AI Just Changed Everything … Again

How will you move forward with regard to AI? Go to brilliant.org/Undecided/ and get 20% off your subscription and a 30 day free trial with brilliant.org/! If you liked this, check out China’s MASSIVE Desert Project Is About To Change The World https://youtu.be/MX_PeNzz-Lw?list=PLnTSM-ORSgi4At-R_1s6-_50PCbYsoEcj

What troubles me most about AI is that our politicians aren't knowledgable enough to regulate AI properly and they are greedy and corrupt enough to let AI companies make the rules.

“I don’t think we need to panic” yeah, not a single chance we will hold the tech companies accountable until its at a complete boiling point and they have either seized full control or sucked up every last dollar possible.

There is a new Turing test. Its how long you will talk to an AI customer service bot before realizing you are getting nowhere and hang up.

Incredibly honored to be featured on one of my favorite channels of all time. Thank you so much for having me!

“Hold these tech companies accountable” how, exactly? The corporations don’t answer to us. The government doesn’t even answer to us. Without explaining how, suggesting we “hold them accountable” is totally empty.

18:12 "..supporting human creators like me" That's exactly what an AI pretending to be a human would say.

We used to say “don’t believe anything you hear, and only half of what you see”. . . We need a provenance system for ANYTHING that is delivered electronically.

The biggest issue in my opinion is the feedback loop. LLMs are mostly trained on data publicly available on the internet and they generate new data from this. This generated data is often wrong but is still being published as some sort of truth. Then the LLMs will train again on this new incorrect data generating even more wrong data. At some point, it will train on so much junk data that it will become pretty much useless but it will have destroyed the internet in the process.

Worst part: AI will earn profit for large corporations - younger generations will get nothing slowly turning into third grade citizen.

You don't push a toddler on a tricycle down a hill! Man I wish I knew that before.

Wow, talk about a trip through memory lane! After my PhD in robotic vision in 1983, I began working for FLAIR (Fairchild Laboratory for Artificial Intelligence Research) where, among several areas, we were doing research in "speech recognition", the precursor of LLMs. So, yes, the roots are tens of years deep.

My favorite example of neural networks in the past was the 1996 game "Creatures". You looked after creatures called Norns and they each had a neural network to learn behaviors and language. It was also paired with a genetic and basic biology simulation for them to learn from things that made them feel good or bad. You could micromanage and be a helicopter parent trying to make your Norns behave the way you wanted or you could just leave the game running and see what happens. I've been really hoping the new AI trend will bring the series back. Not exactly a world changing, helpful use of neural networks but it's always stayed with me and probably what drew me to Biology when young.

The bizarre thing is the voice mode you used in the intro is the old version as far as I know, it's crazy good and will be even better

18:09 unfortunately, humanities tombstone will read "Best of Intentions, but Convenience Ended It All"

For years I have used the internet for information but it now is full of useless information and garbage it is impossible to extract useful information for learning. I am at the point of dumping social media and the internet for some sanity of real life and leave the internet to the bots.

In the mid 1980's, as I was doing a masters degree in process control engineering, we investigated neural networks. Unfortunately at the time, it quickly became apparent that computing power was the limiting factor. Definitely not new technology!

"You don't push a toddler down a steep hill on a tricycle." I felt personally attacked. My kid loved this... lol.

I like your point on the black box. One of my largest problems learning math in school was the way they made it into a black box. It was not until I learned some number theory that the blinds started to be lifted.

Companies like Open AI should basically have to pay a licensing fee for all the data to the whole world. The money should then fund public projects that benefit everyone, possibly toward a UBI. While it's true that we all get inspiration from the work of many others for free all the time, the scale of this is so huge that they should pay the world.