Crypto and AI – Microsoft exec ‘We’re just scratching the surface’
Microsoft believes artificial intelligence (AI) is “the defining technology of our time,” and has been at the cutting edge of both AI research and investment.
But that doesn't mean the Seattle-based tech giant isn't also paying close attention to the cryptoverse, including the ways blockchain technology and AI may one day complement each other.
At the recent Cornell Blockchain Conference in New York, York Rhodes, Microsoft's director of digital transformation, blockchain and cloud supply chain, was asked how the company sees this technological convergence.
“As these two technologies advance, I think you can create agents that combine the power of both. We're just scratching the surface,” he said.
During a panel discussion titled “Crypto x AI,” Rhodes' views were moderated by Alex Lin, co-founder and general partner of Reforge, asking: Will Microsoft one day have its own blockchain?
“There are so many exciting things going on in crypto,” including the open source community, Rhodes replied, “so why try to reinvent something? [already] Does he have that much investment?”
Instead, Microsoft's focus today is on optimizing technologies such as Layer-2 blockchain scrolls. Rhodes added:
“But we would be.” [Microsoft] Ever build L1 blockchain? i don't think so.”
Crypto is “well kept”.
Rhodes and Lin were joined on stage at an April 26 event at Cornell Tech by Neil DeSilva, chief financial officer of PayPal Digital Currencies. Matt Stephens, Pantera's head of research; and Jasper Zhang, CEO and co-founder of Hyperbolic Labs.
“Crypto is well-positioned to be the ‘pick and shovel' of a certain type of AI,” Stephenson opined. This is especially true given the future of “decentralized, multi-agent” AI.
Still, crypto may have to play a secondary role to AI leadership. Rhodes admits that a “huge trend” like AI tends to “suck a lot of the air out of the room” for other new technologies, including crypto/blockchain and Web3.
“It's a hot topic – the intersection or symbiosis between blockchain networks and AI,” comments Lin. But it's also prone to exaggerated claims, and it can sometimes be hard to tell the hype from the real thing.
Much has been said about decentralized graphics processing units (GPUs), for example, Lin continued, “but no one talks about latency, that is, the time it takes to transfer data over a network.
Latest: Does Trump care about crypto? Bitcoin is the latest battleground in the US election.
These days, AI questions or recommendations can be quickly accessed from a centralized network. However, due to “latency”, decentralized networks do not produce those results quickly.
But Hyperbolic Labs' Zhang doesn't think this will be a problem for decentralized networks like blockchain. “It's predictable,” he said.
For example, take a centralized network with a data center in Texas that receives a request from a user in the United Kingdom. That data request “needs to travel across the ocean from the United Kingdom to Texas and then back again. So that has a really big delay,” Zhang said.
In contrast, with a decentralized network at scale, a user can easily find a node in London to process a local request, which “reduces the communication overhead.”
In fact, Hyperbolic Labs recently opened an AI inference interface on the company's decentralized network and achieved latency results comparable to centralized solutions, Zhang said.
A growing trend: minimal language models
Much of the current discussion in AI focuses on large-scale language models (LLMs), which require large amounts of computing power. However, according to Rhodes, “There's a lot going on with what you might call edge AI—getting small language models that work efficiently on phones and laptops.” that is:
There is a lot more computation at the edge as the models are getting smaller for certain workloads [and] In fact, you can get a lot more use out of it.
Microsoft has been developing small-language AI models that require less training data and computational power to develop and run, including the Phi-3 family of open models. The ability is “really starting to approach some of the larger language models,” Rhodes said.
Controllers have AI in their eyes
AI is likely to face intense scrutiny from regulators around the world in the coming years, as will cryptocurrencies. What obstacles did the panelists think there are regarding government laws and regulations?
“I think the United States is particularly bad at this [regulation]” said Lin, citing the US Securities and Exchange Commission's tougher approach to regulating cryptocurrencies. “right now, [SEC Chair] Gensler has come out and said AI will give us tighter control over blockchain digital assets.
Fintech powerhouse PayPal launched its own USD-pegged stablecoin, PayPal USD (PYUSD), seven months ago, and so Lin asked DeSilva to take on US regulation.
“I don't think the United States is bad in terms of regulation,” DeSilva said. “Look at all the innovation here in the United States.
Of course, sometimes dealing with government officials can be frustrating; But “regulators have a mission,” he explained, “don't hurt customers.” They're trying to protect consumers, and there's nothing inherently wrong with that. Or as he said from the stage:
“If you want your technology, your innovation, to be used by millions or billions of customers, you have to engage with regulators.”
Still, other jurisdictions, including the European Union, are becoming more welcoming to stablecoin issuers, and the United States should keep that in mind. “If the U.S. doesn't push too quickly, that's the advantage,” DeSilva said. “America has struggled to find the right level of urgency there.”
Achieving the right amount of control can be even more challenging with artificial intelligence. AI's opaque decision-making process — the so-called black box problem — “and I think regulators will struggle with that,” added DeSilva, making it difficult for regulators to control potential harm to consumers.
That transparency could provide an opportunity for blockchain technology with its transparency, immutability and traceability. Lynn said:
“Hey you [can] ‘Controllers, we have this mechanism that clears the opacity associated with these black boxes.'
Why AGI?
Lin concluded the session by asking panelists to share their visions for the future of AI. Will artificial general intelligence (AGI) become a reality in the next five to 10 years, for example? And what can one expect in the short term?
“In the near future, AI may become powerful enough that everyone will start using it,” Zhang predicted. “Just as every company is an Internet company, every company will be an AI company.”
“I think AGI will be possible in five to 10 years,” Zhang continued. “Look at how fast AI models can improve now, and with the help of a decentralized infrastructure we can generalize the computer,” meaning increasing the total number of GPUs available, which should also allow smaller players to participate.
Elsewhere, zero-knowledge proofs (ZK-proofs) will “give away in three years,” Rhodes predicts, to be replaced by Perfectly Homomorphic Encryption (FHA), a technology that “introduces zero trust by unnecessarily unlocking the value of information on untrusted domains.” to decrypt,” says IBM.
Reader: MakerDAO and Aave's DeFi Conflict Reopens Over DAI's Perceived Risk Growth
FHE solves many privacy problems, Rhodes said, and could be particularly useful for the healthcare industry, including clinical trials involving sensitive personal information.
Rhodes, in conclusion, recalled the words of the Wharton School's Ethan Mollick: “The AI you're using today is going to be the worst version of AI you've ever used.” The same can be said about ZK proofs and fully asynchronous encryption. In general, privacy-preserving computing frameworks will do better, he says.
DeSilva has worked in technology and finance for decades and has seen many realistic predictions come and go. But I got that optimism [often] He will win the day.
“So my prediction is you [will] Get to AGI in time, and it will be useful for people. That takes everyone's work.
