Would like to add some of my experiences and observations. Some of my perspectives are based on enterprise side of systems, both blockchains and off-chain data assets in regulated environments, and with all flavors of public blockchains, private and consortium blockchains.
It is actually a very interesting self-enhancing recursive loop of symbiosis between blockchain and AI.
It is NOT just “ the more data the more accurate and precise AI results are”. Under the hood, there is more than that, both in technology as well as adoption at scale for business operations. Also, yes the public data itself is a treasure cove, but the advances emerging to also deal with private data can take it further, for how blockchain and AI can feed on each other.
From my experience, the hyper-productive efficiencies we were targeting actually came much higher than expected, with a combination of cognitive AI, business knowledge heuristics, concepts emerging from the blockchain area around provenance, and a fine-tuned agreement for AI – human SME loop. However, what was rather interesting was that it completely corroborated with high degree of precision where the AI results were trustworthy in business reliable data, and NOT so much in less reliable base data.
A whole bunch of time in enterprises go on addressing key data quality issues, that feeds their AI models, worse-yet, these cleanup are done unilaterally, sometimes without the full context of the parties they work with.
Step 1 - Where blockchains help is in the feeder data to the AI., where some amount of data quality rigor comes up automatically, and more importantly the truth could be in the context of the party that provides the data, or an in-direction to the data on-chain ( in privacy sensitive flows).
Step 2 - With this increased data quality, AI solutions greatly improve their results, and more importantly, if the AI models can keep the explain-ability high ( which is going to be a problem given how fast the field is growing).
Step 3 - That is where blockchain can help again, with detailed provenances to the results, why- what is prescribed, and the level of aggregations that are duly interesting to the counter-parties, and blockchain based analytics, both what was put on-chain, off-chain, specifically indicating the need for areas of focus on more collaborative data convergence working with ecosystem partners.
Step 4 – With the increased focus comes the multi-party protocols which demand the further convergence, and it will have a gradient of friction, some low, medium and high. I see a day soon where insights shared , some might be based on LLMs where enterprises were able to collect that kind of a corpus, and also be able to share that with even tinyML, to address the asymmetry and to see if the models are workable with that asymmetry, and further tune them. These flows will rely heavily on privacy enhancements, specifically light weight ZKPs, that helps converge on decisions.
That is the fine symbiosis loop,
There has been a clear challenge of blockchain adoption albeit the amazing work that has evolved in short time from the industry, particularly on the public blockchain side and the openAI /LLM side. In my view, the concepts above and the tools / techniques that can promote these will drive the high chance of driving adoption and much synergized results as it puts, particularly as the line between public blockchains and enterprises are blurring, where there is a lot of interest from enterprise but also a lot of caution.
I am in the process of writing more on these concepts and the key inflection points observed, and what it takes to get it to better adoption in various industries.
Good observations.
Would like to add some of my experiences and observations. Some of my perspectives are based on enterprise side of systems, both blockchains and off-chain data assets in regulated environments, and with all flavors of public blockchains, private and consortium blockchains.
It is actually a very interesting self-enhancing recursive loop of symbiosis between blockchain and AI.
It is NOT just “ the more data the more accurate and precise AI results are”. Under the hood, there is more than that, both in technology as well as adoption at scale for business operations. Also, yes the public data itself is a treasure cove, but the advances emerging to also deal with private data can take it further, for how blockchain and AI can feed on each other.
From my experience, the hyper-productive efficiencies we were targeting actually came much higher than expected, with a combination of cognitive AI, business knowledge heuristics, concepts emerging from the blockchain area around provenance, and a fine-tuned agreement for AI – human SME loop. However, what was rather interesting was that it completely corroborated with high degree of precision where the AI results were trustworthy in business reliable data, and NOT so much in less reliable base data.
A whole bunch of time in enterprises go on addressing key data quality issues, that feeds their AI models, worse-yet, these cleanup are done unilaterally, sometimes without the full context of the parties they work with.
Step 1 - Where blockchains help is in the feeder data to the AI., where some amount of data quality rigor comes up automatically, and more importantly the truth could be in the context of the party that provides the data, or an in-direction to the data on-chain ( in privacy sensitive flows).
Step 2 - With this increased data quality, AI solutions greatly improve their results, and more importantly, if the AI models can keep the explain-ability high ( which is going to be a problem given how fast the field is growing).
Step 3 - That is where blockchain can help again, with detailed provenances to the results, why- what is prescribed, and the level of aggregations that are duly interesting to the counter-parties, and blockchain based analytics, both what was put on-chain, off-chain, specifically indicating the need for areas of focus on more collaborative data convergence working with ecosystem partners.
Step 4 – With the increased focus comes the multi-party protocols which demand the further convergence, and it will have a gradient of friction, some low, medium and high. I see a day soon where insights shared , some might be based on LLMs where enterprises were able to collect that kind of a corpus, and also be able to share that with even tinyML, to address the asymmetry and to see if the models are workable with that asymmetry, and further tune them. These flows will rely heavily on privacy enhancements, specifically light weight ZKPs, that helps converge on decisions.
That is the fine symbiosis loop,
There has been a clear challenge of blockchain adoption albeit the amazing work that has evolved in short time from the industry, particularly on the public blockchain side and the openAI /LLM side. In my view, the concepts above and the tools / techniques that can promote these will drive the high chance of driving adoption and much synergized results as it puts, particularly as the line between public blockchains and enterprises are blurring, where there is a lot of interest from enterprise but also a lot of caution.
I am in the process of writing more on these concepts and the key inflection points observed, and what it takes to get it to better adoption in various industries.
Regards,
Sathya Krishnasamy
linkedin.com/in/sathya-krishnasamy...