Using Negative Instructions and Self-Generated Examples in Language Models

3min Snip

00:00

Play full episode

Summary

Transcript

Episode notes

Negative instructions may not work effectively in language models, as they can introduce the concept into context and make it difficult to negate. However, explicitly stating what not to do in detail can work well. Language models like GPT respond decently well to negative instructions. Moreover, having the model generate its own examples can be effective, as it can recall its own canonical examples and move on to solving the problem. Using self-generated examples and chain of thought rationales can be useful, but human-written examples may still be better for accuracy.

In this episode, Nathan sits down with Sander Schulhoff, Cofounder and CEO of Learnprompting.org. They discuss the business model, the keys to prompting that every user of language models should know, negative prompting, prompt hacking, and more. If you need an ecommerce platform, check out our sponsor Shopify: https://shopify.com/cognitive for a $1/month trial period.

LINKS:

- Learnprompting.org: https://learnprompting.org/

- Learnprompting.org Prompt Hacking: https://learnprompting.org/docs/category/-prompt-hacking

- Ignore This Title and Hackaprompt: https://arxiv.org/abs/2311.16119

SPONSORS:

The Brave search API can be used to assemble a data set to train your AI models and help with retrieval augmentation at the time of inference. All while remaining affordable with developer first pricing, integrating the Brave search API into your workflow translates to more ethical data sourcing and more human representative data sets. Try the Brave search API for free for up to 2000 queries per month at https://brave.com/api

Shopify is the global commerce platform that helps you sell at every stage of your business. Shopify powers 10% of ALL eCommerce in the US. And Shopify's the global force behind Allbirds, Rothy's, and Brooklinen, and 1,000,000s of other entrepreneurs across 175 countries.From their all-in-one e-commerce platform, to their in-person POS system – wherever and whatever you're selling, Shopify's got you covered. With free Shopify Magic, sell more with less effort by whipping up captivating content that converts – from blog posts to product descriptions using AI. Sign up for $1/month trial period: https://shopify.com/cognitive

Omneky is an omnichannel creative generation platform that lets you launch hundreds of thousands of ad iterations that actually work customized across all platforms, with a click of a button. Omneky combines generative AI and real-time advertising data. Mention "Cog Rev" for 10% off www.omneky.com

NetSuite has 25 years of providing financial software for all your business needs. More than 36,000 businesses have already upgraded to NetSuite by Oracle, gaining visibility and control over their financials, inventory, HR, eCommerce, and more. If you're looking for an ERP platform ✅ head to NetSuite: http://netsuite.com/cognitive and download your own customized KPI checklist.

X/SOCIALS:

@labenz (Nathan)

@SanderSchulhoff

@learnprompting

@CogRev_Podcast

TIMESTAMPS:

(04:50) What is Learnprompting.org

(06:30) Learnprompting.org's adoption stats and the transition from open source to a business

(10:21) Are we done with prompt engineering or is there more to be discovered?

The key 2-3 things every user of GPT should know

(19:11) The format trick

(21:32) Role casting / persona prompting

(24:44) Does the level of vocabulary you bring to a LM impact its performance

(26:10) Contrastive chain of thought

(28:30) Language models responding well to negative instructions

(32:29) Benchmarking techniques

(34:00) Answer engineering

(37:29) Debugging a prompt

(41:13) Sander's favourite models today

Best practices for image prompting - maybe cut out

(48:50) Productivity improvement with language models

(51:25) Tips for coding with prompts

(56:06) The current state of prompt engineering and how it'll evolve

(58:30) 2024 will be the year of agents and Sander's favourite agents

(59:57) How will agents impact the future of prompt engineering

(1:04:10) How to start agent engineering

(Adversarial attacks

(1:19:40) Model to model hijacking

(1:25:24) Chinese character attack

(1:28:33) Taxonomy of attacks

This show is produced by Turpentine: a network of podcasts, newsletters, and more, covering technology, business, and culture — all from the perspective of industry insiders and experts. We’re launching new shows every week, and we’re looking for industry-leading sponsors — if you think that might be you and your company, email us at erik@turpentine.co.