- BeyondOps
- Posts
- Cloudflare is going on war with AI giants.
Cloudflare is going on war with AI giants.
Greetings, AI Enthusiast’s!
83% developers are using ChatGPT for coding, but the new findings about ChatGPT code accuracy is shocking.
With Clouldflare going on war with tech giants, is this a wake up call for AI models? Would this affect AI training ?
Lets explore!
This Week’s AI Insights:
CriticGPT fixes ChatGPT blunders
Cloudflare blocks AI chatbots
Salesforce's new Tiny Giant model
The European Union AI Act
More Interesting AI news
CriticGPT fixes ChatGPT blunders:
Image Source : OpenAI
ChatGPT might be writing code for you but researchers from Purdue University presented alarming findings that 52% of ChatGPT-generated code was incorrect.
Why does this matter?
If 52% of the code generated by ChatGPT is incorrect and not properly verified, think about the impact on the quality and reliability of the software.
The Essentials:
OpenAI was very much aware of this issue and they were building a new tool under the radar for quite some time. This new tool named “CriticGPT” acts as a helper to ChatGPT.
CriticGPT is a tool designed to identify errors from ChatGPT-generated code output.
CriticGPT is a tool that highlights the inaccuracies in ChatGPT-generated outputs.
The Bottom Line:
While AI models like ChatGPT are powerful tools, relying on them 100% is risky.
There is still a great need for human participation and input in coding and such.
Cloudflare blocks AI chatbot crawler
Cloudflare Inc., a global cloud provider company specializing in providing security and DDoS protection to millions of websites is going to war against the AI tech giants Google, Microsoft, OpenAI and others.
Why does this matter?
Not all would be happy to lend their data for AI models to get trained with. This solution by Clouflare has been made so easy for customers who are reluctant to not let AI bots to crawl through their website.
The Essentials:
Cloudflare's new free tool blocks out AI bot crawlers.
Cloudflare's new free tool is available to all including its free tier customers.
The Bottom Line:
With just one free tool, CloudFlare Customers who do not want to have AI bots to visit their websites and steal their data for training AI models dishonestly will benefit from Cloudflare's new free tool.
Salesforce's new Tiny Giant model
Image Source : Salesforce
Salesforce's new models performed well above their weight and paved the way for on-device Artificial Intelligence.
Why does this matter?
In this AI era, where LLMs are trained with billions or even trillions of parameters, a model which is trained with just 1 billion parameters was able to outperform parameters which are trained with 175 billion parameters is a remarkable achievement.
The Essentials:
Salesforce’s new model xLAM-1B named rightly as “Tiny Giant” outperforms AI giants like GPT-3.5 turbo.
xLAM-1B was trained using Salesforce proprietary APIGen pipelines curated for high-quality, diverse datasets for training.
xLAM-1B, opens new possibilities for edge computing.
xLAM-1B makes it suitable for On-device AI.
The Bottom Line:
Salesforce in this era of AI proved with these models that David can indeed take down Goliath.
In function calling tasks, xLAM-1B, the 1Billion parameter model, outperformed GPT-3.5 turbo models.
On the Berkeley function calling benchmark, xLAM-7B which is the 7Billion parameter model outperformed multiple GPT-4 models.
The European Union AI Act
The AI Act approved by the European Union Council is set to be rolled out soon. This proposed EU's AI act will make significant changes in how companies should approach AI.
Why does this matter ?
AI has been taking its shape in many different forms. But unfortunately there was no regulation in place. This new AI act will keep tab of the models been created at least in the European Union.
The Essentials:
The EU AI act will categorize AI systems into 4 risk system levels.
Unacceptable: The "No-Go" Zone : AI that manipulates people without them knowing will be banned outright.
High: The "Proceed with Caution" Zone : AI is used in job hiring or to determine a credit score and needs strict oversight and testing.
Limited: The "Be transparent" Zone : AI that creates deep fakes, must be identified as AI.
Minimal: The "Free to Use" Zone : AI-powered video games. These can be used without any additional requirements.
The implementation of this AI act will be rolled out first to the "Unacceptable" risk systems with them being eliminated in six months and "High" risk systems will take effect within 36 months.
The Bottom Line:
The countdown to compliance begins. Non-complaint AI systems penalties could reach €30 million or 6%of global turnover.
The future of AI regulation is here and it is the right time to adapt and be compliant.
Interesting AI news
Image Source : Meta
Meta’s 3D Gen can now create high-quality 3D images with just from a text prompt.
The 3D assets and textures can be created with a text prompt in less than a minute.
Image Source : NewScientist
Top Row: What the Monkey SawBottom Row : The images the AI system reconstructed based on the brain activity
Artificial Intelligence can reconstruct what someone is visually looking at based on the recordings of their brain activity.
In this finding, researchers reconstructed images of what a Monkey was looking at from its brain recordings. Fascinating, isn’t it ?
Meta is looking towards bringing AI to its metaverse games, including VR, AR and mixed reality games.
With AI in metaverse, the dynamics of the games will completely be changed.
How about having a talking selfie photo? Sounds cool, right? It is possible now.
With Hedra, you can just have a photo uploaded with audio and it will generate a talking photo in a rapid speed. Go try here.
FEEDBACK
Your feedback is precious and helps us tailor future emails to better suit your needs!
We welcome your feedback and insights. Please reply to this email with any comments or suggestions you'd like to share.
Appreciate your time.
AI-ways forward,Naveen Krishnan, Founder and Writer, The AI Catalyst Solutions Inc.
© 2024 The AI Catalyst Solutions Inc. All rights reserved. This newsletter and its content are protected by copyright law. No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of The AI Catalyst Solutions Inc., except in the case of brief quotations embodied in critical reviews and certain other noncommercial uses permitted by copyright law.