DEV Community: pravintargaryen

DeepMind at Google: Denny Zhou

pravintargaryen — Thu, 12 Dec 2024 17:09:33 +0000

The buzz around Large Language Models (LLMs) is impossible to ignore. If you’ve been keeping up with tech news since 2022, you’ve likely heard the term. But do you really know what LLMs are all about? Despite being marketed as artificial intelligence, LLMs are essentially advanced prediction models trained to excel at guessing the next word or token in a sequence.

Much like Google’s simple search box unlocks access to the world’s knowledge, LLMs have revolutionized AI since their rise to prominence in early 2024. However, beneath their polished outputs lies a fundamental question: Are LLMs truly intelligent, or are they just exceptionally well-trained parrots?

Denny Zhou’s Take

Denny Zhou, a researcher at Google DeepMind, has a grounded perspective. In his recent lecture (https://www.youtube.com/watch?v=QL-FS_Zcmyo), Zhou addressed the reasoning abilities of LLMs, comparing them to teaching a parrot. “It just repeats mostly what we say but with an answer at the end,” he remarked.

Zhou believes AI should mimic human learning, where understanding grows from just a few examples. To illustrate, he shared a charming anecdote about his children:

Kid 1: “What’s 17 times three?”
Kid 2: “I don’t know.”
Kid 1: “What’s 10 times 3?”
Kid 2: “30.”
Kid 1: “What’s 7 times 3?”
Kid 2: “21.”
Kid 1: “Add both of them.”
Kid 2: “51.”

Reflecting on this, his elder child quipped, “Chain-of-thought prompting works on my little brother too!”

The Power and Pitfalls of Chain-of-Thought Reasoning

Zhou highlighted the potential of chain-of-thought prompting, where intermediate reasoning steps between the prompt and response significantly enhance an LLM’s performance. When an LLM reasons step-by-step, its confidence in the final answer increases compared to directly generating a response.

But there’s a caveat. Just like humans, LLMs can be easily distracted by irrelevant context, leading to flawed reasoning. This limitation underscores the complexity of teaching machines to "think."

What Lies Ahead

Denny Zhou is at the forefront of teaching the next generation of LLMs how to reason. His research explores how AI can mimic human-like learning and reasoning, a challenge that could redefine AI as we know it. Zhou’s work doesn’t just speculate; it builds on rigorous experiments and insights, paving the way for future breakthroughs.

For those interested in diving deeper into his research, Zhou’s publications are available at https://dennyzhou.github.io/.

A Final Word

Denny Zhou reminds us: “Always keep in mind that LLMs are probabilistic models of generating next tokens. They are not humans.” His cautionary message emphasizes the need to balance excitement with a realistic understanding of AI’s capabilities and limitations.

Indeed, Denny Zhou’s insights offer a glimpse into the intricate dance between human ingenuity and artificial reasoning. As we navigate this transformative era, one thing is clear—LLMs, while remarkable, are just another step in humanity’s quest to unravel the mysteries of intelligence.

Create a Django app in less than 2 minutes

pravintargaryen — Mon, 28 Oct 2024 17:11:45 +0000

Django is a Popular Web framework in the pythonic world since its inception in 2005 as there's consistent growth of developers and users alike ever since.

If you love Python in the Backend, creating a Django app is more fun than you thought.

Virtual environment (venv)

Virtual environment makes sure the packages and dependencies you install do not end up globally in the site-packges folder of your Python installation directory and in the current project folder.

The following command will create a virtual environment in the current working directory

py -m venv venv

You can activate your virtual environment using the command (in windows)

.\venv\Scripts\activate

Once you have your virtual environment activated, we are ready to go.

Installing Django

Django package can be installed with following command

py -m pip install Django

After the installation is completed, we can start a new Django project

django-admin startproject demo .

