The theme of this podcast episode originates from an interesting project by Google called Dobby. I spent an afternoon discussing with our software engineers, and we envisioned how to use on-device AI technology, especially lightweight models, to enhance the ability to prevent phone scams.

We talked about the advantages of on-device AI technology, including real-time processing, resource constraints, and privacy protection.

We explored the foundations of this technology, including model compression, dedicated hardware, and efficient architecture design.

We proposed a three-layer architecture for an anti-fraud system, including quick filtering, proactive analysis, and an interactive engine, and explained in detail how each layer operates.

Of course, while mentioning specific solutions, we are more interested in identifying the technical challenges and future research directions the system faces, such as latency issues, memory capacity, and adaptability.

I am sharing this idea with you, whether you are a startup founder or an investor, I have highlighted the directions you should focus on.

References
1. Kumar, A., & Soni, S. (2021). "Edge AI: Empowering Intelligent Applications at the Device
Level." IEEE Internet of Things Journal.
2. Yu, C., & Deng, L. (2020). "Deep Learning Applications in Speech and Audio Processing."
IEEE Signal Processing Magazine.
3. Google teases taking Pixel Call Screen 'even further' with AI
4. How to use Android 12’s call screening features - The Verge
5. Pixel Phone app preps contextual ‘AI Replies’ in Call Screen

Content by Wei Sun
Audio by Google
Music by Dopestuff @MelodyLoops
Download free music at https://www.melodyloops.com

Disclaimer:
This podcast is for informational purposes only. The technical content and market analysis shared here, including guest opinions, do not constitute professional or financial advice. We cannot guarantee accuracy or completeness of information. Listeners assume all risks from using our content. Consult professionals before making investment or technical decisions.

Contact: victor@nexture.nz

Episode Overview

Inspired by enlightening conversations with startup founders in founder catalyst 2025, this episode delves into the crucial distinctions between entrepreneurs and businesspeople, exploring how these different roles shape economic development.

Key Points Discussed

1. Core Definitions & Distinctions

• Schumpeter's definition of entrepreneurs as agents of "creative destruction"

2. Approaching Uncertainty

• Businesspeople's preference for certainty (referencing Buffett's "I don't invest in what I don't understand")

3. Paths to Success

• Survival rates comparison: traditional businesses (45%) vs. tech startups (20%)
• Real-world examples from our guests' entrepreneurial journeys

4. Required Skills & Background

• Business school's effectiveness in developing businesspeople vs. entrepreneurs
• Success metrics and failure rates (90% startup failure rate)

5. Economic Impact

• How entrepreneurs create new markets
• How businesspeople optimize existing markets

Target Audience

• Aspiring and current entrepreneurs
• Business enthusiasts
• Individuals seeking career direction in business

Call to Action

1. Reflect on which path aligns with your strengths and aspirations
2. Share your thoughts and experiences in the comments section

Featured References

• Schumpeter's "Theory of Economic Development"
• Peter Thiel's "Zero to One"
• Warren Buffett's investment philosophy

Note: All statistics and data mentioned in this episode are based on current research and market analysis as of recording date.


Thanks to:
Mr. Leo - music entrepreneur & business innovation director

Lady Skye - CEO&Co-founder at settlopedia

Content by Wei Sun
Audio by Google
Music by Dopestuff @MelodyLoops
Download free music at https://www.melodyloops.com

Disclaimer:
This podcast is for informational purposes only. The technical content and market analysis shared here, including guest opinions, do not constitute professional or financial advice. We cannot guarantee accuracy or completeness of information. Listeners assume all risks from using our content. Consult professionals before making investment or technical decisions.

Contact: victor@nexture.nz

In this Episode 4, we explore the evolution of software services from tools to colleagues. I believe software services will become "Software as a Work" (SaaW) in the future. Starting with Salesforce's concept of "The End of Software" in 1999, we review the development journey from SaaS to SaaW, as well as the changes in technology, management, and organizational structure brought by SaaW. I believe that SaaW is not just a technological innovation but a revolution in human work methods that will reshape human-machine relationships and create a new world of human-machine collaboration. This content is based on my conversation with Mr. Yang, after which I wrote a long article: "The Evolution of Software Services: From Tools to Colleagues."
If you're interested, you can read it in depth at the following link: https://www.linkedin.com/pulse/evolution-software-services-from-tools-colleagues-wei-sun-czs6e/?trackingId=pj6%2BZlzoRmiPrOolpnGjSA%3D%3D

Below are the references:
1. Gartner (2023). Market Guide for Generative AI in Customer Service and Support
2. Forrester (2023). The Future of Software Services
3. IDC (2023). Worldwide Digital Transformation Spending Guide
4. OpenAI. (2023). GPT-4 Technical Report
5. Microsoft. (2023). Microsoft 365 Copilot Early Access Program Report
6. Anthropic. (2023). Claude Implementation Case Studies
7. AutoGPT Documentation and Implementation Guide
8. LangChain Technical Documentation

Content by Wei Sun
Audio by Google
Music by Dopestuff @MelodyLoops
Download free music at https://www.melodyloops.com

Disclaimer:
This podcast is for informational purposes only. The technical content and market analysis shared here, including guest opinions, do not constitute professional or financial advice. We cannot guarantee accuracy or completeness of information. Listeners assume all risks from using our content. Consult professionals before making investment or technical decisions.

