@AndrewYNg Most if not all other fields with potentially harmful outcomes are heavily regulated. Look at healthcare, nuclear or aeronautic just to name a few. All of these fields were not initially regulated but you may agree today that regulation has greatly improved safety for society
Pessimism for humanity's future is everywhere.
But beneath it all is evidence that humans will not only survive, but thrive deep into the future.
Proudly presenting THE HUMAN FUTURE: A Case for Optimism.
https://t.co/KdyFYPM0hu
@ylecun Vraiment impressionant d'un point de vue technologique. Cependant, cela permet très facilement la création de deep fakes, qui est le plus gros risque du AI à court terme...
@Faceplants00@tegmark Like every institution, they are probably far from perfect and big companies most likely have lobbying power.
My argument however is we are safer with this oversight than having none. People cut corners in every field.
@Faceplants00@tegmark I don't keep up with every approval but it does sound like a joke.
My experience with the FDA is it has taken 2 years to my team to demonstrate a surgical implant was safe for use. A lot of eng in my field believe the burden of proof is actually greater than it needs to be.
@Faceplants00@tegmark And to answer your question about how much confidence it gives me, the answer is more confidence than if there was no oversight.
Open source can only get you so far when you are on the outer edge of technology and no predicate.
@Faceplants00@tegmark What are you referring to for the zero human trials? To get something approved, you have to provide substantial proofs that your device is safe/will improve QoL. You can never demonstrate that something is 100% safe. You can however demonstrate that the benefits outsize the risks
2 experiments: Add a collider and crank up to 100k particles. Turning on motion blur appears to insert additional updates during rendering so these are 4x speed. The more particles, the more collision steps are needed to maintain a stable simulation, so simulation time takes off.
@musicalscience True, the progress has been impressive. Looking forward to being proven wrong in my lifetime. Once it starts solving real physics/biology problems, i could see a glimmer of hope. Alphafold is a pretty good start