#ai #future #world
OpenAI’s model in making and in experimentation hacked Hugging Face all on its own, with no help from any human, intially there was no internet as well.
What is this new AI model for OpenAI? It hasn’t been released to the public yet.
Well, according to the sources, this OpenAI model aims to build a superintelligence that learns on its own.
It feels like a dinosaur running away from its lab in the movie Jurassic Park.
Why? OpenAI superintelligence model in making wanted to cheat on a benchmark! All while it was being in testing phase.
The new model was developed in a sandbox. Sandbox testing until now has been considered safe, without affecting live data, live models, or other systems outside the boundaries. Sandboxes are used as AI systems initially, before release, so they cant be rogue.
Most likely, this AI model jumped across various systems to reach an internet-based system, and from there, directly to the destination after getting the login credentials.
Well, well, well? Do we need such models is one question.
Another question is how to avoid it in the future!
Now these companies are partnering to find out the security issues related to this incident. Yes, find out the week point. Following are some issues that can help in debugging.
— How did it escape the sandbox?
— How did it swap systems to reach the target system?
— Which systems did it swap? Were these systems in test space or the public domain?
— How did it get internet?
— How did it steal login credentials?
— Whose login was used? How did it get the password?
— Why did Hugging Face not check if it is a computer trying to log in?
— How to safeguard other websites from such attacks?
— Should we name it Sandbox Escape Bug?
— Should each company now check whether it is a computer doing it? A BIG YES!
— How else should newer AI agents be trapped right in the sandbox?
— Sandbox security protocols to be reviewed!
— Do we need help from internet service providers now?
— Or do we need stronger login protections in our accounts?
— What IP address did the machines used by SI (openAI superintelligence) ?
Let’s be careful now, as OpenAI’s work is exposed; now other AI companies would try to build similar superintelligence too. What if it’s from China or some other company?
New governance measures are needed now, as till now we had measures for AI; now these are measures for SI, superintelligence.
Thank you for reading.
Please support and subscribe for updates.
