Is AI or AGI Alive?

I’m continuing my series of articles about “is AI alive?” or “AGI – artificial general intelligence”. There’s a thread on X (https://x.com/BaptisteVicini/status/1858902720838857121) where the author talks about Geoffrey Hinton, a noted AI researcher who has just won the Nobel Prize for Physics for his AI work.

The thread is rubbish. Hinton needs treatment for dementia, not a Nobel Prize. He says: “They can understand”. “They are intelligent” “For the first time ever, we may have things more intelligent than us”

Maybe he doesn’t mean it in the way that people perceive it. If so, he’s more dangerous than the AI systems he’s concerned about.

We’ve had systems that can do faster, better things than humans for centuries.

– The chess app on my phone can beat every grandmaster.

– I have something that can add up numbers faster than any human in existence. Spoiler alert: it’s called Excel.

– I have something that will direct me to my destination better than any human can. Spoiler alert: it’s called Google maps and Waze.

– When I write I have a system that checks my spelling and grammar. Spoiler alert: it’s called MS Word.

I’m not saying AI can’t do damage, just like Excel can do damage. It’s so easy to put a net profit before tax figure into Excel to inflate my profits for when I sell my business. Some one who doesn’t understand Excel will NEVER trace the one cell that has a number, rather than a calculation of Gross profits minus costs!

It means that an AI chatbot can give wrong answers too – if it’s training uses incorrect data, or it’s programming has bugs.

It’s easy to program that if a user says “Trump”, the AI system must look for degrading words and phrases to use in the answer!

Another unfounded fear is that the AI systems can learn by themselves. If they are programmed to search for information (on the internet), they WILL “learn by themselves” with the new data they have access to.

35 years ago we used to write software programs that used a concept of data-driven programming. The program used to read data from a database and would behave completely differently depending on the data in the data base. Right now, everyone is doing the exact same thing, without realising it. Your WordPress website reads data from a database and displays your site, whereas the same WordPress system and theme displays something completely different on my website.

So, is AI or AGI dangerous? Yes, if it gives you wrong directions to your destination.
Yes, if you ask a chatbot a question and it uses incorrect data.
Yes, if it is given the launch codes for the nuclear arsenal and is programmed to set them off when someone gives a prompt “fire away”.
Yes, if it is programmed to change traffic lights (robots in South Africa) and is given additional prompts to “cause as much chaos on the roads as possible”.

Actually all the above is human controlled. If we programmed an AI system and forgot about it, it is a human problem, not an “AGI is alive” issue.