Amy Stapleton

@Cisco announced the launch of their AI Defense solution today. Definitely worth a look for companies starting to deploy LLM-based applications, especially customer-facing. Interesting for #CCaaS providers and users as well. opusresearch.net/2025/01/15/cis…

How does Satya Nadella’s recent prognostication on the demise of SaaS impact CCaaS software? Are agentic systems a real threat? Some thoughts:

Our "Riddle of the Beginning" debate (with the great Scott Aaronson and the eminent Pria Natarayan) is online. (I remember I was a bit flummoxed by how some of my interlocutors seem actually no longer curious about this question...) iai.tv/video/the-ridd…

Ilya Sutskever gave a full talk on "Sequence to sequence learning with neural networks" It's one of the most important talks in 2024 Here are some of my favourite insights from the talk 🧵

Cohere just dropped Command R7B The smallest, fastest, state-of-the-art enterprise-grade LLM. The best part? It’s Open Weights! Here’s everything you need to know 🧵 (Cohere Partner)

@Vapi_AI’s latest fund raise shows the market for voice agent tech continues to heat up.

@Vapi_AI Congrats!

We've raised $20M from Bessemer, joined by Abstract, AI Grant, Y Combinator, Saga Ventures, and Michael Ovitz. Vapi is the platform for bringing voice AI agents to the world at scale. Today, we’re handling millions of calls per month for customers worldwide. Voice is the most natural form of human expression, and we want to use it to put humans back at the foreground of technology. Join us!

Link to full blog post @opusresearch opusresearch.net/2024/12/12/ama…

Amazon's new Nova models give #CCaaS providers another option as they look to right-size models for optimal performance and cost. More in the comments.

Having spent much of my early career crafting Batch Input Processing for controlling SAP systems, this seems like magic!

Starting at about minute 12:50 in this demo by Amazon of a Nova-powered agent updating an order in SAP… youtu.be/KEzL6VywPBQ?fe…

It’s amazing to me how you can now “ask” an AI agent to make updates in your ERP system and have those updates executed. Take a look at the demo here:

I read Google's paper about their quantum computer so you don't have to. They claim to have ran a quantum computation in 5 minutes that would take a normal computer 10^25 years. But what was that computation? Does it live up to the hype? I will break it down.🧵

2/ Some thoughts on how Amazon Q Business now further supports automating workflows… opusresearch.net/2024/12/04/ama…

1/ Employees are taking advantage of GenAI to ask for business insights in plain language. Amazon Q Business is an example enabler of this capability. Yesterday Amazon announced updates to Amazon Q.

Seems positive that they’re focused on offering lower cost alternatives.

Here's the paper: arxiv.org/pdf/2411.18279 So my thought is: if LLMs can do this, why do we even need GUIs anymore?

@rohanpaul_ai I was thinking about the article you posted on how LLMs can navigate GUIs to carry out tasks for a person, all based just on a person's instructions. 1 / 2

@rohanpaul_ai Fascinating

Treating LLMs as actors unlocks their true potential in complex reasoning tasks. Acting-based prompting outperforms traditional reasoning approaches for LLMs 🎯 Original Problem: LLMs struggle with complex reasoning tasks, particularly word puzzles like NYT Connections, where traditional prompting methods achieve low success rates. ----- 🔧 Solution in this Paper: → The paper introduces "Method Actors" - a mental model treating LLMs as actors performing roles rather than thinking machines → Prompts function as scripts and stage directions, while responses are viewed as performances → The approach breaks down complex tasks into smaller, imitable performances → It compensates for LLM limitations through system design and validation checks → For the NYT Connections puzzle, it uses templates based on past puzzle patterns and multi-stage processing ----- 💡 Key Insights: → LLMs perform better when imitating rather than reasoning → Complex tasks need decomposition until imitation matches authentic results → Dramatic scene-setting in prompts increases context window usage → External validation helps filter out hallucinations → System architecture should compensate for inherent LLM weaknesses ----- 📊 Results: → Basic GPT-4 solved 27% puzzles, Chain-of-Thought 41% → Method Actor approach achieved 86% success rate → With GPT-4-preview, Method Actor reached 87% perfect solutions → Surpassed human expert performance in puzzle-solving accuracy