Steven Data Talk

Steven Data Talk

By StevenEducation
Download on the App Store

Steven Data Talk episodes

  • EP3 Can AI Find Hidden Dangers in Your Code? 🤖💻 LLMs vs. Software Vulnerabilities! (feat. Astrid)(EN)

    Steven Data Talk - Can AI Find Hidden Dangers in Your Code? 🤖💻 LLMs vs. Software Vulnerabilities! (feat. Astrid)(EN Dubbed)Ever wondered how cutting-edge AI like Large Language Models (LLMs) are revolutionizing software security? In this episode of Steven Data Talk, we chat with Astrid, an expert researcher from University College London (UCL), specializing in using LLMs to detect code vulnerabilities.Join us as we objectively explore:The New Frontier: How LLM-based vulnerability detection differs from traditional machine learning approaches (like those using NLP or Graph Neural Networks).LLM Advantages: Exploring the potential of zero-shot/few-shot learning, flexibility, and generalization to new, unseen threats.Fine-Tuning is Key: Why pre-trained models often need fine-tuning on specific vulnerability data (often from open-source like GitHub).Current Hurdles: The challenges of accuracy, explainability (critical in security!), handling complex codebases, and the limitations of LLMs with non-source code (like binary or intermediate representations).Hybrid Power: How combining LLMs with traditional program analysis techniques (like program slicing and taint analysis) offers a promising path forward.The Role of AI Agents: Discussing the potential (and current difficulties) of using multi-agent systems for complex code analysis.Real-World Research: Astrid shares insights from her latest work on using LLMs and program slicing for Android malware analysis – a novel approach to tackle huge, complex apps!Whether you're a developer, security professional, AI enthusiast, or just curious about the future of code safety, this discussion offers valuable insights!High-Level Timeline:00:00:00 - Guest Introduction (Astrid) & Research Area (LLMs for Code Vulnerability)00:00:30 - How LLMs Differ from Traditional Methods00:01:44 - Fine-tuning LLMs & Data Sources (Open vs. Closed Source)00:02:51 - Key Challenges: Accuracy & Explainability00:05:06 - LLM Generalization vs. Accuracy Trade-off & New Threats00:06:21 - Overview of Traditional Code Analysis Techniques (ML, GNNs, NLP)00:08:31 - Potential of Multi-Modal & AI Agent Approaches00:10:31 - Combining LLMs with Traditional Program Analysis Tools (Workflow Innovation)00:12:24 - LLM Limitations with Non-Source Code Representations00:13:17 - Guest's Research (Astrid): Android Malware Analysis using LLMs & Program Slicing00:15:01 - Explaining the Program Slicing Technique00:16:21 - Conclusion & Final ThoughtsLike this discussion? 👍 Subscribe for more insights from Steven Data Talk! Let us know your thoughts in the comments below! 👇#LLM #SoftwareSecurity #CodeVulnerability #AI #Cybersecurity #ProgramAnalysis #MachineLearning #DataScience #StevenDataTalk

    17 min
  • Steven Data Talk |EP2| Data Talk with Special Guest - Anh

    In this episode of Data Talk, we are going to discuss various topics about data science, machine learning, llm, with special Guest - Anh.She is very skilled AI practitioner, focusing on language models, natural language processing, and data science, with field experience and research enthusiasm.Don't miss it!Premiere CET 11 Nov 15:30.

    35 min
  • Steven Data Talk |EP1| Data Science and AI in Financial Industry [EN-Machine Translation Version]

    Data Talk [EN-Machine Translation Version]Data Science and AI in Financial Industry with GiorgioGiorgio is an experienced professional financial industry practitioner and now content creator and independent fx trader.The original talk was in Chinese, and the machine translated version is also published here.In this episode of the podcast, host Steven introduces the topic of data science in finance and AI applications. He welcomes Georgio, a trader and self-media entrepreneur, to discuss:- **Data Science in Trading**: Georgio explains the difference between subjective and quantitative trading, highlighting the importance of historical data analysis for strategy evaluation. 📊- **Self-Media Entrepreneurship**: Georgio shares how data analysis shapes video content and audience engagement, emphasizing the need to adapt based on viewer feedback. 🎥- **AI Tools**: They discuss the role of AI, like ChatGPT, in improving efficiency and creativity, while addressing concerns about job displacement. 🤖- **Advice for Students**: Steven encourages college students to engage with AI technologies, stressing the importance of understanding their pros and cons. 🎓Introduction: Career background and entrepreneurial journey.Importance of data science and analytics in past work.Impact of data science and analytics on current entrepreneurship.Use of AI language models in work and learning.Opinion on ChatGPT.Interest in Perplexity tools.Trust in AI tools and replacement of human labor.Blue ocean opportunities in data science and AI in finance.Advice for university students on learning and career.00:00 - Introduction and Podcast Overview00:34 - Host and Guest Introductions01:05 - Importance of Data Science in Financial Industry02:45 - Trading Types: Subjective vs. Quantitative04:05 - Historical Data Analysis in Trading06:30 - Data in Self-Media Entrepreneurship09:20 - Audience Analysis and Content Planning11:30 - Data’s Role in Optimizing Self-Media Content13:55 - Chat GPT and Language Model Applications17:45 - AI in Trading and Potential Replacements20:25 - Chat GPT’s Benefits and Limitations23:45 - Future of Data Science and AI in Financial Industry26:00 - Closing Remarks and Future Topics#LearnByDoing #DataScience #Finance #AIApplications #Trading #SelfMedia #AudienceEngagement #DataAnalysis #QuantitativeTrading #SubjectiveTrading #ChatGPT #AIEfficiency #StudentAdvice #Podcast #Innovation #Entrepreneurship

    27 min

About Steven Data Talk

From the publisher's feed

Machine Learning, AI, Data Science