EPISODE · Apr 30, 2026 · 44 MIN
Computer Scientist Explains: How to Use AI Without Cheating | Faheem Ullah | AFP 52
from Apple Finch Pudding · host Apple Finch Pudding
Is the traditional literature review dead? In an era where AI can summarize 1,000 papers in seconds, the boundary between "working smart" and "losing depth" has never been thinner. As researchers, we are facing a fundamental shift: AI is no longer a futuristic concept—it is a pervasive reality that touches every domain of science. But as we automate our workflows, are we sacrificing the very creativity and learning curves that define a PhD? In this episode, Jeroen Schreel sits down with Faheem Ullah, an Assistant Professor in Computer Science at the University of Adelaide, to dismantle the hype and provide a technical roadmap for the modern academic. Faheem breaks down the hierarchical differences between AI, Machine Learning, and Deep Learning, while offering a pragmatic framework for using these tools without compromising your academic integrity. We dive deep into:- The 1-Hour Literature Review: How to turn a week-long manual search into a high-productivity hour using iterative search strings and pilot studies.- The Technical "Black Box": Understanding feature selection, hyperparameters, and why Deep Learning is uniquely suited for automation.- The Cost of "Free" AI: Why experts predict a jump to $100/month subscriptions as data center costs explode to $600 billion.- The Hidden Biases: A cautionary tale on how AI-generated avatars and data logging can introduce unintended gender and academic bias. (00:00) Intro(00:28) Ad(01:43) The Fascinating Energy Consumption of the Human Brain(03:06) Choosing Computer Science: Why the Future is Pervasive(03:52) AI vs. Machine Learning vs. Deep Learning(06:08) How Deep Learning Automates Feature Selection(08:18) Hyperparameters vs. Parameters in Model Training(10:14) Productivity vs. Creativity: The Two Schools of Thought(12:14) Reducing a Week of Literature Review to One Hour(14:18) Using Pilot Studies to Fact-Check AI Search Strings(15:51) Summarizing 1,000 Papers: Are We Losing Research Depth?(18:51) Listening to Research: The Future of "Reading" Papers(20:25) Why AI-Generated Papers Can't Reach Human Quality (Yet)(21:41) Practical AI: Using Models for Data Analysis and Visuals(22:55) Avoiding Bias: A Cautionary Tale of AI-Generated Avatars(24:41) The Paywall Problem: How AI Excludes Certain Research(26:03) Specialized Tools: Moving Beyond ChatGPT(28:05) The End of Cheap AI? $600 Billion Data Centers(30:47) Will OpenAI Include Ads in ChatGPT?(33:28) The 80% Rule: Responsibility and Fact-Checking(35:55) Why Traditional Essay Assignments are Now Obsolete(38:03) The AI Feedback Loop: Training Models on AI-Generated Data(39:20) Why Social Media Algorithms Suppress AI Content(41:20) Using AI for Research Infographics and Communication This episode is sponsored by ResearchRabbithttps://www.researchrabbit.ai Don't forget to join our monthly AFP newsletter: https://applefinchpudding.ck.page #AI #AcademicIntegrity #ResearcherTools #DeepLearning #MachineLearning #SciencePodcast
What this episode covers
Is the traditional literature review dead? In an era where AI can summarize 1,000 papers in seconds, the boundary between "working smart" and "losing depth" has never been thinner. As researchers, we are facing a fundamental shift: AI is no longer a futuristic concept—it is a pervasive reality that touches every domain of science. But as we automate our workflows, are we sacrificing the very creativity and learning curves that define a PhD? In this episode, Jeroen Schreel sits down with Faheem Ullah, an Assistant Professor in Computer Science at the University of Adelaide, to dismantle the hype and provide a technical roadmap for the modern academic. Faheem breaks down the hierarchical differences between AI, Machine Learning, and Deep Learning, while offering a pragmatic framework for using these tools without compromising your academic integrity. We dive deep into:- The 1-Hour Literature Review: How to turn a week-long manual search into a high-productivity hour using iterative search strings and pilot studies.- The Technical "Black Box": Understanding feature selection, hyperparameters, and why Deep Learning is uniquely suited for automation.- The Cost of "Free" AI: Why experts predict a jump to $100/month subscriptions as data center costs explode to $600 billion.- The Hidden Biases: A cautionary tale on how AI-generated avatars and data logging can introduce unintended gender and academic bias. (00:00) Intro(00:28) Ad(01:43) The Fascinating Energy Consumption of the Human Brain(03:06) Choosing Computer Science: Why the Future is Pervasive(03:52) AI vs. Machine Learning vs. Deep Learning(06:08) How Deep Learning Automates Feature Selection(08:18) Hyperparameters vs. Parameters in Model Training(10:14) Productivity vs. Creativity: The Two Schools of Thought(12:14) Reducing a Week of Literature Review to One Hour(14:18) Using Pilot Studies to Fact-Check AI Search Strings(15:51) Summarizing 1,000 Papers: Are We Losing Research Depth?(18:51) Listening to Research: The Future of "Reading" Papers(20:25) Why AI-Generated Papers Can't Reach Human Quality (Yet)(21:41) Practical AI: Using Models for Data Analysis and Visuals(22:55) Avoiding Bias: A Cautionary Tale of AI-Generated Avatars(24:41) The Paywall Problem: How AI Excludes Certain Research(26:03) Specialized Tools: Moving Beyond ChatGPT(28:05) The End of Cheap AI? $600 Billion Data Centers(30:47) Will OpenAI Include Ads in ChatGPT?(33:28) The 80% Rule: Responsibility and Fact-Checking(35:55) Why Traditional Essay Assignments are Now Obsolete(38:03) The AI Feedback Loop: Training Models on AI-Generated Data(39:20) Why Social Media Algorithms Suppress AI Content(41:20) Using AI for Research Infographics and Communication This episode is sponsored by ResearchRabbithttps://www.researchrabbit.ai Don't forget to join our monthly AFP newsletter: https://applefinchpudding.ck.page #AI #AcademicIntegrity #ResearcherTools #DeepLearning #MachineLearning #SciencePodcast
NOW PLAYING
Computer Scientist Explains: How to Use AI Without Cheating | Faheem Ullah | AFP 52
No transcript for this episode yet
Similar Episodes
No similar episodes found.