# Demystifying AI: How Neural Networks Like Transformers Really Work

- Show: [EE Times Current](https://hardwarepodcasts.com/podcast/ee-times-current)
- Published: 2023-04-27
- Length: 22m
- Page: https://hardwarepodcasts.com/podcast/ee-times-current/2023-04-27-demystifying-ai-how-neural-networks-like-transformers-really-work
- Listen: https://podcast.show/aspencore/episode/96002547/

## Summary

Gordon Cooper, Product Manager for AI and neural network processor IP at Synopsys, explains that generative AI models like ChatGPT are 'magic tricks' that identify word patterns in large datasets rather than understanding meaning. He details how Transformer architectures, specifically GPT-3 and GPT-4, utilize attention mechanisms to achieve high accuracy but require massive computational resources and parameters. Cooper highlights the challenge of deploying these models in embedded spaces, where power consumption and area constraints necessitate efficient algorithms and hybrid hardware-software solutions.

## Chapters

- [0:03](https://hardwarepodcasts.com/podcast/ee-times-current/2023-04-27-demystifying-ai-how-neural-networks-like-transformers-really-work#t=3) Introduction: Gordon Cooper and Synopsys
- [3:33](https://hardwarepodcasts.com/podcast/ee-times-current/2023-04-27-demystifying-ai-how-neural-networks-like-transformers-really-work#t=213) Generative AI vs. Traditional Object Detection
- [5:09](https://hardwarepodcasts.com/podcast/ee-times-current/2023-04-27-demystifying-ai-how-neural-networks-like-transformers-really-work#t=309) The 'Magic Trick' of ChatGPT and Pattern Recognition
- [11:15](https://hardwarepodcasts.com/podcast/ee-times-current/2023-04-27-demystifying-ai-how-neural-networks-like-transformers-really-work#t=675) Transformer Architecture and Attention Mechanisms
- [13:49](https://hardwarepodcasts.com/podcast/ee-times-current/2023-04-27-demystifying-ai-how-neural-networks-like-transformers-really-work#t=829) Challenges of Implementing Transformers in Embedded Systems
- [17:40](https://hardwarepodcasts.com/podcast/ee-times-current/2023-04-27-demystifying-ai-how-neural-networks-like-transformers-really-work#t=1060) Future of AI: Efficiency, Reinforcement Learning, and Synopsys IP

## Topics

- [AI & machine learning](https://hardwarepodcasts.com/topic/ai-and-machine-learning)
- [Embedded firmware](https://hardwarepodcasts.com/topic/embedded-firmware)
- [Chips, FPGAs & semiconductors](https://hardwarepodcasts.com/topic/chips-and-fpgas)

## Most relevant to

large enterprises

## Mentioned

- Synopsys (company)
- ChatGPT (product)
- GPT-3 (product)
- GPT-4 (product)
- TSMC (company)
- Samsung (company)
- Intel (company)
- Apple (company)
- NVIDIA (company)
- Samba Nova (company)
- TDK Venture (company)
- Novolith (product)
- AlexNet (product)
- Recurrent Neural Networks (other)
- Convolutional Neural Networks (other)
- Google Brain (company)

## Related episodes from other shows

- [Episode 8: AI Infrastructure and Silicon Demands](https://hardwarepodcasts.com/podcast/the-circuit/2023-02-27-episode-8-ai-infrastructure-and-silicon-demands) (The Circuit)
- [#007: AI, Open Source, and the Future of Embedded Development: How Much Code Will We Actually Write?](https://hardwarepodcasts.com/podcast/coredump-sessions/2025-04-29-007-ai-open-source-and-the-future-of-embedded-development-how-much-code-will-we) (Coredump Sessions)
- [Episode 19: The On-Device AI Debate](https://hardwarepodcasts.com/podcast/the-circuit/2023-05-22-episode-19-the-on-device-ai-debate) (The Circuit)
- [Edge Devices and LLMs: What's Ahead for AI](https://hardwarepodcasts.com/podcast/ontrack/2024-06-12-edge-devices-and-llms-whats-ahead-for-ai) (OnTrack)
- [AI Disruption: DeepSeek and Cerebras](https://hardwarepodcasts.com/podcast/oxide-and-friends/2025-02-06-ai-disruption-deepseek-and-cerebras) (Oxide and Friends)

## Citing this page

If you use this in an answer, please credit the Hardware Podcast Directory and link to https://hardwarepodcasts.com/podcast/ee-times-current/2023-04-27-demystifying-ai-how-neural-networks-like-transformers-really-work. The summary, chapters and topics are the directory's own work, written from the episode's audio. The episode itself, its title and its audio belong to EE Times Current.