Does the ever-growing artificial intelligence sector need the services of blockchain technology?

The phenomenon of artificial intelligence (AI) is growing at an exponential pace. Although the
underlying technology first caught the attention of masses in 1997, whereby an IBM built machine known as “Deep Blue” outwitted the then World Chess Champion Garry Kasparov, it is only now that researchers are appreciating the practically endless capabilities of what AI can offer.

In fact, according to a recent study, it is estimated than the entire AI sector will be worth close to $1.2 trillion by the end of 2018, up 70% from the previous year. Furthermore, the same study concluded that this figure will grow to just under $4 trillion at the close of 2022. However, whilst demand for AI-driven technology continues to prosper across a magnitude of sectors – whether that’s FinTech, Medicine, Marketing or Defence, there remains a somewhat significant challenge that must be resolved: Computing Processing Unit (CPU) data centers do not have the required computational power and/or hardware capacity to facilitate global requirements of the future.

The Future of Artificial Intelligence

The current state of play in the AI sector places a strong empathis on data centers, which subsequently provide the necessary capacity to facilitate the technology. On top of the aforementioned concerns regarding computational power, these data centers are ultra-energy intensive, with the Office of Energy Efficiency and Renewable Energy estimating that data centers are responsible for close to 2% of all U.S. electricity consumption. As one would expect, this is far from a sustainable model for a technology that many believe will soon play a major role at every corner of the globe. However, there might just be a solution.

Instead of amplifying the problem by placing a heavy reliance on CPU’s, the answer is potentially staring straight at us. With an estimated 2 billion consumer computers now in circulation, the required capacity levels can instead be shared across multiple graphic processing unit (GPU) devices.

Not only do GPU units have the ability to perform faster calculations and facilitate higher levels of memory bandwidth, but they are also 10 times more efficient than their CPU counterparts. As one would imagine, the contribution of surplus computational power does come at a cost, however a new and exciting start-up called Tatau believe they have the solution – blockchain technology.

How blockchain technology can bridge the artificial intelligence gap

In a nutshell, Tatau have created a decentralized platform that will allow those in need of computational power and storage to purchase it, and those that have surplus quantities, to sell it. In order to create an autonomous and seamless marketplace, the platform utilizes a tokenization eco-system that is built on top of the blockchain protocol. Not only does this ensure transactional activity remains cheap, fast and secure, but the entire platform is transparently accessible via the blockchain ledger.

To illustrate the demand for such a solution from within the AI industry, the Tatau team recently entered into a partnership with FaceMe – an innovative New Zealand based organization that specialize in bridging the emotional gap between digital and human interaction. The concept – which is suitably named an Intelligent Digital Human Platform, allows humans to interact with an AI-based machine as a means of receiving customer support. Whether it’s in the banking, retail, educational or government sectors, a digital interaction through the FaceMe technology gives customers that personal touch.

The partnership will see the FaceMe platform facilitate its AI-driven computational power through the Tatau hub, which in turn, is fully supported by those that contribute their surplus GPU capabilities. Ultimately, it is a solution that stands to benefit all involved. Whilst the likes of FaceMe are able to offset their demand for power, GPU owners are able to monetize through a tokenization model that is based on cryptocurrency and blockchain technology.

This great solution also led to Tatau’s back-to-back wins on both the Regional Startup World Cup in Dubai, and the World Blockchain Summit’s “Grand Slam” pitch award. Overall, Tatau beat a field of 33 tech companies in both competitions and were picked winners.

Tatau’s team with the judges and organizers. Image: Tatau

The FaceMe-Tatau partnership is certainly a step in the right direction for an industry solution that has the potential to go global, and could be the first of many. In fact, when one considers the current size of the AI sector, alongside its estimated growth over the course of the next few years, we could see a sea-change in the way organizations facilitate their AI computational demands.

9 thoughts on “Does the ever-growing artificial intelligence sector need the services of blockchain technology?”

