Using AI to audit code and eradicate any weaknesses adversaries could possibly use and keep hidden away?

https://twitter.com/emollick/status/1759774087767900346

Hi guys, I hope this speculative post is suitable for here. This link is about the new Gemini Pro LLM with an astounding 1 million context window. This man got it to read over 1000 pages of pdfs of his previous work and it was finding quotes and themes throughout all of it. This is mindblowing stuff for an AI to be able to do.

Now, this is Gemini, so a google product riddled with privacy issues, but imagine 5-10 years down the line when an open source model is capable of doing this but far better. Imagine how powerful that could one day be for auditing code and creating privacy tools.

Tell it everything about Tor code and how it works, and it can analyse it all and find any possible weakness in the code.

It could be done with any software to completely eradicate potential holes and bugs and weaknesses in the code. Couldn’t this be completely game changing for developers? What are your thoughts on this?

3 Likes