AWS CodeWhisperer… The Copilot Killer?

TL;DR

"AWS releases Code Whisperer, a free code generation tool, along with other developments in AI and machine learning capabilities."

Transcript

it is April 14 2023 and you're watching the code report we're living through a development tooling Space Race right now and GitHub co-pilot is just the tip of the rocket there's a long shaft of other tools competing in the space that's growing girthier every day all of which are seated by Massive foundational models yesterday Amazon web services re... Read More

Key Insights

• 💻 AWS released a new code generation tool called Code Whisperer that competes with GitHub's Co-pilot and is 100% free for everyone. It has unique features and can generate code based on context in VS Code and JetBrains.
• 💡 AWS also launched Web Service Bedrock that allows developers to access large foundational AI models on their infrastructure, making it possible for non-experts to create customized AI applications. It includes image models like Stability AI and LLMS like AI 21 Labs Jurassic models.
• 💰 AWS made its ML training EC2 instances, based on inferentia chips, generally available. These custom-built chips can train ML models at about 50% of the normal cost on EC2, potentially reducing the cost of training an LLM from $10 million to $5 million.
• 🧠 Code Whisperer provides suggestions for code generation but usually suggests fewer lines of code compared to Co-pilot. It excels in understanding project context and generating appropriate code snippets, making it useful for reducing boilerplate code.
• 🧪 Code Whisperer shows potential for generating regular expressions and unit tests, although more complex patterns can be harder to prompt. It also offers transparency by providing references to code in the training data, reducing the likelihood of accidental code theft.
• 🕸️ Code Whisperer includes security scan features that check for OWASP issues in web applications, analyze smart contracts, and provide guidance on using AWS properly. Users get 50 security scans per month.
• ⚖️ The legal landscape around AI tools like Co-pilot and Code Whisperer is uncertain, with GitHub Co-pilot facing a class action lawsuit and Stability API being sued by Getty Images. Legal challenges may impact the future of these tools. ⏳ Code Whisperer represents the rapid development of AI tools and code generation capabilities. It demonstrates the potential for AI to evolve further and revolutionize programming, but the full extent of its impact and limitations is yet to be determined.

Questions & Answers

Q: How does Code Whisperer compare to GitHub Co-Pilot?

Code Whisperer and GitHub Co-Pilot are similar in generating code suggestions, but Code Whisperer tends to generate code line by line while Co-Pilot suggests larger chunks of code at once.

Q: Can Code Whisperer write regular expressions?

Code Whisperer is useful for simple regular expressions like phone numbers, but it becomes more challenging to prompt as the pattern gets more complex.

Q: What are some unique features of Code Whisperer compared to GitHub Co-Pilot?

Code Whisperer provides transparency by offering references to the code in the training data and has security scans for web applications, checking for OWASP issues and analyzing smart contracts.

Q: How does Bedrock enable developers to build AI applications without high-end hardware?

Bedrock allows developers to access large foundational models on AWS infrastructure, making it possible to build highly customized AI applications without the need for GPUs.

Answer: Overall, AWS's Code Whisperer is a promising code generation tool that competes with GitHub Co-Pilot, offering some unique features like transparency and security scans. It is free to use and provides valuable support for developers.

Summary & Key Takeaways

• AWS has launched Code Whisperer, a free code generation tool that competes with GitHub Co-Pilot.

• AWS has also introduced Bedrock, which allows developers to access foundational models on AWS infrastructure for building AI applications.

• AWS has made its ml training EC2 instances available, which use custom-built chips for training machine learning models at half the cost.