People Actually Using AI in Workflows at Large Corporations - Please Chime In.

As someone who doesn't work for a large company and doesn't use AI much at work outside of asking claude an occasional question - I have a very hard time of parsing the news flow and trying to understand how capable currently models actually are, and where things are headed. I would really appreciate people who are much more hands on with this stuff, and ideally involved in integrations at large corporations, shedding some light.

The news flow is a constant ping pong between "This is going to eliminate all white collar work in X years" and "It's vaporware/it doesn't do anything/it isn't good enough" - again, as an average joe, I have no real way of deciphering the truth.

My intuition is that while the models are powerful and its easy to recognize potential use cases, the implementation is the issue. It's cliche to talk about the parallels between the internet bubble and current AI hype - but I think its a useful analogy here.

In 2000 everyone was able to recognize the value of the internet and long term implications, but the thought was that we just needed more infrastructure to realize that long term vision. In retrospect, the value creation didn't necessarily come from the infrastructure. Of course we use a lot of the fiber that was laid at that time now, but I would argue that the main difference between the bubble period and the eventual boom, was people figuring out more complex and valuable use cases/implementations. Yes we had Amazon, google, etc in 2000, but the amazon, youtube, netflix, facebook of today are much more powerful use cases than anything that existed at that time.

I feel like we are perhaps in a similar place with AI - we can see the long term potential, and many believe we "just need more compute" to realize that potential - but my intuition is that we are on an internet-like trajectory. Eventually this compute will be used, and we will need much more than we are even anticipating today, but compute alone is not going to bridge the gap between current capabilities and the real value creation - to do that, some significant innovations will need to occur that drive the technology meaningfully forward in ways that more compute cannot.

As I said - this is just the perspective of an average joe who isn't immersed in the technology, so I would really appreciate the thoughts of those more knowledgeable. Thanks!