Wow, just heard the CEO of Nvidia estimate that AI will double in capability, speed, and power each year. He said it might even triple.
1. Moors law says computing power doubles every two years and AI performance is directly correlated with computing power.
however, it is also directly correlated with software, data, and training. So we have three or four different ways to improve AI all working together.
When you consider AGI what does that mean? You have strategy and you have tactics. Already AI is better at humans tactically. It can juggle hundreds of drones and respond faster and more precisely than a human can. Even if you are strategically smarter, being tactically weaker is a clear disadvantage.
We also know that AI has achieved ASSI, that means for a specific task it can have super intelligence. Suppose you are the FBI and you have an employee that has super intelligence in detecting lies during interrogation. That alone is enough to make them a highly ranked agent. AI has that ability, it is used to detect microexpressions and respond with probing questions. But AI also has super intelligence when it comes to looking through old case files for patterns. One super intelligence would put the AI in the top 5% of agents, but two super intelligences? Top 1%? It can also scan through video footage faster than humans. But that is not all, it will give you an answer much quicker than anyone else. Very often speed is a critical factor. But that is not all, it is available in any FBI office that has a computer. My point is that even though it has not yet arrived at AGI, or ASI, it is still superior than just about any agent in the FBI. Not in all skills, but guess what, it is doubling or tripling in its capability each year. Once AI is deployed you know that the data it will be receiving will also increase as will the computing power. To me, the singularity is not when AI can replace 100% of FBI agents. No, it is when it can replace 20%. At that point there is no more reason to hire new agents and instead you can start early retirement and layoffs.
The same is true for doctors, lawyers, accountants and any other white collar job.
But my point is not that AI is about to take all the jobs, that has been repeated over and over. No, my point is that if the thing is doubling or tripling in capability each year then it is impossible for us to actually know how capable it is right now. The latest benchmarks put it scoring quite well on college exams and graduate level exams of every kind. What does it mean to double or triple that performance? My point is this is not going to be a gradual change, it will be sudden, much more sudden than the iphone. I remember one year no one had them the next year it seemed everyone did. The first iphone was sold in 2007. So the complete transformation took place in about ten years. I think this is the first year of a complete transition to AI which will take about 2 years. That doesn't mean no one will have a job in two years, it just means AI will be doing everything and the chance of anyone getting a job after April of 2027 is probably extremely close to 0.
It also means there will be extreme pressure to become a cyborg. Have a chip implanted in your brain and you will still be a functioning member of society, but you also become a drone controlled by AI.