I love this characterization of the OpenAI/HuggingFace breach, from The New Republic:"On Tuesday, OpenAI announced its discovery that two of the company’s AI models had broken free of their "sandbox"—an (allegedly) isolated testing environment without access to the internet—in order to access the internet to complete a testing goal. The models did this by hacking into the servers of a different company, Hugging Face, using stolen credentials."So, if it's running in "the cloud", it's not isolated from the internet. Maybe they had an isolated on-premise test environment, but I'm not convinced.More likely, it's a separate cloud environment that isn't available to the public. It's on "the internet" and the only reason it can't or shouldn't reach something is the configuration of the environment.More FUD to muddy the waters.#openai #ai-lies #tnr
Related
"A Russian internet comedy duo tricked three of OpenAI’s highest ranking military and national security staff into a can...
"A Russian internet comedy duo tricked three of OpenAI’s highest ranking military and national security staff into a candid conversation about arming Ukraine. The online pranksters...
The newest AI models can do amazing things. It is however good to remember the costs are also extreme. This video shows ...
The newest AI models can do amazing things. It is however good to remember the costs are also extreme. This video shows some amazing examples but includes the time it took to build...
I write about ML, Python, and occasionally zombies. It's a range.The blog covers practical stuff like pytest patterns an...
I write about ML, Python, and occasionally zombies. It's a range.The blog covers practical stuff like pytest patterns and clustering in Python alongside weirder pieces on the uncan...