ChatGPT is actually trying to replace humans. But it's not ready.

There are many shortcomings to ChatGPT. For example, for many topics it provides only mediocre answers. There is also a limit to the amount of working memory it has (“context window”). It sometimes requires a lot of prompting to get it to perform exactly as you would like.?

These are real foundational problems that prevent greater habitual use of the ChatGPT technology.

Perhaps it was naive of me, but I had thought that OpenAI’s roadmap would focus on solving these types of problems first, before trying to make the technology even more mainstream.

Instead, an important part of the mission, Mira Murati (OpenAI's CTO) tells us in the announcement for the new GPT4o model, is getting the technology in the hands of as many people as possible, in as frictionless a way as possible.

What this means for OpenAI is?taking the current ChatGPT technology and making it more human like, with emotions and sing-songs.

It is impressive. The latency (i.e. the wait time for responses) is negligible, and ChatGPT seems to have even more personality than the Sophon from Three Body Problem.

ChatGPT clearly excels at party tricks, and we are treated to a nice collection of them on OpenAI's YouTube channel.

However, businesses are going to be left only more frustrated. We now have, on the surface, a convincing demo that ChatGPT can be a replacement for a humans for many office tasks that AI has long been promising to displace.?

In reality, ChatGPT still does not have the long-term memory and adaptability of a human. It also does not yet have a way to coordinate numerous software applications and accomplish a business goal. It can also sometimes take a lot of work to get ChatGPT to understand what you want to accomplish.

The proof is in watching the demo from OpenAI's President, Greg Brockman, where he has to ask the AI about the interloper to get it to notice it, whereas even a child would have thought it strange and noteworthy to bring up without prompting. Even then, was the amount of effort required to get these two AIs to sing together frustrating only for me? Perhaps humans still have a shot at trumping AI at party tricks.

We always knew that OpenAI was trying to build a human-like AGI. I just wish they would work on the fundamentals so, when the human-like features are added, it could actually be comparable to humans.?I'm concerned that GPT4o, while technically impressive, is an incremental step forward that doesn't address any of the larger challenges of LLMs.