Blockchain, AI and Data: Who Wins the Next Tech Wave? COLLIDE 2023
At COLLIDE 2023 in Atlanta, Intuit Factory founder Roop Singh argued that the winners in enterprise technology will combine AI, blockchain and data. He predicted priced data marketplaces and cheaper models that let small firms compete.
By The AI Enterprise Magazine Team4 min read
Key takeaways
- Look at waves together. Singh said winners combine AI, blockchain and data, as film evolved through many technologies.
- Relevant data decides outcomes. Models cannot learn from irrelevant data, and inconsistent semantics hold companies back.
- Data will be priced. He predicted marketplaces in every domain, with blockchain-based attestation of what sellers hold.
- Models are commoditizing. Efficient LLMs at the edge could let small firms look large.
- Trust has a cost. A shared ledger could cut the reconciliation work among business partners.
At COLLIDE 2023 in Atlanta, Roop Singh, Founder of Intuit Factory, argued that the companies that win the next phase of enterprise technology will be those that combine AI, blockchain and data, rather than betting on any one of them. In a session titled "Blockchain, AI, and the Data Revolution: Disruptive Paradigms," Singh described what he called leading-edge thinking on an intersection that, by his account, had little public discussion at the time.
The views below are Singh's, presented in October 2023, and the session was framed as a set of questions rather than answers.
What can the history of movies teach about technology shifts?
Singh opened without mentioning data or AI at all. He walked through the history of film, from silent movies to sound, color, home video, digital cameras, computer-generated graphics, discs and streaming.
His point was that no single technology produced the experience viewers have today. "It's a synchronization of multiple different technologies," he said, and each wave changed distribution and business models. In every stage, he noted, companies that did not adapt were disrupted.
How do AI, blockchain and data work together?
Singh defined blockchain as a secure, immutable, decentralized way to share data, including smart contracts that run automatically when conditions are met. AI, he said, is a set of tools for cognition and decision making, and data is the fuel.
One link is auditability. Blockchain can record the decisions AI systems make in a trustworthy way, he said, so those decisions can be audited and learned from later.
His broader frame was an equation: a question asked in context needs relevant, usable data plus a learning model, and together they produce an outcome. The question he cared about was game theoretic: who wins, who competes and who cooperates.
Why does relevant, usable data decide who wins?
Singh said that models cannot learn from data that is irrelevant to the question, whether it is structured, such as historical real estate prices, or unstructured, such as social media sentiment.
He blamed decades of separate schemas, built by different companies with their own view of the world, for growing semantic heterogeneity. The semantic layers discussed elsewhere at the conference are one response, but they require everyone to agree on the same standards. Until then, he said, there is no relevant, useful data. Companies that solve this faster, he argued, will win.
Who owns the data, and will it be priced?
As models need training data, Singh said, data becomes more valuable and its owners will price it. He compared the instinct to a child with a toy: "it's mine." The boundaries of the firm are evolving, he said, and companies will have to decide whether to share, sell, syndicate or barter their data.
He pointed to the Reddit community's recent decision not to let APIs scrape its data for free as an example. He predicted data marketplaces in every domain and use case, using the example of a researcher who needs medical training data held by only five databases. Singh said such marketplaces would need blockchain-based attestation, including zero-knowledge proofs, to prove sellers hold the data they claim.
Will large models stay in the hands of big companies?
Singh expected compute-efficient language models to run at the edge of the network. He cited open-source models on Hugging Face and MosaicML that can run on very modest hardware. If models become commodities, he said, "the small can look large and the large might feel small."
On safety, he said he agreed with Warren Buffett's comparison of AI to nuclear weapons in a small share of use cases, which he put at 0.1%. He suggested that new models might need attestation from a central agency, acknowledged this is difficult in practice, and said the other 99% of use cases are benign.
What did Singh say in the Q&A?
He called blockchain a "trust machine." Companies keep separate systems and spend effort reconciling data and adding manual checks, he said. A shared ledger, once partners agree on a transaction, removes the discrepancy and the need to reconcile repeatedly.
On authenticity, he described tracking a diamond from the mine to the retailer so a buyer can scan a code and see its provenance, and said the same could apply to food, medicine and art. On privacy, he said the centralized model of storing data is broken and favored self-sovereign identity, where individuals choose what to share and can retract it. He added that business models, not just technology, have to change.
Continue the conversation at COLLIDE 2027
COLLIDE 2027, the largest AI conference in the Southern U.S. with more than 1,000 enterprise data and AI leaders, takes place Wednesday, October 13, 2027, at the Sandy Springs Performing Arts Center in Atlanta. See the COLLIDE page for details.
Frequently asked questions
How can blockchain work with AI?
Singh said blockchain can record the decisions AI systems make in a trustworthy way so they can be audited later, and could attest to data in marketplaces.
Why will data marketplaces emerge?
He said models need training data, so owners will put a price on it and buyers will need proof that sellers hold what they claim.
Will only big companies run powerful LLMs?
Singh predicted compute-efficient models will run at the edge, making them more accessible to small companies.
What does blockchain add for business partners?
He described a shared ledger where agreed transactions are visible to all parties, reducing reconciliation and manual checks.
This article is based on remarks by Roop Singh, Founder, Intuit Factory at COLLIDE Data + AI Conference 2023 on October 3, 2023. More coverage: COLLIDE AI Conference.
Filed under: AI Strategy & ROI, Data & AI Infrastructure, Security & Fraud
