• AI Emergence
  • Posts
  • Sam Altman Reveals OpenAI’s Plan for GPT-4.5 & GPT-5

Sam Altman Reveals OpenAI’s Plan for GPT-4.5 & GPT-5

Along with - EU Launches InvestAI to Mobilize €200 Billion for AI Development

Hi,

I don’t know about you - but every time I feel I understand Sam Altman, the strategist, he throws a new punch out of the bag. It feels like he has a way to ensure that he leads the world in the direction he wants to, in spite of the challenges thrown at him.

You would have thought Elon’s offer bid would slow OpenAI down - Sam came back with an offer for Twitter. He adds to it by saying “He feels for Elon and his insecurities!”

Next, to ensure this doesn’t blow out - he diverts the world’s attention to GPT 5! 👏👏👏

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

Right now, you get the option to select 7 types of models in ChatGPT.

Since launching ChatGPT-3, OpenAI has kept adding more options - but at some point, it got confusing. With different model names and intelligence levels, users are left figuring out which one actually fits their needs.

Now, Sam Altman has confirmed that OpenAI is simplifying everything with GPT-4.5 and GPT-5, finally bringing back a unified AI experience that blends intelligence, reasoning, and multimodal capabilities into one seamless system.

What’s Changing?

  • GPT-4.5 (Orion) Arriving Soon- This will be OpenAI’s last non-chain-of-thought model, expected in the coming weeks, setting the stage for a more integrated AI system.

  • GPT-5 Will Merge Everything- Instead of picking between GPT-series and o-series models, OpenAI will combine them into one AI that adapts dynamically to different tasks. No more model picker, no more confusion.

  • o3 Will No Longer Be Standalone- The o-series models will be directly integrated into GPT-5, meaning ChatGPT and API users will get all capabilities in one place.

  • Deep Research & Multimodal Power- GPT-5 won’t just generate text—it will search the web, analyze documents, understand images, and engage in voice conversations, all in a single interface.

What This Means for You?

Tiered AI Intelligence- Paid users will unlock higher intelligence levels:

  • Plus Users - A more advanced version of GPT-5.

  • Pro Users - An even higher level of intelligence, sparking speculation about just how capable it will be.

Deep Research for Everyone - Previously limited to Pro users, Deep Research (which allows GPT-5 to search, analyze, and cite sources) will now be available at lower tiers:

  • Plus users: 10 searches/month (scaling up over time).

  • Free users: 2 searches/month.

  • Pay-as-you-go option coming soon.

GPT-5 is shaping up to be the closest thing yet to an AI that “just works”- and we’re here for it. (source)

Following the US announcement of Project Stargate, a $500 billion AI infrastructure initiative, the EU has launched InvestAI, a €200 billion initiative to boost AI investment, including €20 billion for AI gigafactories. Announced by Commission President Ursula von der Leyen at the AI Action Summit in Paris, the initiative aims to position Europe as a leader in AI innovation through open and collaborative development.

Key Highlights:

  • Four AI Gigafactories- Large-scale infrastructure equipped with 100,000 next-gen AI chips, specializing in training complex AI models for industries like medicine and science.

  • Public-Private Partnership- The largest global AI collaboration, ensuring startups and companies, not just big tech, can access supercomputing power.

  • Funding Strategy- InvestAI will combine EU budget contributions with private sector investment, leveraging funds from Horizon Europe, Digital Europe, and InvestEU.

  • AI Innovation & Talent Development- Complementary efforts include AI research funding, education initiatives, and Common European Data Spaces for better AI training.

With €10 billion already committed to AI factories, InvestAI expands Europe's investment leadership, unlocking further private capital and accelerating AI-driven industrial transformation. (source)

OpenAI is moving fast to reduce reliance on Nvidia, finalizing the design for its first custom AI chip in the coming months. The chip will be manufactured by TSMC with mass production targeted for 2026.

Why It Matters

  • Breaking Nvidia’s Grip- OpenAI joins Microsoft and Meta in exploring alternatives to Nvidia’s 80% market-dominating chips.

  • Cost & Supply Chain Control- In-house chips could lower costs and secure AI hardware as demand skyrockets.

