The holy GraphGrail of AI: a review
A blockchain startup with big-but-focused claims
The GraphGrail team, a small Russian blockchain startup, promises to make ambitious play to offer an open-access AI platform. recently reached out to me requesting a review of their concept and plan. They hit all the right points of intrigue for a futurist-humanist to take a look.
I want AI to be decentralized, democratized, openly accessible for a collective human intelligence to experiment and simulate futures.
With this bias, this hope for such AI futures to play out — this is my attempt at an unbiased review of their whitepaper. You might want to scan the litepaper for a quick take of the GraphGrail mission and plan. The following is a collection of my notes and conclusions/reactions from reviewing the paper:
First, GraphGrail starts from a believable place.
Natural Language Processing (NLP) and “text data analysis”. They’re not taking on some experimental application of Artificial Intelligence at the start. NLP is a common focus in the field and has made promising strides in the past decade. This tells me they have a practical mindset in the over-hyped and disparate field of AI startups.
They have designated the “GraphGrail AI Lab” as an open experimentation program to satisfy the natural push for what I see as AI science. Science can broadly advance with digital simulation and machine learning loops. AI can experiment and test without affecting the physical world, without relegating scientists to endlessly hunch over test tubes and terminals.
They also acknowledge and address the common problem of “data tagging” — that is, how does one sort all the tremendous data sets involved in data science modeling. This kind of data refinery is essential to making a platform usable.
Open-access means tools that let anyone use the platform
GraphGrail has plans for this. The AI Designer tool claims you don’t need to code to interact.
They go on to break down the actual methods and list some of their primary algorithms — a refreshing level of detail and implementation strategy I look for in any legitimate whitepaper. Like a sales pitch, any ambiguity and distraction at this stage is a sign of undeveloped and incomplete considerations, i.e. not being ready or focused enough to launch.
As a (potentially) foundational AI platform, they’re taking the marketplace route. How the AI computing infrastructure will scale as more applications request more work.
The GAI token runs on the Ethereum network to allow for “creating, improving and voting for language models.” I see issues with moving any kind of heavy transaction volume on Ethereum network. Layers like the Raiden Network have made Ethereum much more robust by taking actions “off-chain” — but 3rd and 4th gen blockchains address this without the layers from the lessons of BitCoin and Ethereum. Chains like EOS are still unproven, as is most of the scaling solutions out there — unproven “at scale” that is, when big and serious data and transaction volumes start moving.
The early days of neo-AI (not your grandfather’s sci-fi AI that suffered through winters and abandonment) is upon us. It’s what makes all of the new startups in the space exciting. When someone cuts through and avoids too much hype or ambiguous utopian vision — it shows in what they’re promising to deliver, especially if they already built something and made money.
How do they (GraphGrail)make money?
Their marketplace will take a share, their smart contract assistance service (for funding, accounting, business deals) will have transactions fees.
Again, they’re very open and refreshingly transparent — it extends to a breakdown of their current customers. GG has worked with city governments in Russia and mid-size tech companies. They’ve built or supported everything from analytics and speech recognition to chatbot development. Everything is an immediate market need, which says good things about their focus on real revenue in the business model and their ability to consult and deliver to keep making money as they build a common platform. I view this as an essential balance for any B2B tech startup — blockchain and AI or not.
The grasp on AI market trends seems solid. Big Data and Cognitive Systems are addressed broadly, and then segmented to show fit and product roadmap connections. If I was a VC or Angel investor, this part of the plan would be overkill — something for a junior associate to verify or a sign of unnecessary market research and possibly a too-careful or conservative approach from the startup team.
The GraphGrail ICO may not meet its goals. I like the real-revenue and payments/business they’ve already generated. It’s a good sign they can build something even if they don’t raise much.
If the GG team can deliver (always the challenge for any tech startup) on even 25% of what I’ve distilled from the whitepaper, it will be a big step towards democratized AI interaction — and a future I can endorse.