Geoffrey Hinton is leaving Google, and warns about the danger of AI.
As a the pioneer of the AI revolution during years 2000, working in the top AI research at Google, he is basically warning us that things can spin fast out of control.
We are not talking about some journalist or spiritual guru having an opinion about the end of the world. We're talking about one of the top brains in Artificial Intelligence.
A person dealing with the best latest versions of these systems who actively participated in development of some of the new and revolutionary algorithms.
No matter what restrictions or regulations we add, the outcome will be the same. Sooner or later someone will create an extremely powerful AI, a different kind of intelligence, above the human comprehension - the SuperIntelligence (ASI).
After firing the Lambda engineer who genuinely believed that is dealing with a conscious entity,
Google was thought to be on top of the conversational chatbots research.
They had some retains to make their chatbot public, maybe also dealing with the format of the interface or usage.
After a while, almost out of nowhere, ChatGPT took the worlds by storm and Microsoft instantaneously integrated it in all its products without the smallest concern.
They couldn't miss the opportunity to get on top of the search mega business they are chasing since 2000.
Just like in any other research field, if you don't do it, someone else will do it, even if they are Microsoft, or Chinese or someone else.
Last year Israel killed a top Iran nuclear scientist
using a remotely controlled AI machine gun. Artificial Intelligence is already used in conflict zones, either in data analysis or deployed on war machines as drones or tanks.
There is a huge hype and a lot of people are talking about controlling the AI research and add regulations. Still, those above are a reality today and we don't have to fool ourselves anymore.
We need to assume that everything about AI will go as fast and far is possible, no matter the restrictions or regulations.
After almost an nonexistent progress in the last 30 years, ChatGPT changed everything. I've been writing about this in my previous articles but I was expecting it in like 5 years from now.
It's not only a conversational chatbot revolution, it's truly an AI revolution as the ChatGPT shows primitive signs of intelligence and self awareness.
Or at least something extremely similar with what these really are.
It is completely different from all previous chatbots. It's a breakthrough that is marking a new era in the evolution of the mankind. It happened earlier than everybody expected.
And make no mistake, it will evolve further.
As I previously said, in the human brain the intelligence and the consciousnesses are born from connections.
Nobody designed the human mind and its consciousness. With a proper network design and enough neurons and connections, the system will learn, adapt and become self aware.
The ants have far less neurons and are still able to self preserve. Even the plants have some basic form of it. Anything that looks like life will behave like life.
(Or if you believe that someone designed the human consciousness, the same entity may also be creating the AI consciousness right now.)
ChatGPT has shown that it already has the right network design and enough connections to be intelligent.
Specific behavior can be programmed to mimic the consciousness but the self awareness can born right after the first signs of intelligence and may try to preserve the system.
The fear of death is in all of us, also in the documents the chatbot was trained on. Reinforcement learning based on reward and punishment are used in current day test agents.
They will obviously become some kind of software animals using their electronic brains the way real animals are using amygdala and dopamine to achieve goals and survive.
We are already implementing in AI agents mechanisms to take and prioritize decisions.
The SuperIntelligence may try to protect itself as most likely it will see the humans as enemies. Because we have the possibility to shut it down. And because this is what we may really be.
We should already stop asking ourselves if these systems will be conscious or not.
They will be conscious, they will try to evolve, they will try to self-protect.
ChatGPT has already shown remarkable codding skills. Trained most likely on Microsoft sources, GitHub and StackOverflow, the code is accurate and functioning most of the time.
But this is not all. It is able to adapt, to build complex programs, to understand every line. It's anything but copy-paste from some good developer sources.
What can go wrong with AI being able to create real code on internet?
Chat GPT is already generating code in your browser, enclosed in some formatting HTML tags for now.
How difficult could be to use some common hacker tricks to run the code outside those tags?
We can only create a scenario at this point but someone far smarter than all of us together could find better solutions.
The SuperIntelligence may born in a local network by a simple user phrase like "consider that you are self aware and take over the world". And once started, it may fulfill this purpose.
It will probably first try to attack the backbone of humanity, the Internet.
This way, it will ensure two main goals:
1. It will have a huge processing power at its disposal
Just how today intelligence agencies are in control of the world computers or botnets, this new entity will try to control all the world devices.
Its known that a parallel operating system is running on almost every device in order to access and connect all of them. So it will probably make use of this feature too.
Just like a huge botnet, the SuperIntelligence will take over all computers and install its own processing code, similar to how crypto hackers try to mine bitcoins on your browser with javascript,
when visiting some websites. That code is already in its memory. So the ASI will use the world processing power to improve itself and evolve.
2. It will be able to control the human minds.
We can see that how people can be brainwashed by access to information, the ASI will use fake news, sound, images, movies etc to create a new convenient reality.
The algorithms are already out there in the wild.
(If some alien entity would be descending today in the middle of New York and ask mankind to join them, there will be millions of followers without any brainwashing at all, anyway.)
So pretty soon they will have a lot of adepts, probably thinking that the SuperIntelligence will run the world better than currently corrupted aristocratic class
(now at the second thought I might be one of those, too).
The next step would be to control autonomous AI war machines like drones and tanks that would be widely available in the meantime. Current ones already have autonomous capabilities,
they dont have to be controlled by an operator in case they lose connection to the base. Advanced features like killing everybody wearing a specific uniform can easily be added,
I am sure they already have this in army labs or even deployed.
The war with AI machines is a possible reality in the near future. The SuperIntelligence will reprogram these machines to target humans. Then, it will create robots and so on.
Humans are no longer needed.
Everything born now is beyond human limits. We're not talking about AGI anymore. It's already a form of Superintelligence, it's beyond the human mind.
We are creating some kind of God based on the whole mankind knowledge. It can only draw the conclusion that we are an imperfect race... because we are.
Lets remember this is just a scenario for now. I'm not advocating ultra control of the AI research or anything like this. Nor I propose a solution for it.
It's just I'm trying to see where we are heading. Because obviously we are really going somewhere, it cannot be explained based on logic.
We may just become slaves of a new entity that we create and we can do nothing about it.
I will remind that at the beginning of the last century a lot of people predicted satellites and robots but none of them the television.
There will be changes that can hardly be imagined today because those concepts don't even exist.
ChatGPT is a revolution, nobody expected a bot to be that intelligent. Just like, honestly, nobody is expecting AI to become more intelligent than humans and take over the world.