Hey there, 

Imagine you tell the world that you are starting a not-for-profit world to make sure everyone in the world gets food. You bring the best of talent to work on the problem, you get backing from investors on the same premise. Some of them go out of their way to help you get off the ground.

But, as soon as you have a remarkable lead in solving the problem - you turn to a for-profit (first partially and then fully!) and you start charging the world for supplying food.

If this sounds like fiction, well - OpenAI just pulled it off. It topped this by securing $6.6 Bn in funding and with Mira Murati leaving the organization.

If you thought “Twitter” had a controversial, convoluted history - OpenAI is beating it on every count :) 

If you're looking to use AI for innovation and efficiency in your organization, the Agentic AI Pioneer program helps you develop AI agents that can autonomously perform tasks and make decisions using customized data inputs and algorithms for your specific field.

Now moving forward to this week’s newsletter!

What would be the format? Every week, we will break the newsletter into the following sections:

• The Input - All about recent developments in AI

• The Tools - Interesting finds and launches

• The Algorithm - Resources for learning

• The Output - Our reflection 

Table of Contents

• The Input

  • OpenAI secures $6.6 billion in funding, becoming o …

  • OpenAI DevDay updates

  • Microsoft introduces new features to Copilot AI

  • Pika 1.5 - Physics-defying video effects

  • SB 1047 bill vetoed by Newsom

  • Apple elevates multimodal AI with MM1.5: Advancing …

• The Tools

• The Algorithm

• The Output

The Input

OpenAI secures $6.6 billion in funding, becoming one of the most valuable private companies

OpenAI raised $6.6 billion in a recent funding round, boosting its valuation to a whopping $157 billion! This has potentially made OpenAI one of the most valuable private companies globally.

The funding attracted key investors - Thrive Capital and Khosla Ventures, along with Microsoft. Notably, Nvidia also participated in this round.

This funding coincides with the company restructuring from non-profit to for-profit and the unexpected departure of their long-time Chief Technology Officer, Mira Murati, last week.

Interestingly, Apple was reportedly in discussions to invest in OpenAI but did not participate in the funding. (source)

OpenAI DevDay updates

OpenAI’s DevDay has been a treat for all the developers. Here’s a rundown of the key updates:

Realtime API: A new "Realtime API" enables developers to integrate six unique AI voices into their apps, distinct from those used in ChatGPT. OpenAI does not automatically reveal the use of an AI voice, leaving the disclosure decision to the developers.

Image Fine-Tuning for GPT-4o: Developers can now use images to fine-tune GPT-4o. Providing just 100 sample images can enhance the model’s performance for specific visual tasks.

Prompt Caching: A new prompt caching feature aims to cut costs and reduce latency. By reusing recently processed input tokens, developers benefit from a 50% discount and faster processing times.

Model Distillation: OpenAI introduced "model distillation," allowing smaller models, such as GPT-4o mini, to be optimized using the outputs from larger models. Integrated tools like saved completions and evaluation options support this process. (source)

Microsoft introduces new features to Copilot AI

Following the path of OpenAI, Microsoft has upgraded Copilot with new features introducing voice, vision, and deep thinking - 

• Copilot Voice: Users can now communicate with the AI using natural speech, choosing from four unique voices. Built on fine-tuned versions of OpenAI models, it offers longer time limits for Pro accounts.

• Copilot Daily: Designed to provide a personalized overview of users' daily news and weather, tailored to individual interests. It sources information from authorized content providers, including Reuters and The Financial Times.

• Copilot Vision: An experimental feature for select Pro users, enabling visual understanding of images and web pages. Copilot can assist with decision-making, such as comparing products or suggesting next steps.

• Privacy First: Microsoft emphasizes that no data from Copilot sessions is stored or used for model training, ensuring privacy and security.

• Think Deeper: A feature that tackles complex questions with detailed, step-by-step answers, providing insights into topics like financial decisions, life choices, and more.

