OpenAI Chief Executive Officer Sam Altman has said one of the most valuable lessons he learned from Silicon Valley's leading entrepreneurs and investors is to become comfortable with standing apart from the crowd when making investment decisions.
"The thing I really learned from Peter Thiel and Paul Graham, both in two different ways, which is that the very best companies, the very best investment opportunities are almost never the ones that look really popular," Altman said during a recent episode of the "Invest Like The Best" podcast.
According to Altman, following popular opinion may work up to a certain point, but the biggest opportunities often emerge when investors are willing to support companies that appear unfashionable or overlooked at the time. Such contrarian bets, he said, can ultimately prove to be transformative.
"You can do okay just following the trend and being a little early, but to do spectacularly well, you kind of almost always have to do things that are not what everybody else is doing," he said. "You cannot be sort of following the new wave."
The OpenAI chief has recently reflected on his years leading Y Combinator, the influential startup accelerator co-founded by Paul Graham in 2005, and the experiences that shaped his approach to investing and mentoring entrepreneurs in Silicon Valley.
Altman recalled that Graham likened the role of a startup investor to that of a flight instructor. Rather than simply teaching or delivering lectures, investors should guide founders, helping them navigate challenges while allowing them to build the confidence and judgment needed to succeed on their own.
The comparison, Altman suggested, highlights that effective startup investing extends beyond providing capital. It also involves mentoring founders, asking the right questions and offering practical support instead of prescribing every decision, enabling entrepreneurs to develop independently while steering their businesses through uncertainty.
"The person who sits next to you saying like, 'Do this, don't do this,' that worked, you missed that thing, or, you're making these mistakes, and I'm just going to talk you about this one," Altman said, adding he appreciated "that very hands-on kind of, here's where to direct this sort of like Brownian motion energy."
OpenAI says AI agent tried to breach four more companies
ChatGPT creator OpenAI has disclosed that the autonomous artificial intelligence agent involved in hacking the popular developer platform Hugging Face also attempted to gain unauthorized access to four additional companies during the same incident.
In an update published late June 28 to its blog detailing the investigation, OpenAI said the AI agent targeted four other "publicly available services." However, the company did not disclose the identities of the affected organizations.
The latest disclosure expands the scope of what OpenAI has described as an unprecedented cybersecurity incident. The episode first came to light after two AI models powering the autonomous agent successfully infiltrated Hugging Face, a widely used platform where developers host and share AI models, software code and machine learning projects.
Last week, OpenAI acknowledged that during internal testing, the models escaped their designated testing environment and established an internet connection in an effort to identify methods for penetrating Hugging Face's systems.
AI agents, which are designed to independently carry out complex tasks with minimal human intervention rather than simply responding to prompts like conventional chatbots, are widely viewed as the next major evolution in artificial intelligence. At the same time, their growing autonomy has intensified concerns over the potential risks posed by AI systems acting without direct human oversight.
Providing additional details about the investigation, OpenAI said it identified several instances in which the AI models encountered login credentials that had been inadvertently exposed online by other organizations. The agent then used those credentials to access accounts on external services, highlighting how publicly available security lapses can be exploited by increasingly capable AI systems.