Can AI 'Bank On' Blockchain To Power Science & Medicine's Future Progress?

The last few decades have witnessed innovations in modern medicine, science and technology at a much faster rate than at any time before in history. A large portion of the credit goes to computers, for helping us solve problems more efficiently and therefore at a faster pace.

Most recently, this has included the rise of Artificial Intelligence (AI), machine learning as well as neural networks that can simulate human thought patterns. They can then apply their more efficient brains to issues that desperately require resolution, many of which are in STEM fields like medicine or cryptography.

STEM stands for science, technology, engineering and mathematics, but a far wider range of academic disciplines fall under this description. Courses one could study range from aerospace engineering and astronomy to civil engineering and statistics.

Following on from various buzzwords like Big Data, Cloud (computing) and Omni-channel in the technology landscape, more recently we have seen the Internet of Things (IoT) and AI surface.

Last year research firm Forrester expected investment in AI to triple in 2017. And, according to Gartner by around 2020, 85% of customer interactions will be managed by AI. Highlighting the value of AI, market intelligence firm Tractica has projected the sector to worth a not inconsiderable $36.8 billion globally by 2025.

The pace of discovery and progress in the genomics field, which focuses on the eradication of cancer and other genetic-level human defects, is now using AI to record how differences in human DNA might manifest in unique corresponding traits. AI thrives in environments where it can scan billions of genes and learn to identify and correctly interpret emergent patterns.

If a connection is made, targeted drugs can be synthesized as a treatment. AI is also making an impact in cancer research where researchers have already created a convolutional neural network capable of distinguishing between dangerous skin marks and harmless ones. When tested against an international panel of expert dermatologists, it accurately detected skin cancer from imagery at a 95% rate compared to the experts’ 87%.

AI is capable of quickly sifting through hundreds of datasets and is many orders of magnitude more efficient than a human brain at doing so. However, it is also more resource-intensive.

But researchers cannot afford to be liberal with their use of computers due to the nearly prohibitive costs of maintaining a centralized source of processing power for the amount of time it takes to map a human genome. Even at roughly one hour. And, for some perspective, just two years ago it took 26 hours – with the cost of outsourcing this power bill being substantial.

The power demand of AI is also too large to be feasible for our current, regardless of its potential applications. And, in a broader sense, this inefficiency puts a ceiling on innovation.

It is also worth bearing in mind that by 2025 the global data sphere has been projected to grow to 163 zettabytes. And, the current computational paradigm is not scalable or intelligent enough to handle this massive influx of information. The zettabyte is a multiple of the unit byte for digital information, equivalent to one sextillion (one long scale trilliard) bytes, so hardly inconsiderable.

Fortunately, blockchain or Distributed Ledger Technologies (DLT) are touted and looking to offer a solution with more effective resource-consolidation and dissemination, which can help AI do more by being a reliable source of power to draw from.

Putting some context on matters, Siim Õunap, COO of blockchain and crypto digital marketing agency Savii Digital, said: “One of the little-known reasons behind the creation of the Bitcoin protocol was to get enough computing power to solve complex mathematical problems that no one computer could by itself. As the process went on, it evolved and virtual currency was born.”

The Estonian, who divides his time between London and Tallinn, added: “Even though computing power was used for something else, it created an ideal solution to combine hardware efforts to power among other things, AI and IoT. Now, almost a decade since it first started, it can be seen being used widely by many companies and I believe that this is just the beginning.”

Raw Compute Power is the ‘New Oil’

Just as the industrial revolution was driven by factory machines and oil, the next revolution will come on the backs of AI and vast sums of hardware and computational power. These last two factors directly influence the intelligence of computers, so it becomes more important to give the world’s growing artificial intelligences a more efficient source.

Processing power – or compute power – could include parallel distributed CPU and GPU processing.

Even purpose-built centralized processing infrastructure like data centers are not efficient enough to meet AI’s growing compiling, rendering, and predictive analytics needs in a cost-effective manner.

Professor Andrew J. Hacker, Harrisburg University, and founder of Thought AI, which earlier this year launched a public mineable blockchain backed by Harrisburg University in Pennsylvania for what was dubbed the “AI Superhighway”, commenting here said: “Just as important as compute power is to AI, so too is both the data we feed it and how we use the results. Ultimately the input of AI is data that through complex algorithms provide connections, patterns and insight that, we hope, provide valuable output.”