  1. That’s why a lot of people look at the blockchain as a solution in search of a problem. Why go to the blockchain? The only answer I can think of is because of a personal deep dislike and distrust of centralized computing. That’s it.The same can almost be said for Bitcoin. Why would a person use bitcoin? Because they dislike the concept of using a fiat currency, or they think fiat currency is going to become worthless one day and Bitcoin won’t.All of that being said I do think some people are working on cool AI networks on the blockchain, like Singularitynet. They let people create single purpose AI’s to solve certain problems and they can sell their AI to others to put to use. But while that idea is interesting, I don’t see a reason why they can’t do this using a centralized server.

  2. The idea of such projects is to reuse the enormous mountain of hardware used for crypto mining. AI computing pays off many times better than mining even if price is set to 25% of AWS and similar.Quality of service will typically be worse than for dedicated datacenters but for many clients, this is not as important as the cost.What many of these projects are now discovering is that the mining hardware is too limited and need to be upgraded at high costs. A lot of GPUs are not enough. CPU power, memory and storage must be of decent capacity too.

  3. Tatau is one project out of many related, all trying to provide distributed computing combined with blockchain. Here is a little list of some of the most known. I will try and modify the links to make them survive.SONM: sonm:comVectordash: vectordash:comNBAI: nbai:ioSyncAI: syncai:net/pdf/syncai-wp:pdfLeonardo: leonardorender:comEffect.AI: effect:aiDML : decentralizedml:comTatau : tatau:ioMassGrid: massgrid:comNeuromation: neuromation:ioOCEAN: oceanprotocol:com DBC : deepbrainchain:org/index:htmlOf these, I think DBC (DeepBrainChain) may become one of the survivors. They have a solid track record in the AI business and they have addressed some important aspects of providing a quality distributed service. None of these are really up and running yet. Most are in the testnet phase.SONM, which is a general compute platform has probably come furthest to date. However, they are not specialized in AI computing and their network can’t provide the kind of hardware needed for advanced AI training. Not yet anyway.For anyone looking for cost effective compute power, it may pay off to approach some of these projects. Cost is typically 10 – 25% of AWS or similar cloud services.

  4. So… this is a gambit to sell unused computing power (invested in by the BlockChain advocates) … to the AI community in a way which is fast, secure and cheap? Well, what’s not to like about that. Excepting of course for the fact that the price won’t be cheap. With few exceptions, high-profit motivated computer timesharing services are quite expensive.Also, please let’s NOT call “Deep Blue” IBM’s entrée into artificial intelligence computing. It most certainly was not. It was, in fact, an enormously parallelized relatively compact constellation of algorithms to evaluate billions of board-positions hypothetically, and from amongst them choose the relatively few that both “worked” and showed promise of strategically winning the games it was playing.Its like cryptocurrency mining. Ultimately every ‘bitcoin’ (or other unit of cryptocurrency) is a unique pair or triplet of very long prime numbers, which have the unique ability to multiply-and-divide the “standard test equation” and yield. a product with a looooonng string of zeroes in it. That meets the definition of a cryptocurrency “found token”. Since there are no theories as to how to NOT-find these by brute force, well … brute force is used. Rubber stamp the algorithm a million times for the million execution units on the 2,500 NVidea GPU chips (each having 4,000 compute units), on 300 servers, in 15 racks warehoused in Iceland, cooled by the lovely glacial catabatic winds, powered by geothermal vent generators. Wicked good. Just have to do a LOT of brute force calculations. To mine the coins. Still, if that — if millions of GPU microcores — is what your AI research and development needs, and your investors can hardly conjure up the money to actually buy the 300 servers, the 1200 cards, and so on, especially since the actual AI analytical run-jobs aren’t 24 hours a day, 7 days a week, then purchasing unusued crypto-whatever CPU time doesn’t seem like an imprudent option.Just saying,GoatGuy

  5. Not sure how this can beat retail prices by major providers with internal infrastructure.Google, Amazon and Microsoft have adapted to the cloud, by selling computing services with different pricing, depending on the volume.And they maintain their own computing infrastructure with certain minimal standards of quality. That is, they do their own IT and tools for maintaining your business up and working.Probably this could rather help with the “micro-transaction” kind of computing, with very limited and defined execution times and well defined resource consumption. The computer equivalent of the Mechanical Turk.But for most business, buying computing resources wholesale seems to be cheaper.

