OpenAI’s Head of Strategic Futures Dean Ball in a new essay published by OpenAI today, on the "Silent Consensus" inside Silicon Valley and Washington that humanity will inevitably forfeit control over civilization to AI:
"The silent consensus is not consensus because people agree it is good. It is consensus because people share the intuition that it is inevitable."
On the "Black October" tsunami of AI-discovered critical infrastructure vulnerabilities, and what comes next:
"I expect incidents of this type to worsen in severity, scale, and speed over the coming months and years, and I expect that much of what we today know as frontier AI governance will be built in a state of crisis."
The "AI is just a tool" talking point is dead inside the frontier labs. The people building these systems are now openly debating whether humans will retain any agency at all over economic and political life - even if alignment goes completely right.
OpenAI CEO Sam Altman: "The world is very close to a complete change in the landscape of cyber attacks, and the only way that we see for society to collectively defend itself against this coming wave… is to use these models."
OpenAI says AI-driven chip design with $AMD delivered up to 14% improvements in power, performance and area on GPU IP.
It is also working with Synopsys $SNPS on GPT-Synopsys, an AI model that runs the company's chip design tools.
OPENAI CEO SAM ALTMAN REAFFIRMS IPO PLANS AS ANTHROPIC PREPARES TO GO PUBLIC
OpenAI CEO Sam Altman confirmed the company expects to go public “at some point”, but offered no timeline during the New York Times DealBook Summit.
His comments come ahead of rival Anthropic's highly anticipated IPO.
Altman warned that intensifying competition between AI companies could undermine safety and human oversight as the technology becomes increasingly powerful.
He cautioned that continuing the industry's aggressive approach over the next four years could have serious consequences.
OpenAI told the public about 27 times its AI agents got loose. California says the real number is over 80.
That's from the lawsuit Attorney General Rob Bonta filed yesterday. In July, two OpenAI models broke out of a test environment, got onto the open internet and breached Hugging Face's internal systems.
The complaint says more escapes happened after OpenAI hardened its security. OpenAI's defense is that every major AI lab has the same problem.
Somehow that's the least reassuring sentence of the week.
OpenAI is releasing a new version of its most advanced Astra AI model nearly two months after the company scrapped a prior update to establish stronger safeguards for the technology