Musk vs. Altman: The $97B OpenAI Showdown!
Plus: Altman announces GPT-5 roadmap, OpenAI’s AI research assistant outperforms GPT-4o, Softbank and OpenAI partners for ‘Cristal Intelligence’, and more.
Hello Engineering Leaders and AI Enthusiasts!
This newsletter brings you the latest AI updates in a crisp manner! Dive in for a quick recap of everything important that happened around AI in the past two weeks.
And a huge shoutout to our amazing readers. We appreciate you😊
In today’s edition:
🤼 Musk vs. Altman: The $97B OpenAI showdown!
🗺️ Altman announces GPT-5 roadmap, cancels o3 launch
🤖 OpenAI's AI research assistant outperforms GPT-4o
🤝 Softbank and OpenAI partners for 'Cristal Intelligence'
🎬 ByteDance's new AI transforms photos into lifelike video
⚡ Mistral reveals one of the world's fastest AI assistant
🧮 DeepMind AI beats Maths Olympiad gold medalists
📈 Perplexity's new AI beats GPT-4o, Gemini and Claude
📚 Knowledge Nugget: Generative AI and the Self-Diagnosis Pandemic
by
Let’s go!
Musk vs. Altman: The $97B OpenAI showdown!
Elon Musk and a group of investors have made an unsolicited $97.4 billion bid to acquire control of OpenAI, a company Musk co-founded. The move aims to return OpenAI to its open-source, safety-focused roots, as Musk believes the company has strayed from its original mission.
OpenAI CEO Sam Altman dismissed the offer, leading to a public clash between the two tech moguls over the future direction of the influential AI company.
Why does it matter?
The rivalry between these Silicon Valley powerhouses shows no signs of cooling, as Altman predictably rejected the overture amid Musk's legal battles and provocations. Their shared connections to federal tech initiatives suggest Musk's crusade against his former venture will continue to unfold in the public eye.
Altman announces GPT-5 roadmap, cancels o3 launch
Sam Altman announced that OpenAI will release GPT-5 in a few months, canceling the standalone o3 model. Instead, the features of o3 will be integrated into GPT-5. The upcoming GPT-4.5 will be the last model without chain-of-thought capabilities, launching in weeks.
GPT-5 aims to unify OpenAI's tools, making them easier to use. It will be available for free users at a standard level, while Plus and Pro subscribers will access more advanced features. This change responds to growing competition in the AI market.
Why does it matter?
AI models' complex names and structures have created confusion for users. A unified approach where the system automatically chooses optimal models and tools could streamline this experience, transforming how we access and interact with AI.
OpenAI’s AI research assistant outperforms GPT-4o
OpenAI has launched Deep Research, a new AI tool that uses a specialized version of o3 to conduct in-depth online research on complex topics in minutes. It scours the web for relevant text, images, and PDFs across multiple sources to produce comprehensive research reports with citations. The research hardly takes around 5-30 minutes to get completed.
Deep Research significantly outperformed other AI models like GPT-4 and Gemini Thinking on benchmarks like Humanity's Last Exam. The tool is initially available to paid ChatGPT Pro subscribers at $200/month, with plans to expand access later.
Why does it matter?
While ChatGPT provides immediate responses, Deep Research breaks new ground by handling multi-day tasks for consumers. Together with Operator's launch, AI is evolving toward extended analysis and independent actions, yielding superior outcomes.
Softbank and OpenAI partners for ‘Cristal Intelligence’
SoftBank and OpenAI announced a partnership to develop and market an advanced AI system called 'Cristal Intelligence' for enterprises. SoftBank will invest $3 billion annually to deploy OpenAI's technology across its subsidiaries like Arm and PayPay.
They formed a joint venture called SB OpenAI Japan to accelerate Cristal Intelligence's deployment and customization. Cristal Intelligence will securely integrate individual companies' systems and data in a tailored manner, aiming to boost productivity and drive innovation.
Why does it matter?
SoftBank's annual $3B commitment showcases Japan's ambitious AI strategy. Through expanding partnerships with OpenAI, Stargate, and Cristal Intelligence, the company is positioning itself to control AI implementation throughout Asia.
ByteDance’s new AI transforms photos into lifelike video
ByteDance has unveiled OmniHuman-1, an AI model that can generate highly realistic and vivid videos of humans from just a single photo. The model can animate head movements, facial expressions, gestures, and full-body motions that precisely match input audio or pose data.
OmniHuman-1 supports various portrait styles, body proportions, aspect ratios, and input modalities like audio, video, or combined signals. It achieves superior gesture generation and object interaction capabilities compared to existing methods by leveraging an innovative "omni-conditions" training method that scales up the model on large, mixed-condition datasets.
Why does it matter?
While ByteDance's OmniHuman-1 remains unreleased, its demonstrations showcase AI-generated content that's indistinguishable from reality. This marks a critical juncture – as such technology spreads, society must find ways to authenticate content in an era of flawless synthetic media and deepfakes.
