Chain-of-thought can also be interpreted as a simple external memory tool: The model writes to a scratchpad, then reads it, and answers the question. The scratchpads to which the model outputs text and inputs text are powerful tools that need further exploration.
PROMPT ENGINEERING
-
Chatbot Creates Video Code Executing AI Instructions
By
–
Pensad que ese vídeo sale del código ejecutado creado por un chatbot bajo mis instrucciones. Es una locura de herramienta, en sus aciertos y sus errores. Bendita IA.
-
Prompt Engineering Techniques for Automated Code Generation
By
–
Nada, olvidad lo anterior. Con un poco de ingeniería de prompt le tienes picando código fácil.
-
AI Model Limitations Code Generation Prompt Injection
By
–
Dos cosas interesantes: – Aunque puede, no va a generar código de programación. Sospecho que es una de las funcionalidades que OpenAI no ha cedido y que servirán para diferencia a su ChatGPT Plus. – Parece robusto a inyecciones de prompt. No voy a forzar para no ser baneado.
-
Detecting LLM Hallucinations Without Source References
By
–
En este prueba le he pedido un párrafo del paper de PaLM y lo ha alucinado. Toda esa frase de "I am a museum in Paris, France. I have a glass…" no se encuentra en el paper. Lo interesante es que cuando no hay referenciada una fuente, es fácil de detectar que está alucinando.
-
Intermediate representation in music generation from text prompts
By
–
"Intermediate representation" refers to a representation of the music generated from the text prompt. The generator model first turns the text prompt into this intermediate representation, which captures elements of the music such as: -Genre
-Tempo
-Instruments
-Mood
-Era -
Future of music creation via text prompts using diffusion models
By
–
In the not-too-distant future, anyone will be able to create any type of music they want via text prompts.
— AI Breakfast (@AiBreakfast) 9 février 2023
Here's a look into recent research on diffusion models for generating high quality music audio from text prompts:
(more examples below) pic.twitter.com/ywgJRVoxLxIn the not-too-distant future, anyone will be able to create any type of music they want via text prompts. Here's a look into recent research on diffusion models for generating high quality music audio from text prompts: (more examples below)
-
Flan-T5 Public Release with Instruction Tuning Data Ablation
By
–
(2 cont.) Flan-t5 is publicly available along with the dataset (thanks to @ShayneRedford
), so you can even ablate parts of the instruction tuning data to see what unlocks performance. -
Bing’s real star is unnamed next-gen OpenAI model, not Prometheus
By
–
I suspect the real star of Bing Search is the strategically unnamed "next-generation OpenAI model." Prometheus seems to be the prompt + search augmentation. The underlying model is tuned to be a white-label ChatGPT for search applications generically.
-
Contrast between Bing’s detailed prompt and ChatGPT’s simple system prompt
By
–
It’s remarkable how much is in this prompt. Contrast it with ChatGPT’s, which just sets a few parameters. This model doesn’t seem to be FT’ed for Bing — they had to specify in the prompt Sydney is not “an assistant.”