  • Small but Strategic Start- The chip will initially be used on a limited scale, focusing on running AI models rather than full-scale training.

The Bigger Picture

  • AI chip development is costly- up to $500M per version- and OpenAI’s 40-person team is much smaller than Google or Amazon’s efforts.

  • Meta plans to spend $60B on AI infrastructure in 2025, while Microsoft is committing $80B- raising competition for next-gen AI hardware.

  • OpenAI’s participation in the $500B Stargate project signals long-term investment in AI chip infrastructure.

If successful, OpenAI’s custom chip could provide leverage in chip negotiations and pave the way for more advanced processors in the future. (source)

The Musk vs. Altman drama just hit a new level.

Elon Musk has put in a $97.6 billion bid to take over OpenAI, backed by xAI and a group of investors. He claims he wants to bring OpenAI back to its open-source roots, but let’s be real- if this goes through, we might be looking at an xAI + OpenAI merger.

What’s Going On?

  • Musk’s team has officially made an offer to buy OpenAI.

  • OpenAI’s board now has to decide whether to accept or wait for a counteroffer.

  • OpenAI is racing to raise $40B by 2026 to go fully for-profit- Musk’s bid just threw a wrench in those plans.

  • Altman fired back on X: "No thank you, but we’ll buy Twitter for $9.74B if you want.”

Musk pulling takeover moves? Sounds familiar. This whole thing has serious PayPal-X.com merger vibes, where he was pushed out before the company took off. Could history repeat itself? (source)

This year’s Super Bowl wasn’t just about touchdowns and halftime drama- it was an AI showcase. Major tech giants and startups splashed millions on ad spots, proving AI is now a mainstream marketing weapon.

Biggest AI Moments from the Super Bowl Ads

OpenAI’s Super Bowl Debut

OpenAI aired its first-ever Super Bowl ad, using a pointillism-style animation to showcase ChatGPT’s role in human progress- from the invention of the wheel to space exploration. The ad aimed to position AI as the next great technological leap, though reactions were mixed, with some praising its artistic approach and others finding it underwhelming.

Google’s Gemini Gets Sentimental

Google featured Gemini in a heartwarming ad about a father juggling job interview prep while raising his daughter. The AI assistant helped him with tasks like scheduling and resume-building, highlighting its practical applications. However, Google faced a hiccup before the big game, tweaking a YouTube ad after a Gemini-generated cheese fact was questioned for accuracy.

While OpenAI and Google dominated the AI conversation, other companies like Meta, Salesforce, and GoDaddy also showcased AI-powered innovations in their Super Bowl campaigns.

Did any AI ad stand out to you? (source)

Adobe has launched the Firefly Video Model, bringing generative AI to video, audio, image, and vector creation. Designed to be IP-friendly and commercially safe, Firefly aims to help creators generate and edit content without copyright concerns.

Key Features

  • Text-to-Video & Image-to-Video- Create and edit motion visuals with just a prompt.

  • Precision Editing- Adjust lighting, camera angles, composition, and motion with ease.

  • AI-Powered Audio & Translation- Translate and lip-sync videos in over 20 languages while maintaining the original voice.

  • Seamless Integration- Works directly within Photoshop, Illustrator, and Premiere Pro.

  • Commercially Safe- Built on licensed and public domain content, ensuring secure use for creators.

Adobe is rolling this out with Standard, Pro, and Premium plans, offering flexible options for different levels of content creation. (source)

AI is changing the way people work, and now there’s a way to track it. The Anthropic Economic Index looks at anonymized data from Claude.ai conversations to see how AI is actually being used across industries. The goal? To understand where AI is making an impact and how different job roles are evolving.

Key Findings from the First Report

  • AI is mainly used in software development & technical writing- but adoption is growing across various sectors.

  • AI augments work more than it automates- 57% of tasks involve collaboration, while 43% are fully automated.

  • Mid-to-high wage jobs- like programmers and copywriters- see the highest AI adoption, while manual labor and very high-paying professions remain largely untouched.

  • AI adoption is task-based, not job-based- only 4% of jobs use AI for most of their tasks, but 36% use it for at least 25% of their work.