Notice the presence of '.' a dot at the end of the command. Omitting this dot will result in a nested project structure same as our app name 'demo1'(in this case)

Starting the Dev Server

Now that we have our demo project created, let's check if our django app is installed properly

py manage.py runserver

You will notice certain warnings about migrations which you can ignore for now

That's it. We now have our first Django app up and running

Cyclops UI - the browser for your Kubernetes Clusters

pravintargaryen — Mon, 05 Aug 2024 18:41:02 +0000

Have you ever wondered how Google can handle so much requests per minute convincingly without any lag or downtime and that too at Godspeed?

The answer lies in it's Data Center

What is a Datacenter?

A Datacenter is a place where thousands of Linux Servers(mostly) are racked (aka rack servers) working down the clock for serving each requests at a scale larger than anyone can imagine.

Imagine a whole football ground with Servers

Pic Courtesy : Google Data Center Gallery

This is the map of the Google Data Centers around the Globe

You can get a quick snapshot of what a Datacenter is at the moment.

What is a Cluster?

As the name suggests Clusters are group of anything, in this case a group of Servers. The following illustration somehow details it

Pic Courtesy : Google Datacenter

You just got to know about cluster in the Datacenter. Let's take dig deeper.

Kubernetes

What if we want to build or emulate a data center of our own. Here Comes Kubernetes. Kubernetes or K8s as is commonly helps Containerized applications run with resources they need to run. Kubernetes is actually a result of Google's own Cluster Brainchild operating system called Borg.

Kubernetes Cluster Architecture

The basic Kubernetes cluster consists of a Control Pane and a Node

https://www.linkedin.com/pulse/step-by-step-guide-setting-up-kubernetes-cluster-md-rezaul-karim-1fpuc/

The kubelet is the primary "node agent" that runs on each node

What is Containerization?

Containerization as the name suggests helps contain an application with the tools they need to run such as environment variables, packages etc..

Think of it like a Sandbox or Delivery Container where everything you need is kept so that it can run as a single isolated units

How to Containerize an Application?

Well we usually containerize an application with Docker. Docker can package an application and its dependencies in a virtual container that can run on any Linux, Windows, or macOS computer. This enables the application to run in a variety of locations, such as on-premises, in public (see decentralized computing, distributed computing, and cloud computing) or private cloud.

Normally to Containerize an Application you need your application and a Docker File. That's it. A Docker File is a simple YAML File that will have the necessary details like:

Base Image
Working Directory
COPY
RUN Command
PORT

This is more like the build, run command and env variables needed when running an application in production. Think of it like Deploying our application to Vercel or Netlify. We are Deploying our application to Docker Container in this case. The Dockerfile will create a Docker Image first and the Docker Image will be Containerized.

https://www.theserverside.com/blog/Coffee-Talk-Java-News-Stories-and-Opinions/compare-Docker-image-vs-container-difference-compare

Because Docker containers are lightweight, a single server or virtual machine can run several containers simultaneously.

Copyright: Wired Magazine

Think of this ship. Each Container hosts an individual application or file. You can literally containerize anything from Google Colab notebook to a single image file or an entire application. The Choice is yours.

Docker Hub

You might have probably heard Github by now. Similar to Github, which hosts the source code of our applications, we have Docker Hub which hosts the Docker Image of our applications at https://hub.docker.com/.

Minikube

Now that you have a basic understanding of Kubernetes and Docker, let's focus on Minikube.

Minikube is a lightweight and simplified version of Kubernetes, primarily used for local development and testing purposes.While Kubernetes are multi nodes in real world scenario, Minikube allows you to run a single-node Kubernetes cluster locally on your machine. So this will be ideal for playing with it.

That's it we have learnt all the basics, now we can focus on Cyclops UI

Cyclops UI

Cyclops is an user interface for managing and interacting with Kubernetes clusters. This is more like a browser for Kubernetes Cluster without which we have to rely on Command line Terminals for managing clusters.

Cyclops UI Demo

Click the below image to watch Cyclops UI Demo