Hacker News new | ask | show | jobs
by l3x4ur1n 81 days ago
I agree with you. I am personally building small tools and apps to solve my own problems and I would not do that before AI - I would not know how.
1 comments

Would you pay $1000 per month for that?
I have never come close to running into any limits using codex with my $20/momth ChatGPT subscription every day
OpenAI is losing massive amounts of money and it needs to pay back hundreds of billions. What do you think will happen to that price 5-10 years from now?
A) Whatever company I’m working for then will pay for it. Right now we get a $1000 a month budget a piece on Claude. I just hardly ever use it.

B) price of computing always comes down and models will be a commodity

C) If I am an independent consultant by then, I’ll just pay for local inference (I work full time for a consulting company now). You can already get a decent local inference Mac for $500/month.

> B) price of computing always comes down and models will be a commodity

Past results are not an indication of future performance. Also funny to say this when hardware costs have spiked.

> Past results are not an indication of future performance.

Yes. Sun setting and rising tomorrow like it did today is not given either, despite it having doing so without an error every single day for the past 4 billion years.

I have been buying computers myself for over 30 years and before that my parents bought my first two computers in 1986 and 1992 (an Apple //e and then a Mac LCII).

I’ve seen RAM and hard drive spikes plenty of times.

Do you really think the cost of compute is not ever going down? BTW, my second computer was $4000 in 1992 for the full setup - Mac LC II with 10 MB RAM (actually 12 with 10 usable), monitor, Apple //e card, 5-1/4 inch drive for the card, laser printer and SoftPC. Thats over $9000 in today’s dollars.

I made more than that on a three week consulting contract I did when I was between jobs. I wouldn’t hesitate to spend around $8000 on equipment if I went independent.

> What do you think will happen to that price 5-10 years from now?

Who knows? Unless AI hits a brick wall in a year, 5-10 years from now is effectively singularity (in the original sense that things start moving so fast it's no longer possible to make even high-level predictions on what the world will look like).

2 years from now? Open models will be around the level of current commercial SOTA today, so at worst, they'll be able to continue doing what they do for the price of electricity used by inference.

> Who knows? Unless AI hits a brick wall in a year, 5-10 years from now is effectively singularity (in the original sense that things start moving so fast it's no longer possible to make even high-level predictions on what the world will look like).

I appreciate the optimism. I'd put the singularity closer to cold fusion and flying cars than to longer range EVs.