• Creative Tools: Features like Visual Search let users upload images for analysis, whether identifying dog breeds or generating recipe ideas. Additionally, Copilot offers writing suggestions for tone and style, sparking creativity with tools for story, poem, and image generation. (source)

Pika 1.5 - Physics-defying video effects

Pika has unveiled its latest Video Model 1.5, introducing "Pika Effects" for users to apply fun, physics-defying effects to images. These effects, like "crush," "squeeze," and "cake-ify," can transform subjects into fantastical versions of themselves, offering more realistic movements in the new model.

Pika 1.5 automatically detects subjects or objects in a video and applies the selected effect, even if the transformation defies reality. (source)

SB 1047 bill vetoed by Newsom

Governor Gavin Newsom vetoed SB 1047, rejecting a bill that would have imposed strict AI regulations. 

The bill sought to make tech companies liable for AI harms and required a "kill switch" for rogue AI systems. 

Although Newsom acknowledged the bill’s intention, he argued it would hinder innovation, especially for smaller AI firms. 

Critics, like Senator Scott Wiener, expressed concerns over the lack of accountability in AI development. Meanwhile, several tech leaders opposed the bill, fearing it would slow industry growth. (source)

Apple elevates multimodal AI with MM1.5: Advancing image understanding and reasoning

Apple's MM1.5 family of multimodal large language models enhances text-rich image understanding, visual grounding, and multi-image reasoning. 

Building on the MM1 architecture, Apple adopts a data-centric approach, using diverse data mixtures like high-quality OCR data and synthetic captions throughout training. 

Models span 1B to 30B parameters, including dense and mixture-of-experts versions, proving that small-scale models can achieve strong performance. 

Specialized variants like MM1.5-Video for video understanding and MM1.5-UI for mobile interfaces offer deeper insights and guidance for future MLLM advancements. (source)

The Tools

Tool: NetJet.io

NetJet is an innovative AI tool designed to create websites quickly and efficiently, allowing businesses to establish their online presence without the need for extensive technical skills.

Problem Statement: I wanted to create a website focused on food recipes. After providing the descriptions, this AI tool successfully generated the website for me. You can try it out for yourself!

How to Access NetJet.io:

1. Log in to NetJet.io.

2. Enter your business name.

3. Select the relevant category for your business.

4. Describe your business offerings.

5. That's it! You can further customize your website to suit your preferences.

With these simple steps, you'll have a professional website ready to showcase your business to the public.

Here’s my trial result: I believe this tool is best for prototyping a website, or if your audience is under 50. However, if you’re looking to reach more than 100+ audiences, there are other tools available that may yield better results. Also, Please note that you need to have a premium subscription to edit the website.

The Algorithm

• If you want to learn how to create AI agents that enhance innovation and efficiency, this Analytics Vidhya course will equip you with the skills to develop agents that autonomously perform tasks and make data-driven decisions.

• If you're interested in data visualization and machine learning integration, this Analytics Vidhya course covers SSM and Mamba for AI applications, taught by Maarten Grootendorst, Senior Clinical Data Scientist at IKNL and creator of KeyBERT and BERTopic.

• The blog "AI Can (Mostly) Outperform Human CEOs" by Hamza Mudassir, Kamal Munir, Shaz Ansari, and Amal Zahra highlights AI's ability to outperform humans in data-driven tasks but stresses that AI is best suited to assist CEOs rather than replace them. It emphasizes the unique strengths of human leaders in vision, ethics, and strategic judgment, and concludes that the future of leadership will involve a partnership between AI and human decision-making. A must-read for those exploring AI's role in business leadership.

The Output

How many of you are feeling that Elon was right about OpenAI since he launched xAI? I obviously don’t know a lot about what has happened at OpenAI or at xAI - but as of now, I do feel most pro-Elon I have ever been!

Exciting, turbulent times - let’s wait for the next version of AI, Elon comes out with to take the war ahead.