OpenAI Tasks, Perplexity Data Integrations, Anyone Can Build, and More

AI Wanna Talk

In this episode of "AI Want to Talk," host Jacob Norgord explores various new developments in AI, focusing on product updates and innovative applications from different companies. This episode emphasizes the evolving capabilities of AI and its potential impact on various industries.

AI Product and Feature Updates:

• OpenAI's Task Feature: Discussion of OpenAI's task automation feature, which allows users to create automations that enable AI to perform specific tasks at specific times, such as weekly weather forecasts. This is designed to enhance the value of AI for users with frequent, recurring requests.
• TL Draw Computer: Introduction to TL Draw Computer, an online collaborative platform for brainstorming and creating AI and natural language processing workflows. Users can draw out workflows with components like text boxes, images, and audio clips to create complex automations.
• Google's NotebookLM UI: Overview of the new UI for Google's NotebookLM, featuring a layout with sources on the left and document generation tools on the right, including options for creating podcasts, study guides, FAQs, and timelines. The new interface also provides a central pane for source summaries and question prompts.

AI Model and Integration Developments:

• Perplexity's Acquisition of Carbon: Details on Perplexity's acquisition of Carbon, a retrieval engine that connects external data sources to large language models. This will allow Perplexity to integrate data from apps like Notion and Google Docs, creating a centralized hub for user information.
• Eleven Labs Flash Model: Announcement of Eleven Labs' new Flash model, which generates speech in 75 milliseconds, enabling more human-like interactions with AI voice models. This model aims for low latency and is being targeted for integration into various products, such as video games.
• ChatGPT Integration: Discussion of ChatGPT's new feature allowing it to work directly with apps like Apple Notes and Notion, accessing all data within these applications rather than just screen displays. This feature positions ChatGPT as a central interface for interacting with data across various applications.

AI Agents and Data:

• Firecrawl and AI Agents: Introduction to Firecrawl, a company focused on providing AI models with high-quality data by scraping the web for specific data sets. They are also hiring AI agents (not humans) to work within their system and are paying $10,000 to $15,000 for the use of these agents.
• Vertical AI Agents: Explanation of vertical AI agents, which are specialized AI systems designed for industry-specific tasks. Examples include agents for finance or law that can take action within their respective fields.
• AI Software Creation: Discussion of platforms like Windsurf, Bolt, and Cursor, which allow users without coding experience to create software using AI.
• Tempo Labs: Introduction to Tempo Labs, a code-first alternative to Figma, powered by AI. This platform generates functional code by prototyping user interfaces and allowing users to focus on core ideas rather than code.

AI and Human Cognition:

• Human Brain Processing Speed: Exploration of an article highlighting that the human brain processes information at a rate of only 10 bits per second. The podcast discusses how, despite this slow rate, humans are able to distill vast amounts of data efficiently.
• AI Constraints: Speculation on the idea that mimicking human constraints in AI data processing may be key to achieving more nuanced and contextual understanding in AI.

Other Interesting Points:

• Fake AI Band: A story about a person who created a fake band with AI and made $10 million by also creating fake AI fans. This is framed as a humorous example of how AI can be used in unexpected and potentially fraudulent ways.
• The focus on AI Agents: Going into 2025, there