The rapid advancements in research, software libraries, and applications following OpenAI’s ChatGPT launch have confused many professionals. Amidst the hype, this article aims to clarify the opportunities available.
Microsoft now integrates GPT into Bing and Microsoft Office, prompting other tech giants like Google and Meta to develop their own LLMs. The open-source community is also actively creating new tools and research for LLMs.
These developments are truly exciting and phenomenal. The advancements in LLMs open up new possibilities for AI applications and have the potential to revolutionize how we interact with technology. As LLMs become more accessible and integrated, they will play an increasingly significant role in our daily lives.
Many professionals feel overwhelmed and disoriented by the excessive online materials. I am one of them. As a response, online courses on using GPT for business and application development have emerged to offer guidance.
This article focuses on the opportunities for application developers using LLMs like GPT, which span across various skill sets from software engineering to ML engineering. As, like me, everyone wants something out of this hype, I hope my perspective is helpful for developers to see the opportunities clearly.
Levels of activities involving LLM.
Level 1: Using AI Power Tools in Development Workflow
Leverage AI-powered tools, such as GPT-based plugins and APIs, to accelerate development work and enhance productivity. These tools can streamline the development process, discover and learn new code patterns, better code quality, and create innovative solutions.
Engineers who can effectively use tools like CodeGPT and Github Copilot can make efficient development work. These tools like ChatGPT provide better code suggestions, automate tasks, and generate code snippets based on the prompts. Understanding AI capabilities and crafting precise prompts can give 10x productivity boost. The opportunity is to master these AI tools and integrate them into the development workflow, create innovative and efficient applications.
Level 2: Integrating AI Functions into Application Projects
Integrate GPT models and other AI functions via APIs into applications to enable advanced features like recommendations, summarizations, and natural language processing, thereby increasing the application’s value and usability for end-users.
Small businesses use AI-powered APIs like OpenAI’s Chat API and create and sell new SaaS tools with unique features. AI-driven applications can provide personalized recommendations, generate summaries, and automate content creation. Subscription-based solutions have emerged, showcasing AI integration potential in application projects. Tools like docGPT, and siteGPT show how AI functions can offer competitive advantages in applications, opening opportunities for developers to create innovative and marketable solutions using AI capabilities. xGPT FTW!
Level 3: Fine-tuning AI Models and Embedding and Querying Custom Datasets
Fine-tune AI models for specific use cases, industries, or applications, and build GPT indices for custom datasets to create tailored solutions that offer a competitive edge in a crowded market.
Fine-tuning AI models using OpenAI’s GPT, Google’s PaLM, or opensource GPT-J and embedding custom datasets create specialized AI tools for popular ERPs, CRMs, and CMSs, targeting SMEs and various sectors. Developers create solutions that cater to niche markets and specific industry needs, offering unparalleled value to end-users. The opportunity lies in identifying industry gaps and leveraging fine-tuned AI models to deliver tailored solutions addressing the needs for internal customers or external.
Many libraries such as llama-index, LangChain and Semantic Kernel from Microsoft is making things easy for developers to integrate GPT solutions on existing data.
Level 4: Training Custom ML Models for Specific Purposes
Develop custom GPT models using transformer models and proprietary datasets for business applications, enabling highly tailored solutions that cater to specific industry requirements and offer a competitive advantage in the market.
Large enterprises create private GPT models for proprietary datasets, developing specialized AI applications for unique business needs. E.g Bloomberge GPT for This approach requires both ML engineering and application development skills. Developers create AI solutions that leverage internal data for actionable insights, automate workflows, and enable data-driven decision-making. The opportunity lies in harnessing custom-trained AI models and proprietary data to create innovative solutions, taking competitive advantages.
Level 5: Developing New Training Tools and Models
Work on creating new training tools, models, and algorithms for AI applications as an ML engineer. Research focuses on making models run on low-resource machines using techniques like quantization and LoRA, for example llama.cpp and xTuring etc.
AI labs and open-source communities like BigScience Project and BLOOM develop new models and tools. They work on AI models that run on low-resource machines, like Raspberry PI, using techniques such as quantization and LoRA. This active research area creates opportunities for next advancement with examples like llama.cpp and xTuring. Developers can contribute to and benefit from these cutting-edge projects to stay at the forefront of the AI landscape.