Contact: victor@nexture.nz

In this episode, we want to talk about a very interesting robot-related technology. It's a fascinating experiment. Driven purely by visual data, the robot relies on LLM, VLM, and RAG to achieve autonomous understanding and action. Without GPS, NFC, or radar, it can execute voice commands after visual learning and training.

It is an advanced robotics technology called "remember" that NVIDIA is developing, which aims to give robots the ability to remember and understand the meaning of their experiences, similar to human memory. This technology involves using AI to create a detailed log of the robot's experience, so that it can understand the importance of its observations, not just visual data. The query system allows the robot to search for relevant information in its memory according to the user's query, such as locating items or suggesting places with beautiful vision.

The potential applications of Remember are very wide, including improving elderly care, search and rescue operations, and enhancing daily interactions by allowing robots to predict human needs.

Although the reliability, accuracy, and even moral problems of the technology are still very obvious, it does not prevent its interesting and shiny side.

The following is the reference link of this project:https://developer.nvidia.com/blog/using-generative-ai-to-enable-robots-to-reason-and-act-with-remembr/

Content by Wei Sun
Audio by Google
Music by Dopestuff @MelodyLoops
Download free music at https://www.melodyloops.com

Disclaimer:
This podcast is for informational purposes only. The technical content and market analysis shared here, including guest opinions, do not constitute professional or financial advice. We cannot guarantee accuracy or completeness of information. Listeners assume all risks from using our content. Consult professionals before making investment or technical decisions.

Contact: victor@nexture.nz

We explores the applications, opportunities, challenges, and future prospects of AI in medical research. We first introduces the application of artificial intelligence in medical research, including intelligent literature analysis, data processing and analysis, assisted paper writing, and research trend analysis. Then, We illustrates how AI can optimize medical research workflows, including improving time efficiency, expanding access to information, enhancing research quality, and facilitating interdisciplinary research. Next, We discusses the challenges in AI-assisted research, including ensuring academic integrity, maintaining content originality, avoiding over-reliance on AI, and data security and privacy protection. We can reasonably expect that medical research will enter a new era of greater efficiency and innovation, which will ultimately benefit human health.
References：
[1] IDC. "The Digitization of the World: From Edge to Core." 2018.
[2] Zhang et al. "NLP-powered systematic review in biomedical research." Nature Machine Intelligence, 2020.
[3] Johnson et al. "Machine Learning for Electronic Health Record Analysis." JAMA, 2019.
[4] Smith et al. "AI in Academic Writing: A Survey." Journal of Scholarly Publishing, 2023.
[5] Lee et al. "Predicting Scientific Breakthroughs using Machine Learning." Science, 2021.
[6] Cancer Research UK. "AI in Literature Review: A Case Study." 2022.
[7] WHO. "Global Research Collaboration and AI." 2023.
[8] Brown et al. "Quality Assessment of AI-Assisted Medical Research." The Lancet Digital Health, 2022.
[9] Nature Materials. "AI Predicts Novel Biomaterial." 2023.
[10] Wang et al. "Advanced Plagiarism Detection in the AI Era." Ethics in Science and Technology, 2022.
[11] IBM Research. "Deep Learning for Semantic Plagiarism Detection." 2023.
[12] Harvard Medical School. "AI in Research Innovation Assessment." 2022.
[13] NIH. "AI-Assisted Ethical Review in Biomedical Research." 2023.
[14] Taylor & Francis. "AI Use and Citation Rates in Medical Research." 2023.
[15] Stanford University. "Critical Thinking in the Age of AI." 2022.
[16] IEEE. "Homomorphic Encryption in Medical AI Systems." 2023.
[17] Google AI. "Personalized Research Assistants: A Prototype Study." 2023.
[18] The Lancet. "AI-Driven Collaboration in Multi-Center Clinical Trials." 2023. [19] Nature. "AI in Research Planning and Funding." 2022.
[20] European Commission. "Ethics Guidelines for Trustworthy AI." 2023.
Content by Wei Sun

Content by Wei Sun
Audio by Google
Music by Dopestuff @MelodyLoops
Download free music at https://www.melodyloops.com

Disclaimer:
This podcast is for informational purposes only. The technical content and market analysis shared here, including guest opinions, do not constitute professional or financial advice. We cannot guarantee accuracy or completeness of information. Listeners assume all risks from using our content. Consult professionals before making investment or technical decisions.

Contact: victor@nexture.nz

This episode explores the application of AI agents in business, particularly the “one-person's big company” model.
Big up to Mr. Wang, company commander, my mentor, who first put forward the concept of "one person's big company."

Content by Wei Sun
Audio by Google
Music by Dopestuff @MelodyLoops
Download free music at https://www.melodyloops.com

Disclaimer:
This podcast is for informational purposes only. The technical content and market analysis shared here, including guest opinions, do not constitute professional or financial advice. We cannot guarantee accuracy or completeness of information. Listeners assume all risks from using our content. Consult professionals before making investment or technical decisions.

Contact: victor@nexture.nz