AI Today

Botto's 15,000 curators are celebrating a big win this week after six of their carefully-chosen, pixel-pushed masterpieces, sold for more than $350,000 at a Sotheby's auction in New York. It's a story that belongs in a museum. Just when we thought it was safe to come out after NFTs' baffling popularity flare-up of the very early 20s, we're here again. At least Beeple created his own art...

Imagine if you had massive balls - crystal ones - to accurately forecast future business needs.

That's one of the thousands of ways autonomous agents - popularised in organisations of all sizes through Microsoft Copilot Studio - can build better businesses.

These agents can send reports to your senior leadership team identifying inefficiencies or opportunities across the organisation. Then the HR squad can decide whether to upskill colleagues or hire in new ones. All before shit hits the fan!

Autonomous agents will change everything, Taste the future on today's episode of AI Today!

Retrieval augmented generation is how we used to chunk content in huge corpuses of data. Now there's a new sheriff in town - contextual retrieval preprocessing, or contextual RAG. No more relying on keywords; now we're talking hidden relationships between data, which means you can better respond to context inside queries.

We're talking about boosting agent productivity, empowering them to become knowledge ninjas, and transforming customer interactions into personalized experiences. Get ready to unlock growth by maximizing efficiency and creating a customer service dream team!

Let's take a look at how the latest version of Copilot can change the game for your business.

Imagine a manufacturing company developing a new electric vehicle (EV) charging station. This complex process involves multiple steps and teams, using various applications and datasets.

Market research: Copilot can analyse large datasets and generate visualisations to quickly identify key insights by analysing market trends using Excel with Python to forecast demand, pricing, and competitor analysis.

Design and engineering: Copilot can summarise key design discussions from Teams meetings and highlight potential issues, saving engineers time, and facilitating quicker decision-making, by tracking changes and feedback, and summarising discussions, while engineers collaborate on designs using OneDrive and SharePoint.

Sourcing: Copilot compares supplier bids and contracts using OneDrive, ensuring compliance with internal policies. AI surfaces discrepancies and highlights areas needing attention, streamlining negotiation and contract finalisation.

Prototyping and testing: Share test results and feedback across teams using SharePoint and Teams. Copilot can automatically generate reports summarising test data from various sources and identify key performance indicators, helping engineers iterate designs efficiently.

Marketing and sales: Create compelling marketing materials and sales presentations using PowerPoint and Copilot. Copilot can generate presentation drafts based on product specifications, market research, and competitor analysis, ensuring marketing messages are impactful and consistent.

Another major development is Copilot Agents - completing seemingly disconnected tasks forming a single activity.

In the EV charging station development process, a critical step involves collecting and analysing customer feedback during the pilot testing phase.

Traditionally, this process is manual and time-consuming:

Technicians gather feedback from pilot customers through surveys, emails, or phone calls.

This data is collated and manually entered into spreadsheets or databases.

Data analysts then process this information to identify trends, issues, and areas for improvement.

These insights are then shared with engineering, design, and marketing teams.

Copilot Agents can radically transform this process by automating these tasks and unlocking unimaginable possibilities:

Automated feedback collection: A Copilot Agent could be deployed to automatically gather feedback from pilot customers through various channels like in-app surveys, SMS messages, or even voice assistants. This agent could be trained to understand natural language and extract key insights from customer responses.

Real-time data analysis: As the agent collects feedback, it can use Excel with Python to perform real-time data analysis, identifying recurring issues, sentiment trends, and feature requests. This eliminates the need for manual data entry and processing, providing instant insights.

Proactive issue resolution: The agent could be programmed to automatically generate support tickets for reported issues, routing them to the appropriate teams for resolution. It could even suggest solutions based on previous support interactions or knowledge articles, significantly speeding up the resolution process.

Personalised communication: The agent can also be used to communicate with customers proactively, acknowledging their feedback, providing updates on issue resolution, and even offering personalised tips or recommendations based on their usage patterns.

By automating tasks, connecting data, and surfacing insights, Copilot reduces time to market and enhances the final product.

If there's anyone left in the world yet to be convinced AI is changing it, have a chat with Meng To (@mengto on X). He just wrapped up Dreamcut.ai - what he calls his perfect video editor - after spending three months writing 50,000 lines of code with Anthropic's Claude AI.

The application is incredible. As is Meng's generosity - he's giving away the codebase.

Enjoy this deep dive into the amazing story of Dreamcut.ai.

Forget everything you thought you knew about AI assistants. We're not talking simple chatbots that can barely string a sentence together. Claude 3.5 Sonnet, the latest iteration of Anthropic's groundbreaking language model, can actually use computers like a human. It can see the screen, move the mouse, click, type, and even navigate complex software. Think about that for a second. An AI that can control your computer. That can access the vast world of information online, manipulate data, and even write and execute code. It's a power we've only dreamed of, and it's here now.

Engineering teams are frazzled. And we've all been down the Cursor, Aider, Cline, Bolt, and Replit rabbit roles questing for AI coding nirvana. But there are more potholes in the process than a worn-out highway. Enter Solver - self-proclaimed level 4 FSD for coding. Let's take this one for a spin...

Now anyone can know what everyone is doing in real time!

xAI's API gives you access to the 560Gb of data generated every day by the millions of users on X, formerly Twitter.

There are some fantastic opportunities to integrate this unique and ever-changing dataset into your applications.

And you've never before had access at scale to the thoughts and behaviours of your current and future customers.

Let's dive in!

Business is a spider's web of complex processes. And up until now, it's been super hard to get AI working on multi-task processes, without some PhD-level hacking which, unless you're using open source, locally-hosted LLMs, could come unstuck at any time.

Agora solves this. It suggests a rational approach to agentic AI by connecting the bots through a single communication protocol.

In turn, business tasks previously doable only by humans suddenly become perfectly feasible for our silicon siblings.

Super excited about this one. Please enjoy!

Podcast based on this research paper:

A SCALABLE COMMUNICATION PROTOCOL FOR NETWORKS OF LARGE LANGUAGE MODELS

Available at: https://arxiv.org/pdf/2410.11905

Efficiency. Productivity. Growth. That's why we're here. And on today's show, we're hitting all three with the force of a bullet train. 3 AI experts share how they create the best results, in three totally different ways. This is a listen that you can't afford to miss. Enjoy the show!

How product manager Tal Raviv saves a ton of time and generates incredible insights: https://creatoreconomy.so/p/how-experienced-pms-use-chatgpt-tal-raviv

Christopher Penn's advanced AI prompting: https://open.substack.com/pub/almosttimely/p/almost-timely-news-advanced-prompt

No more bad outputs with structured generation, with Remi Louf: https://www.youtube.com/watch?v=aNmfvN6S_n4