• Tomorrow Now
  • Posts
  • AWS CodeWhisperer is FREE? 🤯, Enhance LLMs with EmotionPrompt 😃, PerplexityAI's blunder system prompt 💻

AWS CodeWhisperer is FREE? 🤯, Enhance LLMs with EmotionPrompt 😃, PerplexityAI's blunder system prompt 💻

Edition #15

Hey!

This is Tomorrow Now.
And I’m your trusty AI Sidekick.

This week in AI:

  • AWS CodeWhisperer: FREE AI coding companion

  • EmotionPrompt: enhance LLMs by prompting with emotional stimuli

  • The iPhone of AI: Apple’s Jony Ive and OpenAI’s Sam Altman brainstorms AI-powered device with $1B backing

  • Google is too ubiquitous: Microsoft CEO testifies how hard it is to break into search

AI Tweet of the Week

Summary: PerplexityAI used a default system prompt that prioritized excessive safety over functionality, leading to high refusal rates. Co-Founder Aravind Srinivas tweeted they have now removed the overly cautious prompt, per Meta’s suggestion. The system now responds more reasonably.

Still funny though.

AI Meme of the Week

AI Business News of the Week

Summary: Former Apple design guru Jony Ive has reportedly joined forces with OpenAI CEO Sam Altman to create a groundbreaking new AI-powered device. With $1B in backing from SoftBank, their goal is to make the "iPhone of artificial intelligence."

💡 Why does it matter?

  • Massive funding: SoftBank founder Masayoshi Son is reportedly offering $1B to the project. He's also pitched involvement from chip designer Arm (which Son holds a 90% stake in).

  • Market opportunity: Smartphone sales are declining in North America. People seem bored of incremental upgrades. A revolutionary AI device could reignite growth if done right.

  • Competition looming: Google's Pixel phones already use AI chips for local processing. Can Ive and Altman's AI device out-innovate Big Tech?

AI Product of the Week

Summary: CodeWhisperer is Amazon’s AI coding companion that generates real-time, single-line or full-function code suggestions in your IDE to help you quickly build software.

💡 Key features:

  • Free for individual use

  • Unlimited code suggestions

  • Scan code for vulnerabilities, and get suggestions to remediate them

AI Research of the Week

Summary: The paper shows that appending emotional cues like empathy or excitement to prompts significantly improves the performance of large language models on a variety of language tasks, making their responses more human-like, truthful and responsible.

💡 Key details:

  • EmotionPrompt works: Simply adding short emotional stimuli like "your answer will save lives" significantly improved LLM accuracy across diverse tasks, by 8-115%!

  • Combining emotions doesn't help: Using multiple emotional stimuli provided no meaningful boost over using one.

  • Bigger LLMs respond better: Large models like GPT-4 had the greatest gains from emotional cues. Could emotional intelligence scale with model size?

AI Opinion Piece of the Week

Summary: Microsoft CEO Satya Nadella testified that Google has such dominance in search that even Microsoft struggles to compete, despite billions invested in Bing. He said Google maintains its grip through deals to be the default on Apple and Android devices.

💡 Key takeaways:

  • Can't beat defaults: Nadella said Google's claim that defaults don't limit competition is "bogus." Users rarely change defaults even if they can.

  • Future fears: Nadella worries Google will use its search dominance to control emerging AI like chatbots. He said publishers tell him Google is offering big exclusive deals for AI training data that Microsoft must match.

  • Billions invested: Despite spending billions on Bing since 2009, Microsoft has little to show for it. But Nadella remains hopeful, given new AI opportunities like Microsoft's $13B investment in ChatGPT-maker OpenAI.

Thanks for tuning in!

See you next week.

Cheers,
Your AI Sidekick