Crazy Wisdom

Crazy Wisdom

Episode #392: From Digital Footprints to Transhumanism: Navigating the AI-Driven Future

Episode #392 · September 16, 2024 · 58 min

MP3 · Apple Podcasts · Spotify

About this episode

In this episode of the Crazy Wisdom podcast, Stewart Alsop speaks with Anand Dwivedi, a Senior Data Scientist at ICE, returning for his second appearance. The conversation covers a range of topics including the evolution of machine learning models, the integration of AI into operating systems, and how innovations like Neuralink may reshape our understanding of human-machine interaction. Anand also touches on the role of cultural feedback in shaping human learning, the implications of distributed systems in cybersecurity, and his current project—training a language model on the teachings of his spiritual guru. For more information, listeners can connect with Anand on LinkedIn.

Check out this GPT we trained on the conversation!

Timestamps

00:00 Introduction and Guest Welcome
00:25 Exploring GPT-4 and Machine Learning Innovations
03:34 Apple's Integration of AI and Privacy Concerns
06:07 Digital Footprints and the Evolution of Memory
09:42 Neuralink and the Future of Human Augmentation
14:20 Cybersecurity and Financial Crimes in the Digital Age
20:53 The Role of LLMs and Human Feedback in AI Training
29:50 Freezing Upper Layers and Formative Feedback
30:32 Neuroplasticity in Sports and Growth
32:00 Challenges of Learning New Skills as Adults
32:44 Cultural Immersion and Cooking School
34:21 Exploring Genetic Engineering and Neuroplasticity
38:53 Neuralink and the Future of AI
39:58 Physical vs. Digital World
41:20 Existential Threats and Climate Risk
45:15 Attention Mechanisms in LLMs
48:22 Optimizing Positive Social Impact
54:54 Training LLMs on Spiritual Lectures

Key Insights
  1. Evolution of Machine Learning Models: Anand Dwivedi highlights the advancement in machine learning, especially with GPT-4's ability to process multimodal inputs like text, images, and voice simultaneously. This contrasts with earlier models that handled each modality separately, signifying a shift towards more holistic AI systems that mirror human sensory processing.
  2. AI Integration in Operating Systems: The conversation delves into how AI, like Apple Intelligence, is being integrated directly into operating systems, enabling more intuitive interactions such as device management and on-device tasks. This advancement brings AI closer to daily use, ensuring privacy by processing data locally rather than relying on cloud-based systems.
  3. Neuralink and Transhumanism: Anand and Stewart discuss Neuralink’s potential to bridge the gap between human and artificial intelligence. Neuralink’s brain-computer interface could allow humans to enhance cognitive abilities and better compete in a future dominated by intelligent machines, raising questions about the ethics and risks of such direct brain-AI integration.
  4. Cultural Feedback and Learning: Anand emphasizes the role of cultural feedback in shaping human learning, likening it to how AI models are fine-tuned through feedback loops. He explains that different cultural environments provide varied feedback to individuals, influencing the way they process and adapt to information throughout their lives.
  5. Cybersecurity and Distributed Systems: The discussion highlights the dual-edged nature of distributed systems in cybersecurity. While these systems offer increased freedom and decentralization, they can also serve as breeding grounds for financial crimes and other malicious activities, pointing to the need for balanced approaches to internet freedom and security.
  6. Generative Biology and AI: A key insight from the episode is the potential of AI models, like those used for language processing, to revolutionize fields such as biology and chemistry. Anand mentions the idea of generative biology, where AI could eventually design new proteins or chemical compounds, leading to breakthroughs in drug discovery and personalized medicine.
  7. Positive Social Impact Through Technology: Anand introduces a thought-provoking idea about using AI and data analytics for social good. He suggests that technology can help bridge disparities in education and resources globally, with models being designed to measure and optimize for positive social impacts, rather than just profits or efficiency.
Episode transcript
Stewart Alsop III00:00

Welcome to the Crazy Wisdom Podcast. This podcast is for you. If you have an insane drive to find the truth of things, it's not the good answers that we seek, but the good questions. I interview a range of different guests from many different fields, all with the intention to uncover the simple truths that are hidden in plain sight. Most people don't want to go there. I go there, my guests go there, and you benefit. Please let me know if you enjoy these episodes and as always, subscribe on itunes, Spotify, or wherever you listen to the podcasts.

Stewart Alsop III00:37

Welcome to the Crazy Wisdom Podcast. My guest today is Anand Dwivedi. He is a senior Data scientist at ice. Welcome to the show for the second time.

Anand Dwivedi00:47

Oh, thanks you thanks for having me back. Last time was so much fun. I always love having these interesting crazy discussions with people. So yeah, happy to be back.

Stewart Alsop III00:59

Very cool. We'll jump right into machine learning. What's the craziest thing you learned how to do in the last week or last month in regards to machine learning?

Anand Dwivedi01:08

In the last months things have been like I have been truly more intrigued by these newer models that now like GPT4O that processes everything together is now such a fascinating.

Stewart Alsop III01:22

What do you mean processes everything together?

Anand Dwivedi01:25

So they previously generally people were translating between tokens, right? Like image got into tokens, then those image tokens got transferred into either text tokens, video tokens and doing these translations from one to the other GPT4. Oh and I've been learning a lot about it still because there's so much always going on being one of the initial models to truly process all of the three types at the same time. So imagine text of a dog, video of a dog, voice of a dog, all of the tokens being processed at the same time. For the models to train on is a different strategy than what the world was in before where people just had the large language model which was just language into tokens and then they did language to images, language to sound. So in that way models now processing like getting closer to how we process all sensory inputs together like the GPT4.0 essentially they made the claim that they have trained that 4O model and mixture of experts using that technology. So I've been trying to use to see how different it is from GPT4. Like so many people are trying it out and does that have a difference in its ability to do one type of tasks or the other. But now from at least the last conversation we had and I truly think about it like have more things become novel. But because of this GPU grab of all big resources Truly like allocating a lot of it with them. I feel we have now we've reached an interesting space where we are still trying to figure out how these technologies will get more impacted. But has nothing been so mind blowing to be like oh wow, the dynamic again is going to shift again?

Your first full transcript is free. After that, an email opens every transcript in the index — a list of readers we can write to, not a guest book.

Subscribe