Using LLMs for loan approvals, the future of generative AI, and why I'm all in on HONDA
Plus upcoming panel discussion, text-guided image-to-image generation with Stable Diffusion, and a framework for generating synthetic data for LLMs
What’s up everyone!
Thank you to everyone who joined the LangChain ZoomCamp on Friday!
Don't worry if you missed it; I’ve got the recording for you. I was having issues uploading the video to the podcast section today. Instead, here’s the link to the Zoom recording where you can watch or download the videos.
The notebooks are available on GitHub.
Webinar: Using LLMs for Loan Approvals
My friends at SingleStore invite you to a hands-on code sharing and live demo session to learn firsthand how LLMs, OpenAI, and Vector DBs are used for loan approvals in the Fintech space.
Can't make the live session? No worries! All registrants will receive the recording and access to the GitHub assets post-webinar.
What You’ll Learn:
The core concepts of Large Language Models and their pivotal role in shaping predictive analytics for loan approvals.
Dive into Vector DBs, OpenAI, and Langchain, understanding their seamless integration to bolster the predictive prowess of your applications.
Experience a real-time demo and code sharing, empowering you with the practical tools to translate these technologies into tangible outcomes.
Acquaint yourself with the emerging paradigms in banking technology and how predictive analytics stands as a beacon for streamlined and intelligent loan decisions.
The Future of Generative AI: Vision and Challenges
Last week, I hosted a panel discussion called The Future of Generative AI: Vision and Challenges, where Meryem Arik, Rajiv Shah, Sandeep Singh, Niels Bantilan, and Christophe Chabbert provided insights into the future of generative AI, its challenges, and the opportunities it presents across various industries.
This session was brought to you and sponsored by the Generative AI World Summit, which takes place at MLOps World Summit in Austin, TX, from October 25th through October 26th (with a virtual component).
You can register for the conference with the discount code 'harpreet' for 75$ off your ticket price.
It was awesome to have experts from various fields come together to explore the boundless potential of generative AI and the hurdles it presents. What emerged from this engaging conversation was a glimpse into the exciting advancements and the critical considerations that lie ahead.
The panellists unanimously agreed that generative AI is set to revolutionize multiple industries in the coming decade.
They foresee language models as pivotal in fields like biology and material sciences, where AI can aid groundbreaking discoveries and innovations. But it doesn't stop there. The experts also believe that AI and language models will become integral to every business workflow, transforming how organizations operate and interact with customers.
One of the most fascinating aspects discussed was the evolving role of generative AI in shaping human interaction with technology.
Imagine a future where seamless interactions between text, image, and audio modalities become the norm. It's a world where technology understands and responds to us more naturally and intuitively. The panellists envisioned this paradigm shift and highlighted its potential for enhancing our relationship with machines.
However, amidst the excitement, concerns were raised about the spread of misinformation.
With the power of generative AI, there is a pressing need for platforms, technology, and regulators to address this issue and safeguard the authenticity of information. The panellists recognized this challenge and stressed the importance of responsible deployment and ethical considerations.
Deploying language models, as it turns out, is without its challenges.
Scaling and efficient multi-model inference were discussed as crucial hurdles to overcome. The panellists emphasized the need to architect safe systems around these models, ensuring their potential is harnessed responsibly and securely.
As with any promising technological advancement, generative AI also has its limitations. The computational expense required for training and inference was acknowledged as a significant obstacle. Furthermore, the panellists stressed the importance of developing ethical and safe systems, ensuring that AI remains a tool for the betterment of society.
This entire summary was generated using LangChain. You can see how I got here by checking out the following notebook.
Preparing for the widespread adoption of generative AI emerged as a key theme throughout the discussion.
The panellists emphasized the need for industries to invest in data infrastructure and ensure clean and labelled data. They recognized that experimentation and exploration are essential to harness generative AI's potential fully.
Lastly, the panellists grappled with the question of a saturation point for adding more knowledge to language models. Opinions were divided, with some expressing concerns about the potential risks of an ever-expanding knowledge base.
In contrast, others saw it as an opportunity to deepen our understanding of the world.
As I reflect on the insights shared during the panel discussion, it becomes clear that we stand at the precipice of a remarkable transformation. Generative AI can reshape industries, revolutionize human-machine interaction, and unlock unprecedented possibilities. However, we must navigate this path cautiously, investing in responsible deployment, ethical considerations, and robust systems.
If we embrace these challenges and seize the opportunities that lie ahead, the future of generative AI holds immense promise for a better tomorrow.
