Deep Learning

Unlocking Language Models: The Power of Prompt Engineering (Ep. 238)

Join me on an enlightening journey through the world of prompt engineering. Explore the multifaceted skills and strategies involved in harnessing the potential of large language models for various applications. From enhancing safety measures to augmenting models with domain knowledge, learn how prompt engineering is shaping the future of AI.
 
References
https://arxiv.org/pdf/2109.01652.pdf
https://arxiv.org/abs/2201.11903
https://arxiv.org/abs/2302.00923
https://www.mihaileric.com/posts/a-complete-introduction-to-prompt-engineering
https://www.axios.com/2023/02/22/chatgpt-prompt-engineers-ai-job

The new dimension of AI: Vector Databases (Ep. 236)

Let's delve into the emerging trend in database design – or is it really a new trend? The realm of vector databases and their revolutionary influence on AI and ML is making headlines. Come along as we investigate how these groundbreaking databases are revolutionizing the landscape of data storage, retrieval, and processing, ultimately unlocking the complete potential of artificial intelligence and machine learning. But are they genuinely as innovative as they seem?
 
References
https://partee.io/2022/08/11/vector-embeddings/
https://blog.det.life/why-you-shouldnt-invest-in-vector-databases-c0cd3f59d23c
https://medium.com/@ryanntk/choosing-the-right-embedding-model-a-guide-for-llm-applications-7a60180d28e3
 
 

Debunking AGI Hype and Embracing Reality (Ep. 233)

In this thought-provoking episode, we sit down with the renowned AI expert, Filip Piekniewski, Phd, who fearlessly challenges the prevailing narratives surrounding artificial general intelligence (AGI) and the singularity. With a no-nonsense approach and a deep understanding of the field, Filip dismantles the hype and exposes some of the misconceptions about AI, LLMs and AGI. Join us as we delve into the real-world implications of AI, separating fact from fiction, and gaining a firm grasp on the tangible possibilities of AI advancement. If you're seeking a refreshingly pragmatic perspective on the future of AI, this episode is an absolute must-listen.
 
Filip Piekniewski Bio
Filip Piekniewski is a distinguished computer vision researcher and engineer, specializing in visual object tracking and perception. He approaches machine learning with a pragmatic mindset, recognizing its current limitations. Filip earned his Ph.D. from Warsaw University, where he explored neuroscience and later joined Brain Corporation in San Diego. His extensive study of neuroscience inspired him to develop innovative, bio-inspired machine learning architectures. Filip's unique blend of scientific curiosity and software engineering expertise allows him to quickly prototype and implement new ideas. He is known for his realistic perspective on AI, debunking AGI hype and focusing on tangible advancements.
 
Sponsors
Finally, a better way to do B2B research. NewtonX The World’s Leading B2B Market Research Company
Explore the Complex World of Regulations. Compliance can be overwhelming. Multiple frameworks. Overlapping requirements. Let Arctic Wolf be your guide.Check it out at https://arcticwolf.com/datascience
Amethix works to create and maximize the impact of the world’s leading corporations and startups, so they can create a better future for everyone they serve. We provide solutions in AI/ML, Fintech, Defense, Robotics and Predictive maintenance.
 
References
https://twitter.com/filippie509
http://blog.piekniewski.info/ (On limits of deep learning and where to go next with AI.)
 
 

Warning! Mathematical Mayhem Ahead: Demystifying Liquid Time-Constant Networks (Ep. 228)

Hold on to your calculators and buckle up for a wild mathematical ride in this episode! Brace yourself as we dive into the fascinating realm of Liquid Time-Constant Networks (LTCs), where mathematical content reaches new heights of excitement.
In this mind-bending adventure, we demystify the intricacies of LTCs, from complex equations to mind-boggling mathematical concepts, we break them down into digestible explanations.
 
References
https://www.science.org/doi/10.1126/scirobotics.adc8892
https://spectrum.ieee.org/liquid-neural-networks#toggle-gdpr
 

Efficiently Retraining Language Models: How to Level Up Without Breaking the Bank (Ep. 227)

Get ready for an eye-opening episode! 🎙️
In our latest podcast episode, we dive deep into the world of LoRa (Low-Rank Adaptation) for large language models (LLMs). This groundbreaking technique is revolutionizing the way we approach language model training by leveraging low-rank approximations.
Join us as we unravel the mysteries of LoRa and discover how it enables us to retrain LLMs with minimal expenditure of money and resources. We'll explore the ingenious strategies and practical methods that empower you to fine-tune your language models without breaking the bank.
Whether you're a researcher, developer, or language model enthusiast, this episode is packed with invaluable insights. Learn how to unlock the potential of LLMs without draining your resources.
Tune in and join the conversation as we unravel the secrets of LoRa low-rank adaptation and show you how to retrain LLMs on a budget.
Listen to the full episode now on your favorite podcast platform! 🎧✨
 