  6. The idea of such projects is to reuse the enormous mountain of hardware used for crypto mining. AI computing pays off many times better than mining even if price is set to 25% of AWS and similar.
    Quality of service will typically be worse than for dedicated datacenters but for many clients, this is not as important as the cost.

    What many of these projects are now discovering is that the mining hardware is too limited and need to be upgraded at high costs. A lot of GPUs are not enough. CPU power, memory and storage must be of decent capacity too.

  7. Tatau is one project out of many related, all trying to provide distributed computing combined with blockchain. Here is a little list of some of the most known. I will try and modify the links to make them survive.

    SONM: sonm:com
    Vectordash: vectordash:com
    NBAI: nbai:io
    SyncAI: syncai:net/pdf/syncai-wp:pdf
    Leonardo: leonardorender:com
    Effect.AI: effect:ai
    DML : decentralizedml:com
    Tatau : tatau:io
    MassGrid: massgrid:com
    Neuromation: neuromation:io
    OCEAN: oceanprotocol:com
    DBC : deepbrainchain:org/index:html

    Of these, I think DBC (DeepBrainChain) may become one of the survivors. They have a solid track record in the AI business and they have addressed some important aspects of providing a quality distributed service.

    None of these are really up and running yet. Most are in the testnet phase.
    SONM, which is a general compute platform has probably come furthest to date. However, they are not specialized in AI computing and their network can’t provide the kind of hardware needed for advanced AI training. Not yet anyway.

    For anyone looking for cost effective compute power, it may pay off to approach some of these projects. Cost is typically 10 – 25% of AWS or similar cloud services.

  8. So… this is a gambit to sell unused computing power (invested in by the BlockChain advocates) … to the AI community in a way which is fast, secure and cheap? Well, what’s not to like about that. Excepting of course for the fact that the price won’t be cheap. With few exceptions, high-profit motivated computer timesharing services are quite expensive.

    Also, please let’s NOT call “Deep Blue” IBM’s entrée into artificial intelligence computing. It most certainly was not. It was, in fact, an enormously parallelized relatively compact constellation of algorithms to evaluate billions of board-positions hypothetically, and from amongst them choose the relatively few that both “worked” and showed promise of strategically winning the games it was playing.

    Its like cryptocurrency mining. Ultimately every ‘bitcoin’ (or other unit of cryptocurrency) is a unique pair or triplet of very long prime numbers, which have the unique ability to multiply-and-divide the “standard test equation” and yield. a product with a looooonng string of zeroes in it. That meets the definition of a cryptocurrency “found token”.

    Since there are no theories as to how to NOT-find these by brute force, well … brute force is used. Rubber stamp the algorithm a million times for the million execution units on the 2,500 NVidea GPU chips (each having 4,000 compute units), on 300 servers, in 15 racks warehoused in Iceland, cooled by the lovely glacial catabatic winds, powered by geothermal vent generators. Wicked good. Just have to do a LOT of brute force calculations. To mine the coins.

    Still, if that — if millions of GPU microcores — is what your AI research and development needs, and your investors can hardly conjure up the money to actually buy the 300 servers, the 1200 cards, and so on, especially since the actual AI analytical run-jobs aren’t 24 hours a day, 7 days a week, then purchasing unusued crypto-whatever CPU time doesn’t seem like an imprudent option.

    Just saying,
    GoatGuy

  9. Not sure how this can beat retail prices by major providers with internal infrastructure.

    Google, Amazon and Microsoft have adapted to the cloud, by selling computing services with different pricing, depending on the volume.

    And they maintain their own computing infrastructure with certain minimal standards of quality. That is, they do their own IT and tools for maintaining your business up and working.

    Probably this could rather help with the “micro-transaction” kind of computing, with very limited and defined execution times and well defined resource consumption. The computer equivalent of the Mechanical Turk.

    But for most business, buying computing resources wholesale seems to be cheaper.

Comments are closed.