🗓️ Schedule of upcoming series of panel discussions
I’ve partnered with the Generative AI World Summit to bring you a series of four-panel discussions. These are interactive sessions, and I welcome your participation. I’ll have some seed questions prepared, but your participation will guide the direction of the conversation.
One registration link will get all the sessions on your calendar, and I look forward to having you there. You can register here.
🗓️ Thursday, October 5th at 4pm CST: From Academia to Industry: Bridging the Gap in Generative AI
This panel will discuss translating academic research in Generative AI into real-world applications. Experts will discuss the challenges of this transition and share success stories.
🗓️ Thursday, October 12th at 4pm CST: Generative AI in Production: Best Practices and Lessons Learned
This will discuss the practical aspects of deploying Generative AI solutions, the challenges, and the lessons learned.
🗓️ Thursday, October 19th at 4pm CST: Ethics and Responsibility in Generative AI
Given the power of Generative AI, this session can address the ethical implications, potential misuse, and the responsibility of researchers and practitioners in ensuring its safe and beneficial use.
You can register for the conference with the discount code 'harpreet' for 75$ off your ticket price.
✨ Blogs of the Week
This is a short, sweet, hands-on tutorial for doing text-guided image-to-image generation with Stable Diffusion.
It breaks down the process in a clear, easy-to-follow manner without any fluff. You can, of course, swap out the Stable Diffusion model for any other model on HuggingFace. I work at Deci AI, and we recently released an open-source diffusion model called DeciDiffusion, which you can check out here.
Here’s another blog that’s a beginner’s guide to LLMOps.
This blog is also hands-on with code and breaks down the LLMOps process into the following steps:
Selection of a base model
Adapting to tasks
Model evaluation
Deployment and monitoring
It also touches on the difference between LLMOps and MLOps.
This is another hands-on blog but with minimal coding. The blog provides a tutorial for non-engineers on how to build a chatbot using the LLaMA 2 base model without writing code. It introduces Hugging Face's tools—Spaces, AutoTrain, and ChatUI—which facilitate fine-tuning and deploying the model to a chat app. The tutorial is structured step-by-step, explaining how to create an AutoTrain Space, launch model training, and more to show that machine learning and chatbot development can be accessible to everyone, regardless of their technical background.
🛠️ GitHub Gems
SciPhi is a configurable Python framework for generating synthetic/fine-tuning data, and for robust evaluation of LLMs.
At its core, SciPhi offers:
Configurable Data Generation: Efficiently produce LLM-mediated synthetic training and tuning datasets tailored to your needs.
The Library of Phi: The Library of Phi is an initiative sponsored by SciPhi. Its primary goal is to democratize access to high-quality textbooks. The project utilizes AI-driven techniques to generate textbooks by processing information from the MIT OCW course web pages.
I’m participating in the NeurIPS Large Language Model Efficiency Challenge with some community members from the Deep Learning Daily community, and we’ll be testing out their Textbooks Are All You Need dataset (hosted on HuggingFace).
This seems like a cool project and one worth starring. Check it out on GitHub.
💡 My Two Cents
In the realm of data science, big names often captivate aspiring professionals. Initially, I too, was enchanted by the allure of FAANG (Facebook, Apple, Amazon, Netflix, Google) – these companies (from the outside) were known for their cutting-edge technologies and lucrative positions.
However, as I delved deeper into the domain, a different acronym started resonating with me: HONDA.
HuggingFace - This isn’t just a company; it’s a revolution in code and community. In a world guarded by paywalls, HuggingFace throws open the gates of knowledge, beckoning all curious minds. Their open-source camaraderie is a rallying cry for the democratization of information.
OpenAI - The audacity of OpenAI’s models epitomizes the zenith of what deep learning can achieve, pulling me further from the banal analytics of my past.
NVIDIA - From a graphics titan to the backbone of AI, NVIDIA’s metamorphosis is nothing short of legendary. They’re not just fueling computations; they’re crafting the sinews and bones of the AI behemoth.
DeepMind - This is where deep learning is not just practiced but worshipped. DeepMind doesn’t follow the playbook; it writes it. Every innovation challenges the status quo, daring us to dream bigger.
Anthropic - In the relentless tide of AI advancement, Anthropic emerges as the sage, urging us to wield the sword of innovation with a grip of understanding and control.
It's not just about the big names or hefty paychecks. It's about being at the vanguard of deep learning, treading the uncharted waters that HONDA navigates with a community-centric ethos and a daring vision.
In HONDA, I found not just companies but movements shaping the future of AI, pushing the envelope, and inviting one to be more than just a cog in the machine.
That’s it for this one.
See you next week, and if there’s anything you want me to cover, shoot me an email.
Cheers,
Harpreet