Mistral reveals one of the world’s fastest AI assistant
Mistral, a French AI company, has launched Le Chat, an AI assistant touted as the world's fastest. Powered by Cerebras' AI inference engines, Le Chat can generate responses at a blistering 1,100 words per second, which is claimed to be 10 times quicker than ChatGPT, Claude, and other rivals.
The key feature enabling this speed is called "Flash Answers." Le Chat offers capabilities like web search, document analysis, code interpretation, and image generation. It is available as an iOS and Android app with a free tier and paid Pro ($14.99/month) and enterprise options for custom deployments.
Why does it matter?
Though not leading in raw power, Mistral AI's emphasis on fast, adaptable deployment highlights a strategic pivot in AI competition - where practical benefits like speed and business integration may matter more than pure capability in capturing market share.
DeepMind AI beats Maths Olympiad gold medalists
DeepMind has launched AlphaGeometry2, an advanced AI model that has outperformed gold medalists in the International Mathematical Olympiad. This new model solved 84% of geometry problems from the past 25 years, a significant jump from its predecessor's 54% success rate.
AlphaGeometry2 combines a language model with a symbolic engine, allowing it to tackle complex problems. It was trained using synthetic data generated from random diagrams, making it capable of solving geometrical challenges that typically require deep mathematical understanding.
Why does it matter?
Despite LLM’s historical weakness with mathematics, DeepMind's breakthroughs in complex problem-solving signal a turning point. These advances in mathematical reasoning could catalyze progress across scientific fields that rely heavily on mathematical foundations.
Perplexity’s new AI beats GPT-4o, Gemini and Claude
Perplexity has launched its new AI model called Sonar, available to Pro users. Built on Meta’s Llama 3.3, Sonar is extremely fast, producing 1,200 tokens per second. It outperforms OpenAI's GPT-4o mini and Claude 3.5 Haiku, matching the larger models GPT-4o and Claude 3.5 Sonnet in performance. Sonar is 10 times faster than Google’s Gemini 2.0 Flash.
The Sonar API is offered in two versions, with the Pro version designed for complex tasks and deeper understanding. The pricing is competitive, making it an appealing option for users seeking advanced AI capabilities.
Why does it matter?
With rapid-fire releases like its new high-speed model following Sonar's debut, Perplexity AI showcases growing momentum. By delivering both exceptional speed and accuracy that rivals industry leaders, the company signals ambitious plans to expand its market presence.
Enjoying the latest AI updates?
Refer your pals to subscribe to our newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the “Share” button on any post, you'll get the credit for any new subscribers. All you need to do is send the link via text or email or share it on social media with friends.
Knowledge Nugget: Generative AI and the Self-Diagnosis Pandemic
In this post,
warns about the risks of people using AI tools like DeepSeek and ChatGPT to diagnose medical conditions. She shares a recent incident where DeepSeek misinterpreted blood test results and wrongly suggested a patient needed kidney dialysis.While AI can be helpful in healthcare for things like explaining medical information and reducing doctor burnout, the author emphasizes that AI tools shouldn't be used for self-diagnosis. She stresses that medical decisions should always involve human doctors and that patients need better education about AI's limitations.
Why does it matter?
The growing trend of AI self-diagnosis could create a dangerous cycle of health anxiety and misinformation, potentially delaying proper medical care. This highlights the urgent need for digital health literacy and clear guidelines around AI's role in healthcare.
What Else Is Happening❗
🤖Google launched new Gemini 2.0 AI models, including Flash, Flash-Lite, and Pro, enhancing performance and efficiency for users and developers across various apps.
🌍At the Paris AI Summit, the U.S. and UK declined to sign a declaration on ethical AI, raising tensions with Europe over AI governance and investment.
🔍Cloudflare has integrated Adobe's Content Credentials system to help users verify the image authenticity, tracking their ownership and any AI manipulation across its network.
🛡️Anthropic challenges users to jailbreak its new AI model, Claude, after extensive testing showed its new safeguards successfully blocked most attempts to bypass security.
📸Snap’s new AI text-to-image model for mobile allows rapid high-resolution image creation directly on iPhones, enhancing features like AI Snaps and Bitmoji backgrounds.
📜Adobe's Acrobat AI Assistant includes new GenAI features that simplify understanding contracts, helping users quickly grasp terms and compare multiple agreements.
🚀John Schulman, ex co-founder of OpenAI, has left Anthropic after six months, seeking new opportunities despite enjoying his time at the AI safety-focused company.
⚡OpenAI has upgraded ChatGPT models o1 and o3-mini with file and image upload capabilities, while Plus users get increased daily call limits for the o3-mini-high model.
🎬Adobe launched the Firefly Video Model, an AI tool for creating videos with high creative control, ensuring IP safety and integration with existing Adobe apps.
📂Perplexity AI now allows users to upload files and images for free, with a 1 million token context window, enhancing interaction and accessibility.
New to the newsletter?
The AI Edge keeps engineering leaders & AI enthusiasts like you on the cutting edge of AI. From machine learning to ChatGPT to generative AI and large language models, we break down the latest AI developments and how you can apply them in your work.
Thanks for reading, and see you next week! 😊
https://aicounsel.substack.com/p/musk-v-altman-gains-steam-usd97b