Anthropic is open-sourcing the dataset and inviting researchers, economists, and policymakers to analyze AI’s economic effects. As AI evolves, the Index will track its long-term impact- helping shape future labor policies. (source)

Google DeepMind’s AlphaGeometry has set a new benchmark in AI-powered mathematics, solving Olympiad-level geometry problems with near gold-medalist performance. In a test of 30 International Mathematical Olympiad (IMO) problems, AlphaGeometry solved 25, surpassing previous AI models, which solved only 10.

How It Works

  • Neuro-symbolic AI- Combines a language model (for intuition) and a symbolic deduction engine (for logical reasoning).

  • 100M Synthetic Training Examples- AI-generated geometry proofs, eliminating reliance on human demonstrations.

  • Human-like Proofs- Solutions are verifiable, readable, and structured, avoiding brute-force algebra.

Why It Matters

  • AI is mastering complex reasoning, an essential step toward advanced problem-solving AI.

  • First AI to pass the IMO bronze threshold based on geometry alone.

  • Potential beyond math – AI-powered reasoning could transform science, engineering, and AI research.

This breakthrough brings AI closer to deep mathematical reasoning, paving the way for more general, knowledge-discovering AI systems. (source)

ByteDance, the company behind TikTok, has introduced Goku AI, a model designed for realistic video and image generation from simple text prompts. Built on Rectified Flow Transformers, Goku addresses major limitations in AI-generated media, including temporal coherence, fine-grained detail, and scalability.

Why Goku Stands Out

  • Joint Image & Video Generation- Excels at text-to-video, image-to-video, and text-to-image tasks.

  • Smooth, Realistic Motion- Overcomes flickering and abrupt transitions using Rectified Flow technology.

  • State-of-the-Art Performance- Scores 84.85 on VBench, ranking #2 in text-to-video generation.

  • Scalable & Efficient- Uses 100M+ curated training examples for superior detail and consistency.

How It Works

  • Multi-Stage Training- Starts with text-to-image learning, progresses to image-and-video joint training, and fine-tunes for specific tasks.

  • Advanced Motion Prediction- Ensures natural, fluid movement rather than step-by-step frame corrections.

  • AI-Assisted Editing- Enhances video generation with precise control and customization. (source)

Apple is exploring expressive robotics with a Pixar-inspired research project, showcasing a lamp-like robot designed to enhance human-robot interaction. In a recently published research paper, Apple highlights how subtle movements can make robots feel more intuitive and engaging- similar to Pixar’s Luxo Jr. lamp. 

A video demonstrates how the robot mimics human-like behaviors, such as turning toward a window before answering a weather query, making interactions more natural. While still in the research phase, this concept hints at Apple’s future consumer robotics ambitions, possibly leading to a more advanced smart home assistant. (source)

Krea AI- It is an AI-powered design tool that enables users to create stunning visuals, graphics, and animations effortlessly. With real-time AI generation and advanced editing features, Krea makes creative ideation faster and more intuitive for designers and content creators.

How to Access:

Step 1: Log in or Sign Up- Go to Krea AI and log in or create a free account if you're a new user.

Step 2: Choose Your Creative Mode- Select what you want to do:
Generate an Image / Edit an Image / Enhance Images / Generate Video - Create AI-powered motion visuals from a prompt or images.

Step 3: Enter Your Prompt

Step 4: Get Your AI-Generated Visual

  • Krea AI will process your request and display the generated content.

  • Refine or tweak the output if needed, and download your final creation!

  • OpenAI aired its first-ever Super Bowl ad to showcase AI’s role in shaping the "Intelligence Age", where technology empowers people to achieve more. The ad aimed to spark curiosity and highlight AI’s potential to enhance creativity, productivity, and innovation- just as past technological breakthroughs have. The deliberate use of Sora as a brainstorming tool rather than the end product acknowledges that AI’s greatest potential may lie not in replacing human creativity (yet?), but in amplifying it. 

  • Last week, we launched some more free AI & ML courses to build smarter AI systems. These free courses offer practical, hands-on insights into the latest AI tools and techniques- 

For me, the last few weeks have been dominated by open-source vs. closed-source. There have been times when you feel open source is crushing it, but just when you feel it - closed source comes back into the ring.

What has been the top of your mind?

Reply

or to participate.