🤖 OpenAI’s “GPT” Improvements Are Slowing Down

⚙️OpenAI’s “GPT” Improvements Are Slowing Down

🎙a16z General Partner Says AI Regulations Focus on Fantasy, Not Reality

🦾Nvidia Unveils Robotics AI Toolkit

🛠Trending Tools

💰Funding Frontlines

💼Who’s Hiring?

Employees who tested OpenAI’s next flagship AI model, codenamed “Orion” or “GPT-5,” claim it displayed less impressive improvements than they saw between “GPT-3” and “GPT-4.”

In response, OpenAI has created a Foundations Team to figure out how the company can continue improving AI models despite the lack of new training data.

All high-quality data could be exhausted by 2026. So, Tech Giants are searching for real-world videos to continue training their AI models.

OpenAI is reportedly using “OpenAI o1” to generate high-quality synthetic training data (i.e., artificially created information) for “Orion.”

A lack of training data could cause Overfitting in AI models, which occurs when AI models learn the training data too well, leading to poor performance on unseen data.

During a Y Combinator fireside chat, OpenAI CEO Sam Altman boldly stated that Artificial General Intelligence (AGI) is coming in 2025. However, the company’s employees claim the “rapid pace of innovation” is slowing down.

Experts believe Large Language Models (LLMs) aren’t the true path to AGI because of their overreliance on data, lack of self-awareness, and lack of real-world understanding.

Andreessen Horowitz (a16z) is a Venture Capital (VC) firm that invests in technology startups in exchange for equity (i.e., “ownership”).

Casado leads the VC firm’s $1.25 billion infrastructure practice. Notably, he’s invested in ActionIQ, Ambient.ai, Braintrust, Coactive, and Ideogram.

“I think we’re a little bit early,” he said. “We need to understand what we’re going to regulate.”

Casado believes AI hasn’t matured enough to regulate it. “How can we regulate something we don’t understand yet?”

“Have you actually seen the definitions for AI in these policies? Like, we can’t even define it,” he said.

“Like, how is AI today different than someone using the Internet? If we’ve got a framework for how it’s different…we can apply policies that address those differences,” he said.

Casado and several Silicon Valley founders rejoiced when California Governor Gavin Newsom vetoed “SB 1047,” a law that put a kill switch into AI models to turn them off at any moment.

During the PC Revolution, many experts feared job loss, social unrest, and even the end of humanity. However, this fear didn’t lead to robust regulations of PCs before they caused harm.

If not for the patience of regulators in the 1980s resisting the urge to vet each PC, we may not have had widespread, affordable PCs in our homes as quickly as we did.

ActorShield raises a $3M Seed Fund to protect actors from AI deepfakes.

UnifyApps grabs a $20M Series A to help companies adopt AI applications.

Cloverleaf AI secures a $2.8M Seed Fund to uncover qualified government leads before anyone else.

Cultivarium (Watertown, MA): AI Software Engineer Intern, Summer 2025

Skydio (San Mateo, CA): Embedded Software Engineer Intern, Summer 2025

Rivian (Palo Alto, CA): Software Engineering Intern, Autonomous Features, Systems Engineering, Summer 2025

Nvidia (Santa Clara, CA): AI Software Engineer, AI Copilots, New College Grad 2024

Synechron (New York City, NY): Technical Business Analyst, Early Career