Ask HN: AI, obsolete before you begin? | Hacker News Reader

Ask HN: AI, obsolete before you begin?

2 points13years3 comments

Looking for some perspectives pertaining to the struggles of rapid advancement in the AI space:

1) How many have attempted to start some project/business in the AI space and ultimately found it eclipsed by something else before you could finish it?

2) How many are currently hesitant to start a new venture due to the concern of #1? You may invest a lot of time, money and resources into a project only to find it becomes irrelevant when the next new capability of some AI model is released.

3) Are these currently topics of conversations for any companies that you work for? Are businesses struggling with these questions as well?

4) What have you learned from any of the challenges above?

View on HN

When OpenAI releases training for GPT that is probably going to make a lot of developers redo a big part of their system because it will likely perform much better. People like me who are currently using vector search or other techniques to substitute for training.

But it's not entirely starting over because you have to start to create your training dataset anyway, at least as far as recording the core information.

Isn't this a common problem for building a solution for "now"?

Isn't the goal to use your vision and think about what the world is going to be like in 2 to 5 years and focus on building that?

Those are the products/solutions/businesses that generally win...not people jumping on the hot new thing...

> Isn't the goal to use your vision and think about what the world is going to be like in 2 to 5 years and focus on building that?

No, this is precisely the problem. It is the planning problem. Unpredictable technological acceleration eviscerates your ability to plan and have a functional business on the longer timeline.

The hot new thing we are accustomed to failing is because of a product that was all hype and never delivered. AI certainly has hype, but there are real products delivering real value being eclipsed faster than ever before.

This is precisely what AI is designed to do. It is the goal. If it is not replacing everything at faster and faster pace, then it is failing at the intention of the builders. AI is not about stability, it is constant accelerating disruption.

Reply on news.ycombinator.com