Decentralisation: More Advances in artificial intelligence (AI) Are Being Made Than in Hardware.

Can the Gap Be Closed via Decentralisation?

 

 

Decentralisation of AI computing power

The Rapid Rise of Artificial Intelligence (AI) Capabilities

 

Over the past two years, artificial intelligence (AI) capabilities have skyrocketed. Large language models (LLMs) like ChatGPT, Dall-E, and Midjourney have become indispensable tools. Generative AI programs now write emails, create graphics from simple prompts, compose music, and develop marketing copy, all in real time as you read this post.

The speed at which individuals and businesses adopt AI is even more astonishing. According to a recent McKinsey poll, the percentage of businesses using generative AI in at least one function rose from 33% at the start of 2023 to 65% in less than a year.

However, like any new technology, AI faces its own set of challenges. One major problem is AI centralisation, which demands immense resources to train and run AI systems. Currently, big tech companies seem to dominate this space.

Decentralisation of AI computing power

The Growing Computational Barrier

The World Economic Forum reports that the demand for AI compute is growing rapidly. The computing power needed for AI development increases annually by 26% to 36%. This growth trajectory matches findings from Epoch AI, which predicts that training and running AI algorithms will soon cost billions.

Ben Cottier, a researcher at Epoch AI, explains, “The cost of the largest AI training runs has doubled or tripled every year since 2016. By 2027, we expect billion-dollar price tags, maybe even sooner.”

It seems we have already reached this stage. Last year, Microsoft invested $10 billion in OpenAI. More recently, rumours suggest they plan to build a data centre housing a supercomputer powered by millions of specialised processors at a staggering $100 billion price tag, ten times the initial investment.

Other tech giants such as Google, Alphabet, and Nvidia are also heavily investing in AI computing capabilities. These companies are engaged in a fierce arms race, pumping billions into AI research and development.

Big Tech’s Dominance and Its Risks

While these massive investments promise significant AI advancements, the downside is that AI development remains a “big tech” playground. Only a handful of companies can afford to pour tens or hundreds of billions into AI projects.

This concentration raises concerns. What can be done to avoid repeating the mistakes of Web 2, where a few firms controlled innovation and shaped the internet’s future?

James Landay, Faculty Director of Research at Stanford’s Human-Centred AI Institute, warns that competition for GPU resources and internal use by big tech will raise the cost of computing power. This may push the industry to develop cheaper hardware alternatives.

In response, China is stepping in. After chip export restrictions hindered access to vital processors, the Chinese government now supports AI startups with incentives. Local authorities recently promised AI firms computing vouchers worth $140,000 to $280,000. The goal is clear: lower the price of AI processing power.

Decentralisation of AI computing power

Decentralising AI Computing Power

One thing is clear: the AI computing sector is currently highly centralised. Most processing power and AI software belong to a few tech giants. This monopoly limits innovation and keeps costs high.

Thankfully, decentralised computing infrastructures offer hope. The Qubic Layer 1 blockchain is one such example. It uses an advanced proof-of-work mechanism called uPoW. Unlike Bitcoin’s energy-intensive mining, uPoW harnesses computational resources to train neural networks and perform other valuable AI tasks.

In simple terms, Qubic decentralises AI compute by shifting away from hardware owned or rented by big tech. Instead, it taps into its network of thousands of miners who contribute computational power.

Although decentralisation is more complex than relying on centralised providers, it offers clear benefits. It lowers costs and spreads AI innovation across more participants, reducing big tech’s grip on this crucial technology.

The Importance of Decentralisation for Trust and Innovation

What happens if all major tech companies fail or act against the public interest? This question worries many experts. Tech giants have already proven unreliable in handling life-changing innovations. Issues like data privacy violations and societal manipulation erode trust.

Decentralised AI development offers more transparency. It makes it easier to monitor advancements while lowering entry barriers. This democratisation could lead to fairer innovation and less reliance on a few powerful players.

Decentralisation of AI computing power

Bridging the Gap Through Decentralisation

AI innovations are only beginning, but access to computational power remains a significant hurdle. Currently, big tech controls most resources. This control slows innovation and risks concentrating power over our most valuable digital assets: our data.

However, decentralised infrastructures like Qubic present a promising solution. They can reduce computational costs and break big tech’s monopoly on AI technology. This change would benefit the entire AI ecosystem, encouraging wider participation and accelerating progress in the 21st century’s most important technological frontier.

 

“decentralisation of AI computing power”