Theory, ClosedAI (M$) has focused on coding to win over tech people, tech people tell average people to use their products and it snowballs.

Do you think it’s a coincidence Microsoft bought GitHub and all of a sudden they have a coding focused bot ? Weird right ? Think of all that sweet training material.

It's a fair theory. I think coding is so pivotal, or scary, because it starts to prove these things can write or fix themselves. I'm still not sure whether it's impressive, depressing, or scary.

Just answering questions is a glorified Ask Jeeves.

Hmm not sure if it means it can self-improve itself exactly. It feels like a lot of variables involved there and often the code produced is just an approximation of an existing requirement / something in the training data.

But yeah, I’m not quite sure what the end goal of l this research is supposed to be either. Seems like we’re not acting very intelligently on many fronts.

In an ever-evolving AI landscape, data moats have become increasingly important. Companies that possess large amounts of generated human content, such as Microsoft's acquisition of GitHub, are likely to see an increase in value as a result of this development. The unique data generated by these platforms can give companies an advantage in training and fine-tuning AI models, potentially allowing them to stay ahead of the competition. In this context, a data moat serves as a protective barrier against competitors, ensuring that the company continues to hold a unique and valuable asset in the industry.

Bard even hallucinated when I asked about Youtube Data API v3.

If you can't properly answer about your own API's what good are you ?

>That said, as a user, the moat between ChatGPT and say, Bard seems as wide as the Pacific.

From my observation bard is better at sql than chatgpt

It is hard for me to compare two black boxes. I agree that ChatGPT looks leaps and bounds better than Bard, but it’s hard for me to tell the power of the core transformer model ends and the heuristics and secondary models begin.

Knowing what those little components look like, for me, would help me evaluate which model is actually better. Right now it feels to me that the difference between the two comes down to more copyrighted materials in ChatGPT’s training set… and more sophisticated ways to hide that, but that’s just my hunch.

“That fundamental formula has not really changed much for years,” says Jakub Pachocki, one of GPT-4’s developers. “But it’s still like building a spaceship, where you need to get all these little components right and make sure none of it breaks.”

https://www.technologyreview.com/2023/03/14/1069823/gpt-4-is...