Artificial Intelligence, Digital Mining & The Need for More Compute


Date Posted:

When Bitcoin mining first came onto the scene, it was a very niche industry that only few seemed to understand, let alone appreciate. Dismissed by many as a passing fad, Bitcoin’s continued market volatility has only added to the skepticism shared by experts and amateurs alike.

Today, however, cryptocurrency mining is proving its value, but in a different way than perhaps we might have imagined: by leveraging its computing power to support the exponential rise of artificial intelligence (AI) and machine learning.

The Need for Computing Power

As artificial intelligence continues its infiltration into nearly every industry and sector, there is an increasingly urgent need for compute resources to power all this machine intelligence.

According to Stanford University’s AI Index, the computing power required by AI has been doubling roughly every three and a half months since 2012 – increasingly 300,000 times between 2012 and 2018 alone. That’s faster than the rate at which computing power historically increased, the phenomenon known as “Moore’s Law” (named after Gordon Moore, cofounder of Intel). Computing power itself is what has largely driven the development in various technological landscapes, such as AI, graphics computing, and 5G.

Consider, for example, the amount of data required for an AI system to recognize someone’s voice or identify an animal or a human being. In order to reach an adequate level of precision, it first needs to process millions of audio, video, or image samples. It then learns to differentiate between two different pitches of voices or to differentiate faces based on various facial features.

But this is only possible by having powerful computers that can process millions of data points every single second. The more the computer power, the faster data can be fed to train the AI system, resulting in a shorter span for the AI to reach near human-like intelligence.

Similarly, AI is being applied to practical problems across multiple industries, such as medical diagnoses, financial forecasting, customer profiling, and geospatial mapping. It’s also being used for fraud detection, research automation, and compliance & security measures, among others. In order for all these systems to achieve even higher levels of artificial intelligence, a tremendous amount of compute resources is required.

And with greater compute power comes greater costs. Training a model like ChatGPT, for instance, costs more than $5 million. Even running the early demo before its current usage level cost OpenAI around $100,000 per day.

From Mining to Machine Learning

The growing need for compute power is precisely why many are turning to crypto mining infrastructure to help push the AI revolution forward. In fact, Mining firm Hut 8 has led the way, leveraging its formerly mining-dedicated compute facilities for machine learning and other HPC (high-performance computing) applications.

Similar to AI, crypto mining relies heavily on computation power to solve complex mathematical problems and validate transactions on blockchain networks. The more computing power miners have, the more likely they are to solve the calculations and earn the bitcoin mining reward.

While the specific computing hardware required for high-performance computing (HPC) or AI processing differs from mining, the biggest challenge is often in setting up the necessary physical infrastructure, such as power, cooling, and security systems. These aspects remain relatively similar, whether hosting RAM-light GPUs (graphics processing units) appropriate for ETH mining or RAM-heavy GPUs appropriate for AI model learning.

In fact, much of AI can’t rely on CPUs – central processing units – due to their inability to support parallel processing. While CPUs are good for general purpose applications, they’ve struggled to keep up with the massive and complex computations required for deep-learning models and AI operations. Furthermore, while CPUs can only process tasks in a sequential manner, GPUs can multitask, running many smaller tasks at once with greater speed and efficiency.

Consider, for instance, an autonomous vehicle: the system needs to collect and understand data from sensors and GPS and recognize the images captured by its cameras, which requires sophisticated machine learning capability. This becomes especially vital when facing an unexpected situation and having to make the right split-second decision. A high-speed, low-latency HPC-powered 5G signal makes it possible to offload that complex decision making to a powerful computer connected to the vehicle.

Similarly, many of our mobile applications depend on legions of computers to store trillions of data and perform split-second operations, such as calculating travel time based on distance and traffic volume. As other types of advanced AI become more commonplace and are utilized by businesses and consumers, the challenge will be to maintain sufficient compute capacity to support them.

The need for computing power and more efficient networking infrastructure to process, analyze, and transfer vast amounts of data is only growing. In fact, the high-performance computing market itself is expected to grow from $36 billion in 2022 to $49.9 billion by 2027, a compound annual growth rate of 6.7 percent.

Leveraging the computing power and infrastructure of cryptocurrency mining operations for AI, machine learning, and other HPC applications might help meet this demand – and is something to pay attention to in the years to come.