But as Professor Hacker pointed out “not all data and insight is created equal”, adding that: “Blockchain technologies hold the promise of adding structure and accountability to AI algorithms and the quality and usefulness of the intelligence they produce.”

Now, for the first time in history, the human species are faced with dealing with a technology as powerful (and growing more powerful every second) as global supercomputing and AI that affect our lives and the world around us in profound ways.

“Security, privacy and quality of the data collected to feed more and more sophisticated algorithms and the transparency and safety of how the results they produce are utilized is paramount,” added Hacker.

In the case of Thought AI, it leverages hybrid data/algorithm superstructures and blockchain to create an intelligence layer on the Internet. As Hacker explained: “Thought uses a 3-layer model to efficiently utilize compute power to create and commoditize transparent, responsive and valuable intelligence algorithms while preserving data privacy and increasing quality.”

Given this reality, several companies have decided that blockchain is the ticket to a new status quo in computational capacity and data processes, and preliminary successes have highlighted what some have described as the startling potential in such an idea.

For a geneticist who needs to map a client’s genome or a 3D animator who needs to render a movie, the only current alternative is to outsource the task to centralized processing farms like Amazon, for instance. These companies use their existing data centers and return a finished product at a high price.

As such, the process is too expensive and too slow to do often, which keeps the technology from being truly feasible currently.

Now, the same geneticist or artist can complete their project less expensively by uploading it to a service like Tatau, a decentralized marketplace of idle GPU power. Tatau matches a project’s estimated processing needs with its connected members who are willing to let the system borrow that much power (and at that price) from their own machines.

The blockchain is responsible for organizing the system of authority that acts as a substitute for the structure of a centralized system.

Andrew Fraser, co-Founder and CEO of TATAU.io, commenting on the landscape said: “Demand for AI computation is doubling every 3.5 months with costs increasing proportionately. Traditional suppliers, such as Amazon and Microsoft, use price as a lever to control usage – this restricts innovation.”

Having completed the concept development last December year, Tatau finalized a $1.5 million seed round of investment this April before completing a proof-of-computation in May. According to its roadmap, it aims to release an AI beta version of the product in January 2019 and start beta processing of commercial AI contracts the following month. Ultimately the goal is to launch the Tatau platform in around June 2019.

Fraser, based in Auckland, New Zealand, added: “We want to unleash AI innovation by dramatically reducing the cost of computing by harnessing the globally distributed GPUs used by crypto miners and making them available for compute by AI companies via our platform.”

Golem Network, a Polish-based peer-to-peer (P2P) network with no central server, is another decentralized power platform that calls itself the “Airbnb for Computers”, though it has different rules that govern its processing co-op. It allows both application owners and individual users (requestors) to rent the resources of other users’ (providers) machines, and be paid in cryptocurrency.

The company, which raised more than $8.6 million back in just a matter of half an hour back in November 2016 via an Initial Coin Offering (ICO), also leverages users’ available processing power into a massive pool that any user can tap into for a variety of projects.

Both Golem and Tatau are based on a token ecosystem. But while Golem uses CPUs, Tatau draws power from GPUs instead for what is held up as “more agile” processing capabilities.

Blockchain Comes Just In Time

Just over the horizon is a time when raw processing power is commoditized, thanks to blockchain. This should lead to a liquid market for this power, which will have dynamics much more favorable to an AI project than today’s centralized market model according to industry protagonists in the space.

These new ideas therefore come at an opportune moment, with humanity right now facing questions about sustainable existence relevant to medicine, technology, mathematics and other relevant disciplines.

These are arenas in which AI is better prepared to fight. Platforms like Tatau, Golem and others for now satisfy the demand for a more economical computational power deal. That said, it remains to be seen if or how the idea will go mainstream.

The other concern is whether it will be enough. This borrows from an idea called Jevon’s Paradox, which states that as it becomes more efficient to use any resource, demand for and consumption of that resource will rise more as a result.

And, if blockchain successfully frees the power trapped in retail-level machines will it simply drive demand up beyond what the blockchain is prepared to produce?

Finally, hardware is a relevant consideration. Blockchain may be efficient at collecting and distributing power. But at some point, new power must come from somewhere. These are questions for tomorrow. However, for now, nothing can distract from the stellar progress coming from blockchain’s ability to boost AI scalability. Carpe diem.

Original Source