A statement is only meaningful to the degree to which the audience is not free to overload it with arbitrary interpretations and can arrive at a shared interpretation of whay has been said. “Situated in life” is so unprecise that it does not capture much meaning
PROMPT ENGINEERING
-
AI Guardrails Dilemma: Balancing Safety and User Reassurance
By
–
I know this is kind of goofy but it does illustrate how hard guardrails are when users can ask AI anything at all Do you reassure someone who seems genuinely worried? Play along? Take it seriously? Refuse to answer for fear that this is part of a jailbreak for a hacking attempt?
-
AI Speech Writing: When Everyone Uses Generative AI for Public Speaking
By
–
A friend mentioned to me that he was just at a school event and there are no longer any bad public speeches, since everyone clearly used AI to punch up their talks. Everyone should just stand up and give their prompt "Funny, heartwarming graduation speech with inspiring quote."
-
Green Screen Background Feature Simplifies Image Generation Process
By
–
Really cool not to have to try to prompt for a simple background and then photoshop it out.
— Rob Lennon 🗯 | AI Whisperer (@thatroblennon) 15 mars 2024
When you tell your prompt stuff "green screen background" it can heavily influence the output.
Love this feature https://t.co/m4XJoRglEiReally cool not to have to try to prompt for a simple background and then photoshop it out. When you tell your prompt stuff "green screen background" it can heavily influence the output. Love this feature
-
Devin AI Coder: New Modality of Human-AI Interaction
By
–
The thing that makes Devin "the AI coder" really interesting is the new modality of interaction. You can "talk" to it at any time, like you would a person, and it just keeps chugging away in the background executing and debugging your ideas.
— Ethan Mollick (@emollick) 15 mars 2024
Here is a time lapse of my project. pic.twitter.com/nMLesgzFfuThe thing that makes Devin "the AI coder" really interesting is the new modality of interaction. You can "talk" to it at any time, like you would a person, and it just keeps chugging away in the background executing and debugging your ideas. Here is a time lapse of my project.
-
Comparison of Claude and ChatGPT for generating logos
By
–
Very curious what Claude is doing differently to achieve this. For comparison, ChatGPT 4 with the same prompt couldn't produce any (non-trivial) logos containing my requested text in 15 attempts — a few representative examples of the failures:
-
Claude 3 Opus: Demonstrating ASCII art logo rendering with common but retryable errors
By
–
Demo: Claude 3 Opus renders ASCII art logos LLMs are generally awful at this. Claude 3 is the best yet; errors are common, but with retries you can get legible text. Note how similar the errors are to diffusion models — e.g. bad spelling/glyphs but consistent drop shadows
-
Romain Huet Requests Model Feedback and Prompts
By
–
Hey Tomaz, we’d love to hear your model feedback, especially if you have specific prompts or evals. My DMs are open, thanks!
-
Command-R Tool Use Mode Now Includes Citations and Multi-Hop
By
–
As an added bonus, Command-R and Tool Use mode comes with citations out-of-the-box, allowing users to trace back the actions the model has taken on their behalf. We're just scratching the surface! Keep an eye out for multi-hop Tool Use coming soon. Join the beta program!
-
Command-R Demonstrates Strong Multi-Turn Tool Use Performance
By
–
Tool Use with Command-R shows strong performance when evaluated against real-world tool use capabilities in a multi-turn conversational setting.