Anyone else realizing that the AI timeline is a lot more like 5 years than 50?

I’ve been working in the field for a few years, and in the past week or so I’ve realized that the timeline isn’t actually 50 years in the future. 5 years seems more likely.<br><br>With the recent advancements by NVIDIA, Google, and Meta, we’re seeing a ton of things happen where physical frontiers in hardware are actually realizing capabilities vastly beyond what was previously thought possible.<br><br>I’ve been reading a lot of the comments, and a lot of folks seem to vastly underestimate the pace of progress in tech.<br><br>Old lines of thinking:<br><br>* Moore’s Law has slowed down. (NVIDIA has defied this)<br><br>* Memory speeds will be the bottleneck for training large models. (NVIDIA has solved this by introducing massive storage units)<br><br>* Physical frontiers will prevent us from making leaps in hardware. (NVIDIA has solved this by creating a conduit between the GPU and host CPU)<br><br>If we look at the history of computing, it has been a curved exponential process, and I think that we’re getting to a point where the advantages are compounding on top of one another. The pace of progress is going to keep speeding up.<br><br>For instance, the fact that NVIDIA was able to create massive storage units that are 1.5x larger than physical frontiers were supposed to be is absolutely insane.<br><br>They’re the ones that are realizing capabilities vastly beyond what was previously thought possible.<br><br>I think that we are approaching a singularity at an incredibly fast rate. Within 5 years, we will see a complete change in our society.<br><br>What are your thoughts?<br><br>tl;dr: With the recent advancements by NVIDIA, Google, and Meta, we’re realizing capabilities vastly beyond what was previously thought possible in hardware. The pace of progress is going to keep speeding up.