Will AI Fail in 2025? Thoughts on Decentralization and Fixing Problems?

Seeking Advice
1

I came across an article by Dr. Max Li talking about how AI could face big issues in 2025 if everything stays too centralized. He suggests decentralizing AI by using blockchain or other distributed networks to reduce risks and make things more transparent.

What do you all think? Does this sound realistic, or is it just hype?

2

Welcome to this forum. Here are some basic posting tips:

  • Make sure your post has enough details to help people respond.
  • Use the search bar to see if your question has already been answered.
  • Feel free to discuss both the pros and cons of AI, but keep it respectful.
  • Always provide links to back up your points.

Thanks for participating, and reach out to the mods if you have questions!

3

I don’t think it’s practical. The cost of training these massive models is way too high for most companies, and we’re already short on GPUs. The whole system is bottlenecked by supply.

4

Jaden said:
I don’t think it’s practical. The cost of training these massive models is way too high for most companies, and we’re already short on GPUs. The whole system is bottlenecked by supply.

Fair point, but wouldn’t a decentralized setup let GPU owners rent out their hardware and make money?

5

@Ash
Not really. Most consumer GPUs aren’t powerful enough for serious AI training. Even for high-end server farms, the demand is way higher than supply, which keeps prices high. That’s why companies like Nvidia and OpenAI dominate the space—they’ve locked in those resources.

6

It depends on costs, priorities, and how safe it is to implement.

7

We’ll probably see both centralized and decentralized AI systems competing. Decentralization makes sense, but we need a cost-effective way to do it first.

8

Fallon said:
We’ll probably see both centralized and decentralized AI systems competing. Decentralization makes sense, but we need a cost-effective way to do it first.

Good point. It does seem like both systems will have their strengths.

9

I wrote about this recently. Here’s a quick summary of steps we can take:

  • Support Open-Source Projects: Like Meta Llama and Hugging Face, which help make AI accessible.
  • Encourage Distributed AI: Using platforms like Ocean Protocol or OpenMined to reduce reliance on central systems.
  • Push for Policy Reforms: Advocate for regulations that prevent monopolies over AI tech and data.
  • Promote Community-Driven AI: Projects like DataKind and Mozilla Foundation work on local AI solutions tailored to specific needs.
10

@Rory
Thanks for sharing! I’ll take a look at those.