Except the "something" isn't building a firm, it is the tasks done by that firm. I don't have software that acts as an accounting firm in a box (though that may happen in a few years). I have tools to automate things we do day after day for every client.

And I thought the results I was getting were bad...

The issues with that logic became apparent eventually. A couple years after I graduated, we had a boom in software engineering demand greater than the dotcom bubble.

I bring this up for no reason whatsoever. There's certainly no contemporary analogy.

For sure. I've felt like the old man yelling at clouds for trying to find libraries with few dependencies.

That sounds good, but I think it is easier said than done. Looking at the past is easy. It happened so the events were guaranteed. The cause less so, but there's still comfort in that certainty.

That being said, the "It's never happened before" mentality has led to more than a few crises...

Curious to hear if there are any decent ideas on a resolution to that.

I'm sympathetic to how hard it can be to generate new business. I'm living that life.

These tactics are counter productive though. I don't want engineering candidates associating me with a bait and switch.

I'd like to think that all the people trying to opt out of Copilot would be sending a good enough message that Microsoft should stop...

but I'm guessing too many decision makers have their compensation tied to Copilot adoption.

Maybe we can get an LLM to do an analysis there...

I'd like to say I'm surprised, but I'm not.

It is curious at how much of a difference it makes though. All the emails have your name first as the sender. I feel like the subject line would matter more. Maybe men write titles that attract other men better.

What's the idea?

I'm some rando on the internet though. People are scared because they're listening to Sam Altman and Elon Musk, who are more credible.

They'd be less scared if they also balanced those views with Gary Marcus or Melanie Mitchell.
garymarcus.substack.com/p/breaking-n...

That's why I don't think we're anywhere close to true AGI, which is effectively human intelligence and where the definition has been muddied recently.

How can we build an algorithm to replicate something we don't understand? It'd be like trying to build an airplane without understanding gravity or air resistance.

It has been debated for decades of whether human intelligence can be boiled down to patterns and whether ML is enough. The main issue I think is the fact that we as a society have only scratched the surface of how the human brain works. We don't understand all the fundamentals for intelligence.

ML algorithms are fundamentally based on pattern matching. ChatGPT is a more advanced version of what Watson could do, but it runs on a lot of the same principles as Watson. Watson also generated a lot of hype in 2011, but failed to live up to that hype.

The best analogy I can think of is comparing a car to a spaceship. You can iterate and make a car faster, but it will never be a spaceship no matter how fast you make it. Going into space requires a different set of fundamentals to account for. Similarly, ML will hit a ceiling on performance.

The planning pieces, which include most product management tasks, involve a lot more critical thinking which current AI has not proven to do well.

Will it improve to do better? Sure. But there comes a point where the fundamentals need to change.

I think we're about to start delving into the AGI debate here. A lot of coding tasks are repetitive. The hard part is the planning, but once the plan is made the coding kind of just falls into place. That makes it easy for most ML algorithms, which include LLMs.