Claude, the product of Anthropic, is an AI that was able to create a C language compiler using several “agents” communicating with each other. While some of the claims are a bit exaggerated, it is still an impressive feat.
It also makes you realize, it could be given a far more nefarious task - feed it all the worlds cybersecurity flaws, every hack, every compiler trick, all the computer science embedded into silicon, etc., and create the most sophisticated AI cyberattack ever conceived.
I believe it’s inevitable, especially considering the amount of security flaws per 1000 lines of code are *increasing* due to the use of AI to generate “vibe code”.
The difference between AI and the A-bomb is stark, those striving to build the bomb knew exactly what they were working toward (aside from possibly setting the atmosphere alight!). With AI, we’re all competing, to not be in 2nd place, with a technology that we don’t quite understand what the end goal looks like.
We’re headed for an historical event unlike anything the world has ever seen...but I have no idea what it’ll look like, it just seems inevitable.
Specs will be the new coding language. I suspect they will come up with a standard format for the requirements doc, and can pretty much compile the code straight into machine language.