Hacker News Reader

Comment by tictacttoe | Hacker News Reader

I’m using AI in my daily life. ChatGPT has largely replaced Google. My programming output and ability is greatly enhanced. I’m working on projects that were impossible a few years ago. How can people say this is another hype bubble?

View on HN

Some people say that it has greatly enhanced their abilities, other people say that it's mostly useless. I am myself in the latter camp. I never build the habit of using it because it failed miserably on relatively simple things. However, even if it was smarter, I can hardly see how it could be faster to design the prompt, correct it multiple times vs. just typing code in flow state.

What about things like copilot that are simply predictive autocomplete? It doesn’t pull you out of flow state but dramatically increases typing speed, especially for boilerplate.

It absolutely kills flow state for me personally because I’m being shown code completions that may or may not have anything to do with what I am actually about to write.

When I allow myself to review the relevance and correctness of the attempted completion, I have to set aside the actual code that I was envisioning. This may or may not be easily recoverable after the (generally wrong) suggested completion is dismissed.

I am not neurotypical, and we are all different anyway, so it is obviously different for others. But I disabled the active code assistance option at work almost immediately due to it being actively harmful to maintaining flow state.

Investors aren't hyped about AI being a nice quality-of-life tool for the general populace. They are investing on the notion that it will eventually replace a lot of employees and make the requirement of expensive skilled labor obsolete.

Was the internet a revolutionary technology? Was there a very real dot-com bubble?

To expand on your point:

My lesson from the dot com boom is that's it's very hard to pick the winners. For every Amazon, there were a thousand pets.com. Investing in DEC, Sun, Yahoo, and friends would not have paid off.

That's a broader lesson too. I have an okay -- far from perfect but decent -- track record knowing which national economies will grow, but investing in index funds for those economies rarely yields good returns. I have a tiny portion of my portfolio in various developing markets, and the GDP can rise tenfold with no investment gains.

Many very real revolutions have bubbles. Bubbles are the period when people over-invest.

Because some other people are operating under the assumption that this wave of AI models will automate all labor on the planet, instead of making the average desk jockey 10-25 percent more efficient.

A 25 percent efficiency boost to desk jockey labor is a profound impact.

> A 25 percent efficiency boost to desk jockey labor is a profound impact.

In terms of 25 productivity, yeah, I think that is a good thing and could be true. However, the problem is that it is sold under the pretense that you can fire a large chunk of workers and rely in large part on AI, which makes it a whole lot more attractive to investors.

That is precisely what you can do if your workers are 25% more efficient.

Yes just a matter of recognizing it’s 25% of various job titles, not 100% of 25% of job titles

Or alternatively, drastically lower the skill requirement for the job and replace your skilled workers with cheap unskilled workers

> That is precisely what you can do if your workers are 25% more efficient.

If AI is just a tool that improves performance by 25% it changes what investors believe. It is a lot more expensive than 25% increase in productivity as investors were sold a different reality.

> Or alternatively, drastically lower the skill requirement for the job and replace your skilled workers with cheap unskilled workers

With the error rate of GenAI I'm curious to see how this plays out in the long run. It seems that only more skilled workers can pick up on the wrong answers.

Shrug. I’m not the one claiming 25% is a good figure. Just that it’s a huge deal, if true.

An impact on what? Most desk jockey labor hasn't been optimizing for efficiency for decades. I'd say the amount of "work" will simply go up by 25%, and the position of the human-based processing nodes will shift a little bit. But it's a Schelling point attractor that mutually-antagonistic companies will be pushed into adopting. You certainly won't want to hire 25% more human employees at your own company to handle your counterparties' increased paperwork generation from their use of AI.

For who ?

Nearly all the wealth generated by productivity gains of the last 50 years has been pocketed by business owners and investors

https://assets.weforum.org/editor/HFNnYrqruqvI_-Skg2C7ZYjdcX...

Feel like you’re answering your own question here

I find it insanely useful for moderately basic tasks I don't do often e.g. "write me a regex that will match XYZ", or basic latex formatting. Much more efficient than googling for docs!

Whilst the benefits of AI are real, the public's expectations exceed reality.

In other words, a bubble occurs when expectations >> reality.

Reply on news.ycombinator.com