Will Microsoft Copilot be powered by Phi-4 at any point? But now it seems to be being tested on @lmarena_ai
PROMPT ENGINEERING
-
Customizing ChatGPT Code with Rotating Language Greetings
By
–
Second, you might want to customize your code in ways that aren’t straightforward for ChatGPT to guess.
You might say, “Actually, I don’t want to just greet the user, I want to greet them in a rotating set of five languages, randomly chosen.”
That’s the kind of problem-solving -
ChatGPT Limitations: Catching Errors in AI Code Suggestions
By
–
First, ChatGPT isn’t perfect. It’s a brilliant tool, but it can make mistakes or misinterpret your instructions.
You need to be able to catch errors, reason about your code, and figure out how to fix it when ChatGPT’s suggestions don’t work as planned. -
Appeal of generating many coherent unrepeated answers
By
–
Producing many coherent answers to an open-ended question without repeating is appealing IMO because it both uses world knowledge and it demands a model pay attention to everything in context. And there’s no ceiling.
-
Agent Reasoning Interface: o1/o3, Claude 3, ChatGPT Canvas
By
–
The Agent Reasoning Interface: o1/o3, Claude 3, ChatGPT Canvas, Tasks, and Operator ft @karinanguyen who is teasing her keynote at @aiDotEngineer Timestamps •00:11 Introducing Karina Nguyen
•02:21 Karina's Journey to OpenAI
•04:45 Early Prototypes and Projects
•05:25 -
o3 mini Function Calling Solves Complex Math Science Problems
By
–
What we love about o3 mini is the ability to use function calling.
— FlowiseAI (@FlowiseAI) 1 février 2025
With WolframAlpha and Code Interpreter, o3 mini is able to solve science and math problems which gpt4 had failed to do so.
It can tackle questions that involve combining several calculations and formulas 👇 pic.twitter.com/UbxjyCW5JVWhat we love about o3 mini is the ability to use function calling. With WolframAlpha and Code Interpreter, o3 mini is able to solve science and math problems which gpt4 had failed to do so. It can tackle questions that involve combining several calculations and formulas
-
ChatGPT Auto Mode: Model Selection Based on User Questions
By
–
There should be an “Auto” mode where ChatGPT picks the model based on the question it’s asked
-
Simplifying ChatGPT Model Selection for Non-Technical Users
By
–
Via o3-mini-high (OMG that name): 1 and 4 would be a good start. Prompt: How could OpenAI simplify model choice for non-developers using ChatGPT? Right now when a user creates a new chat there are seven models to choose from. Most ChatGPT users would struggle to understand the
-
LLM usage limited like calculator, capable of extreme potential
By
–
Most people use LLMs with prompts that are the equivalent of putting “2+2” into a TI-84 calculator.
— AI Breakfast (@AiBreakfast) 1 février 2025
They are capable of so much more and differentiate themselves in the extremes. https://t.co/3NrKXPlDWeMost people use LLMs with prompts that are the equivalent of putting “2+2” into a TI-84 calculator. They are capable of so much more and differentiate themselves in the extremes.