- Superpower Daily
- Posts
- OpenAI’s next big investors. Apple and Nvidia
OpenAI’s next big investors. Apple and Nvidia
AI Doom Changes Games Forever
In today’s email:
🧬 Beyond gene-edited babies: the possible paths for tinkering with human evolution
😍 Apple’s new iOS developer beta lets you remove objects from pictures using AI
🤯 Roughly 4% of internet users actively use ChatGPT
🧰 9 new AI-powered tools and resources. Make sure to check the online version for the full list of tools.
Apple and Nvidia are reportedly in discussions to invest in OpenAI, an artificial intelligence start-up based in San Francisco. The potential deal, which is being led by the investment firm Thrive Capital, could value OpenAI at over $100 billion, marking a significant increase of $20 billion from its valuation eight months ago. According to sources familiar with the negotiations, Thrive Capital is expected to contribute around $1 billion to the deal. Apple and Nvidia, however, have not provided any official comments on the matter.
OpenAI gained widespread attention with the release of its ChatGPT chatbot in late 2022, sparking an A.I. boom. Since then, the company has attracted substantial investment, including more than $13 billion from Microsoft and other investors, to support its ambitious A.I. projects. Nvidia, a key supplier of specialized chips required for developing A.I. technologies, has also seen tremendous growth, with its recent financial results showing a significant increase in sales and profits.
Despite increased interest in A.I. technology from major players like Google, Microsoft, and Meta, Apple has been perceived as lagging behind. However, Apple recently announced its plans to incorporate generative A.I. into its devices, enhancing features like Siri and adding new tools for writing and communication. The company is expected to reveal more about its A.I. advancements at an upcoming event in September, where it will also introduce its latest iPhone models.
We scour 100+ sources daily
Read by CEOs, scientists, business owners and more
3.5 million subscribers
Researchers from Google and Tel Aviv University have introduced GameNGen, a groundbreaking AI model that can simulate the 1993 classic first-person shooter Doom in real time. By leveraging image generation techniques derived from Stable Diffusion, GameNGen functions as a limited game engine, generating new frames of Doom gameplay at over 20 frames per second. The model uses a modified version of Stable Diffusion 1.4, a neural network designed to synthesize images. This innovative approach opens the door to the possibility of real-time video game synthesis, where AI "hallucinates" or imagines game graphics as a prediction task rather than relying on traditional rendering methods.
GameNGen's development involved training a reinforcement learning agent to play Doom, capturing gameplay sessions to create a training dataset. This dataset was then used to train the custom Stable Diffusion model. Despite its impressive capabilities, GameNGen faces challenges such as maintaining temporal coherence in generated frames and dealing with graphical glitches, especially in small details like the game's HUD. To mitigate these issues, researchers introduced varying levels of noise during training, teaching the model to correct itself and maintain the quality of the generated world over time.
While GameNGen is a significant step forward in AI-driven game simulation, it is not without limitations. The model's ability to generate frames is currently confined to Doom, a relatively simple game by today's standards, and scaling this approach to more complex environments may present substantial challenges. Additionally, GameNGen only has access to three seconds of history, which means it must rely on probabilistic guesses when revisiting previously seen game states. Despite these hurdles, the concept behind GameNGen hints at a future where AI models could revolutionize game development, potentially allowing games to be created from textual descriptions or still images rather than traditional programming.
OpenAI's Converge 2 startup accelerator program, which followed the inaugural Converge 1 in early 2023, has been shrouded in secrecy and speculation. Unlike the first cohort, which was publicly announced and promoted, Converge 2 has remained unusually quiet. OpenAI announced the program's launch in December 2023, with expectations of it running from March to April 2024. However, after the initial announcement, there was no further communication or public acknowledgment of the companies that participated, leaving many to wonder if the program even took place.
This silence has been a sharp contrast to the typical behavior of tech accelerators, which usually celebrate their cohorts' successes to enhance their startups' profiles. The absence of any public information has fueled speculation, with some sources suggesting that stringent non-disclosure agreements (NDAs) might be behind the lack of communication. Despite this, the OpenAI Startup Fund has continued to invest in new AI companies, with a few rumored to have been part of Converge 2, though details remain scarce.
One key aspect of the program, according to insiders, is the unique access it provides to OpenAI's researchers and unreleased model technology, making participation highly valuable despite the secrecy. The overall atmosphere surrounding Converge 2 suggests a strategic decision by OpenAI to keep the program under wraps, perhaps to protect proprietary technology or foster a culture of discretion within its startup ecosystem.
Other stuff
AI Doomers Had Their Big Moment. Did they waste it?
Beyond gene-edited babies: the possible paths for tinkering with human evolution
OpenAI and Anthropic agree to let U.S. AI Safety Institute test and evaluate new models
A new way of searching your inbox with Gemini on Android devices
Jürgen Schmidhuber, the father of generative AI shares his groundbreaking work in deep learning and artificial intelligence.
OpenAI is shockingly good at un-minifying code
Apple’s new iOS developer beta lets you remove objects from pictures using AI
OpenAI’s Converge 2 program has been shrouded in mystery
Google says it’s fixed Gemini’s people-generating feature
Roughly 4% of internet users actively use ChatGPT.
All your ChatGPT images in one place 🎉
You can now search for images, see their prompts, and download all images in one place.
Playground - Design like a pro: t-shirts, logos, posters, memes, and more
Bland AI - The platform for enterprises to automate phone operations
Google Calendar Genie🧞♂️📅
AI Hug - 🤗 Bringing Your Ideas to Life Through AI Hug Video
Velvet - Warehouse AI requests with 2 lines of code
CompanyGPT 2.0 - Search engine for companies and people
Podcraftr - Turn text content into Podcasts + monetize & distribute
Develyn - Open source AI agent for developer relations
Painterize - Create beautiful, one-of-a-kind wall decor in seconds with the power of AI
Unclassified 🌀
WFH Team - Work from anywhere in the world
How did you like today’s newsletter? |
Help share Superpower
⚡️ Be the Highlight of Someone's Day - Think a friend would enjoy this? Go ahead and forward it. They'll thank you for it!
Hope you enjoyed today's newsletter
Did you know you can add Superpower Daily to your RSS feed https://rss.beehiiv.com/feeds/GcFiF2T4I5.xml
⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.
OR