
OpenAI could possibly be the next gateway platform for the next major tech revolution. Developers are increasingly stating that ChatGPT has been saving them so much time through AI-assisted code that they can finally invest more time into other projects. The time of ‘code to cloud’ maybe end as ‘idea to code’ takes the spotlight. Of course, AI-assisted code is not perfect but the potential is there and only improving. Stephen O’Grady’s work in The New Kingmakers details the ever-increasing sway software developers hold over businesses as the apps or programs they develop have the power to make or break companies. This influence of software developers was driven by three main megatrends, open source, cloud, and online learning and sharing. Now the fourth has appeared, AI-assisted code.
Open Source, the Age of AWS, and GitHub
Prior to AWS, there was open source which was already impacting the startup industry. However, startups still required sizeable capital investments to build the necessary infrastructure. With AWS and cloud, leading to an increase in the number of startups being funded. Following AWS, was the onset of online learning and knowledge sharing for developers that came in the form of GitHub and other platforms like Stack Overflow. These platforms enabled increased collaboration across developers, faster development, and overall better codes.
AI-assisted Code
Now we’re at a new inflection point where knowing AI is likely to become essential for developers moving forward. AI-assisted code use cases include:
Writing code - ChatGPT or GitHub Copilot can generate code for simple functions or repetitive tasks such as database manipulation or database queries. The code itself may not always be accurate or optimized unless a clear specification is given.
Debugging - developers can input their codes into the AIs to have it suggest potential causes of errors and recommendations on how to fix them.
Code completion - AIs can complete portions of developers' code by predicting the subsequent lines or sections based on the existing code above.
Optimization - beyond troubleshooting, AIs can also generate recommendations for developers on ways to optimize and improve their code structure, readability, and performance.
Documentation - developers can input their code into ChatGPT and prompt it for suggestions on the appropriate documentation templates for the specific language and type of code being documented.
Explaining code - AIs can provide examples to explain coding concepts and techniques that developers are new to or have forgotten. Furthermore, AIs can also provide examples of best practices for various languages and code documentation.
Despite all these functionalities, we must remember that it is AI-"assisted" coding and is unable to replace developers. If you recall, people were also concerned that cloud could replace the need for developers, but today we know how unfounded that concern was. AI-assisted code is shaping up to be an essential tool for accelerating developers' work, making it easier for them to learn new skills and launch new projects.
Microsoft and OpenAI’s Competitive Moat
The lead OpenAI’s LLM and ChatGPT have over competing AIs is going to further reinforce Microsoft’s competitive moat that has already been established since its acquisition of GitHub and Visual Studio Code. So, Microsoft is going to own where developers collaborated and also the new essential tool that developers will need going forward. When these tools are integrated, Microsoft has the opportunity to create the developer platform of the future that can finally deliver a major blow to AWS and boost Azure’s position.
AWS is Playing Catchup by Betting on Other Foundation Models
AWS is now trying to catch up by releasing its CodeWhisperer for general use and announcing new tools for building generative AIs using foundation models (FM) that are not from OpenAI. The FMs in question are:
AI21labs' Jurassic-2 - multilingual LLMs for text generation in Spanish, French, German, Portuguese, Italian, and Dutch.
Anthropic's Claude - an LLM for conversations, question answering, and workflow automation very much similar to ChatGPT.
Stable Diffusion - a direct competitor to OpenAI's DALL-E 2, Stable Diffusion is an extension of the Diffusion Probabilistic Models (DPMs) that generate images by diffusing noise.
Amazon Titan - a text summarization, generation, classification, open-ended Q&A, and search model that is also very much similar to ChatGPT.
All these FMs are currently offered under AWS Bedrock service for building generative AIs that can be used together with AWS Sagemaker which will then build, train and deploy the chosen model. These FMs are currently less developed than OpenAI's offerings, however, we believe that AWS intends to leverage its market presence and drive its clientele to these FMs in hopes of eventually overtaking OpenAI. Meanwhile, AWS is likely going on the hunt for more FMs to add to its collection.
We are at a point in time where the hyperscalers are now fighting to see who can produce the platform or AI of the future for developers, but Microsoft and OpenAI are definitely in the lead. One thing is for sure, developers are the main beneficiary and we can expect more exciting AI or tech startups to pop up in the coming years and a larger war for who can acquire the leading AI startups.
Comments