References
LoRA: Low-Rank Adaptation of Large Language Models https://arxiv.org/abs/2106.09685
Low-rank approximation https://en.wikipedia.org/wiki/Low-rank_approximation
Attention is all you need https://arxiv.org/pdf/1706.03762.pdf
 

Deep learning vs tabular models (Ep. 217)

Deep learning methods are not as effective with tabular data. Here is why, and what to do about it.
 
Sponsors
If you're ready to take your WiFi game to the next level, head over to asus.click/ZenWiFi_XD5 or check out the show notes for this episode. Trust me, with ASUS ZenWiFi XD5, you'll get the best WiFi experience ever!
 
References
https://paperswithcode.com/methods/category/deep-tabular-learning 
https://m-clark.github.io/posts/2022-04-01-more-dl-for-tabular/
 

Accelerating Perception Development with Synthetic Data (Ep. 214)

In this episode I am with Kevin McNamara, founder and CEO of Parallel Domain. We speak about a very effective method to generate synthetic data that is currently in production at Parallel Domain.
Enjoy the show!
 
 
References
Parallel Domain Synthetic Data Improves Cyclist Detection (blog post):
https://paralleldomain.com/parallel-domain-synthetic-data-improves-cyclist-detection/ 
 
Beating the State of the Art in Object Tracking with Synthetic Data:
https://paralleldomain.com/beating-the-state-of-the-art-in-object-tracking-with-synthetic-data/ 
 
Parallel Domain Open Synthetic Dataset:
https://paralleldomain.com/open-datasets/bicycle-detection 
 
How Toyota Research Institute Trains Better Computer Vision Models with PD Synthetic Data (interview):
https://www.youtube.com/watch?v=QIYttoVxf2w
 
Career Opportunities:
https://paralleldomain.com/careers

Edge AI applications for military and space [RB] (Ep. 213)

Our Sponsors
NordPass Business has developed a password manager, that will save you a lot of time and energy whenever you need access to business accounts, work across devices, even with the other members of your team, or whenever you need to share sensitive data with your colleagues, or make payments efficiently. All this with the highest standard of cyber secure technology.
See NordPass Business in action now with a 3-month free trial herehttps://nordpass.com/DATASCIENCE with code DATASCIENCE
 
 
Amethix works to create and maximize the impact of the world’s leading corporations and startups, so they can create a better future for everyone they serve. We provide solutions in AI/ML, Fintech, Healthcare/RWE, and Predictive maintenance.
 

Machine learning is physics (Ep. 211)

What if we borrowed from physics some theories that would interpret deep learning and machine learning in general?Here is a list of plausible ways to interpret our beloved ML models and understand why they works, or they don't.Enjoy the show!
Our Sponsors
NordPass Business has developed a password manager, that will save you a lot of time and energy whenever you need access to business accounts, work across devices, even with the other members of your team, or whenever you need to share sensitive data with your colleagues, or make payments efficiently. All this with the highest standard of cyber secure technology.
See NordPass Business in action now with a 3-month free trial herehttps://nordpass.com/DATASCIENCE with codeDATASCIENCE
 
 
Amethix works to create and maximize the impact of the world’s leading corporations and startups, so they can create a better future for everyone they serve. We provide solutions in AI/ML, Fintech, Healthcare/RWE, and Predictive maintenance.
 

Autonomous cars cannot drive. Here is why. (Ep. 210)

If you think that the problem of self-driving cars has been solved, think twice.As a matter of fact, the problem of self-driving cars cannot be solved with the technical solutions that companies are currently considering. Don't get fooled by marketing and PR on social media. Whoever is telling you they solved the problem of driving a vehicle fully autonomously, they are lying.Here is why.
 
Our Sponsors
Explore the Complex World of Regulations. Compliance can be overwhelming. Multiple frameworks. Overlapping requirements. Let Arctic Wolf be your guide.Check it out at https://arcticwolf.com/datascience
 
Amethix works to create and maximize the impact of the world’s leading corporations and startups, so they can create a better future for everyone they serve. We provide solutions in AI/ML, Fintech, Healthcare/RWE, and Predictive maintenance.