By BowTiedRaptor on Substack · 4K+ subscribers · 8 free and 4 paid among the latest 12.
By Data Professionals, for Data Professionals. This is your centralized Website that has all of your data professional needs: We cover: - Money Making Guides -…
In 2023, “prompt engineer” briefly looked like it might become a real profession. Anthropic advertised a Prompt Engineer and Librarian role paying between $175,000 and $335,000 per year . The job involved figuring out the best…
Once an AI workflow works, the next problem is usually making it fast and cheap enough to use regularly. That, is an inference problem . Inference is the process of using a trained model to produce an output. Every time you send…
Most teams working with LLMs eventually run into the same problem, they know what they want the model to do, but they do not have enough examples of it. Maybe you want an internal AI assistant to classify support tickets, write…
A generic model such as GPT or Claude already knows how to write, summarize, classify, reason, and generate code. What it does not know is how your organization wants those tasks performed. It does not automatically know your…
A general purpose AI model can know a great deal about finance, law, programming, accounting, and business. Unfortunately, it still knows almost nothing about how your specific organization works. It does not know your internal…
AI agents are exciting because they promise more than a single response. A good agent can plan, use tools, gather information, execute actions, reflect on results, and keep moving toward a goal. That is what makes agents…
AI agents are one of the most talked-about ideas in modern AI, and for good reason. If chatbots gave people their first glimpse of what LLMs can do, agents are what made many people start imagining truly autonomous software. An…
Foundation models are powerful, but they still make a very human kind of mistake that is: “they answer confidently when they do not have enough information.” Aka, what we call “Hallucination” That is the core problem RAG solves.…
Once an AI application is public, it is no longer interacting only with well-behaved users. It is interacting with curious users, careless users, malicious users, hostile webpages, poisoned emails, weird documents, and tool…
Prompt engineering is the fastest way to improve an AI application. That is why everyone starts there. You do not need to retrain a model. You do not need new GPUs. You do not need to touch the model’s weights at all. You just…
One of the weirdest ideas in modern AI is this: “You can use AI to evaluate AI” At a first glance, that sounds bizarro, maybe even a little stupid. If the model is already unreliable sometimes, why would you trust another model…
Perplexity going down usually feels like you are making progress…and during pretraining, it usually is. But there’s a hidden gotcha here, you can make a model better for users (more helpful, more instruction-following, more…