puzzle constraints** (07:47) — Analyzing how language models approach ancient riddles by mixing literal pattern matching with incorrect logic rather than true reasoning.
1. **Investigating board state comprehension through language model chess simulations** (12:05) — Simulating chess matches against language models reveals reliance on recognizable move sequences rather than functional board state comprehension.
1. **Extrapolating complex world rules from sequence data in Othello-GPT** (14:26) — How a model trained solely on sequential gameplay generated an accurate internal embedding of functional board rules.
1. **Visualizing token correlations to track artificial neural knowledge flows** (16:47) — Developing analytical tools that surface positive and negative pattern correlations to make knowledge flows inside models inherently transparent.
1. **Building auditable information workflows and human-in-the-loop verification systems** (21:05) — Designing interactive systems that leverage underlying token weights to detect hallucinations and trace the origin of output knowledge.

## Related Moments

- [Balancing AI regulation with technological innovation in human resources](https://www.wearedevelopers.com/videos/1356-from-learning-to-leading-why-hr-needs-a-chatgpt-license) (from "From Learning to Leading: Why HR Needs a ChatGPT License")
- [Defining modern generative artificial intelligence and agent applications](https://www.wearedevelopers.com/videos/1512-building-ai-applications-with-langchain-and-node-js) (from "Building AI Applications with LangChain and Node.js")
- [Mitigating the inherent challenges of generative AI tools](https://www.wearedevelopers.com/videos/844-enter-the-brave-new-world-of-genai-with-vector-search) (from "Enter the Brave New World of GenAI with Vector Search")
- [Exploring breakthroughs in large generalist machine learning models](https://www.wearedevelopers.com/videos/509-a-walkthrough-on-responsible-ai-frameworks-and-case-studies) (from "A walkthrough on Responsible AI Frameworks and Case Studies")
- [Discussion on AI hallucinations and practical developer workflows](https://www.wearedevelopers.com/videos/805-openai-for-fintech-building-a-stock-market-advisor-chatbot) (from "OpenAI for FinTech: Building a Stock Market Advisor Chatbot")
- [Overview of generative AI and the presentation agenda](https://www.wearedevelopers.com/videos/1001-langchain4j-an-introduction-for-impatient-developers) (from "Langchain4J - An Introduction for Impatient Developers")

## Related Articles

- [WWC24 Talk - Scott Hanselman - AI: Superhero or Supervillain?](https://www.wearedevelopers.com/magazine/469-wwc24-talk-scott-hanselman-ai-superhero-or-supervillain)
- [MLOps And AI Driven Development](https://www.wearedevelopers.com/magazine/82-mlops-and-ai-driven-development)
- [Graph and AI Trends 2026: Why Is AI Running but Not Yet Delivering?](https://www.wearedevelopers.com/magazine/680-graph-and-ai-trends-2026-why-is-ai-running-but-not-yet-delivering)
- [How to Use Generative AI to Accelerate Learning to Code](https://www.wearedevelopers.com/magazine/530-how-to-use-generative-ai-to-accelerate-learning-to-code)

## Related Jobs

- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**
- [AI Operations Manager (all genders)](https://www.wearedevelopers.com/jobs/48263-ai-operations-manager-all-genders) at **envelio**
- [Principal Software Engineer, Enterprise AI Platform](https://www.wearedevelopers.com/jobs/ext/1467292-principal-software-engineer-enterprise-ai-platform) at **GitHub**
- [Head of AI Applications](https://www.wearedevelopers.com/jobs/ext/1456210-head-of-ai-applications) at **ZEISS Group**
- [AI & Machine Learning Engineer (all genders)](https://www.wearedevelopers.com/jobs/48217-ai-machine-learning-engineer-all-genders) at **msg**