Prompt Engineering

Prompt Engineering explores effective communication with AI, focusing on techniques like prompt crafting and 'chain of thought' prompting. It covers advancements in AI models like LLaMA, Orca, and ChatGPT, discusses anthropomorphizing AI, and addresses concepts like the black box problem and AI's societal impact.

AI Communication Techniques AI Model Advancements Prompt Crafting Strategies Societal Impact of AI AI in Image Generation AI Role Assignment Speech Recognition

Top posts of the year

And their main takeaways
216 implied HN points 29 Apr 23
  1. Effective communication with AI models depends on providing quality prompts.
  2. When interacting with AI, avoid asking it to rephrase or rewrite text directly; instead, focus on asking for correctness and improvements.
  3. Maintaining your unique writing style when engaging with AI is important to preserve your voice in the text.
196 implied HN points 05 May 23
  1. The biggest deal in AI is the open-source model LLaMA, not ChatGPT.
  2. ChatGPT was impressive but had weaknesses like generating nonsense and being easily fooled.
  3. The rapid innovation cycle after the leak of LLaMA weights led to significant advancements in AI models.
Get a weekly roundup of the best Substack posts, by hacker news affinity:
39 implied HN points 04 May 23
  1. Breaking down problems into a series of steps is powerful.
  2. Using large language models like ChatGPT can help with the 'chain of thought' prompting technique.
  3. Understanding this technique is crucial for maximizing the benefits of these models.
39 implied HN points 22 May 23
  1. AI is rapidly advancing, especially in the medical field.
  2. New technology like ImageBind can link different types of data with images as a common basis.
  3. Fine-tuning language models with a small number of prompts can significantly improve performance.
39 implied HN points 10 Jul 23
  1. Our understanding of how AI works is limited due to the black box problem.
  2. Engineers build based on understanding, while scientists dissect to understand.
  3. It's more important for AI to work as intended and be continuously tested than to fully understand how it works.
19 implied HN points 28 May 23
  1. ChatGPT conversations are now shareable to prevent screenshot sharing and misinformation.
  2. Tree-of-thoughts prompting is a new approach where LLM is prompted with multiple initial steps and evaluates each one.
  3. A new highly performant open-source model called Guanaco outperforms previous models and was fine-tuned using a new approach named QLoRA.
19 implied HN points 30 May 23
  1. Large language models perform better when given a specific role in conversations
  2. Assigning roles to language models can lead to more relevant and engaging responses
  3. Providing clarity on the intended role of a language model is a powerful way to enhance its performance
1 HN point 11 May 23
  1. AI advancements are fast and significant, leading to uncertainties about the future models of society.
  2. AI may result in mass unemployment, but historically, technology revolutions have not led to catastrophic unemployment levels.
  3. Challenges with AI include the misuse of data leading to bias, the potential for AI to outsmart humans, and the widening class divide due to unequal access to AI tools.
0 implied HN points 05 Jul 23
  1. OpenAI's Whisper model is a powerful tool for audio to text transcription, trained on 680,000 hours of data.
  2. Voice interfaces are often tied to specific software, but a general-purpose voice transcriber like Whisper could be very useful.
  3. Whisper can be integrated with tools like ChatGPT for recording and transcribing text to work on creating a stronger narrative.
0 implied HN points 14 Jun 23
  1. OpenAI has updated the GPT-3.5 model with a larger context length of 16,384 tokens, allowing for more data input and better results.
  2. There is a 75% cost reduction on OpenAI's models, and a 25% cost reduction on input tokens for gpt-3.5-turbo.
  3. OpenAI announced functions that allow the model to make API calls, write emails, query databases, and more, taking LLM applications to the next level.
0 implied HN points 28 Apr 23
  1. New content is coming soon on engineeringprompts.substack.com.
  2. The post includes an image and a link to Marcel Salathé's profile.
  3. The post encourages sharing and subscribing for updates.