Artificial intelligence has captured the hearts, minds and wallets of the technology industry. So it’s little surprise that Microsoft, which has several irons in the AI fire, is working to expand its footprint in the area. On Wednesday, the company announced that it is partnering with layer-1 blockchain Aptos Labs to work on AI and web3.
“The primary focus for both of us is solving our respective industries problems,” Mo Shaikh, co-founder and CEO of Aptos Labs, told TechCrunch+.
The collaboration allows Microsoft’s AI models to be trained using Aptos’ verified blockchain information, Shaikh explained. Aptos will also run validator nodes for its blockchain on Microsoft’s Azure cloud, which it anticipates will bring greater reliability and security to its service.
Microsoft is also looking to the future. “We predict that AI will be infused into web3 solutions at greater scale in the coming months and years,” Daniel An, global director of business development for AI and web3 at Microsoft, said in an email to TechCrunch+.
There’s little doubt that AI is having a massive impact on society. “We can become incredibly efficient in using these tools every day in our lives,” Shaikh said. “Whether it’s searching and putting together an index of the best restaurants in your neighborhood or helping you write code for your job or research.”
For technologist’s AI aspirations to come true, there is a growing need for transparency, trust and verification of AI-generated content, An said. “For example, how do we know that LLM-generated outputs are authentic [and] trustworthy? How do we know that the training data is bias free in the first place? Blockchain-based solutions can help with verifying, time-stamping and attributing content to its source, thereby improving credibility in a distributed digital economy.”
An compares large language models to content creators “on steroids” and blockchains as “a yardstick for transparency and trust.” To help people become more comfortable with AI — and LLMs more specifically — companies have to make sure users trust how the technology works. “The openness and immutability of blockchain can improve the trust that people place in AI-generated content and provide confidence that they’re making the right decisions.”
AI needs to evolve responsibly, and web3 could help it earn the required credibility, Shaikh said. “Everything we capture onchain is verified and that verification can [help] train these models in a way that you’re relying on credible information.”
Microsoft’s interest is in having credible information to train models in a way that’s verifiable, which is where Aptos can fit in, Shaikh said. “In order to do that you need an incredibly performant blockchain with high throughput.”
The throughput of Aptos’ blockchain can support up to 160,000 transactions per second and has a goal of reaching the ability to handle “hundreds of thousands by the end of the year,” Shaikh said. It also ranks as one of the fastest blockchain networks, alongside Avalanche, with a time-to-finality of less than one second, according to Messari research.
With fast throughput, quick settlement times and a cost to use at a fraction of a cent, Aptos’ blockchain may have all three things needed to make it attractive to Big Tech companies interested in building AI-related products and services — like Microsoft.
AI to web3, web3 to AI
Helping mature tech companies use web3 tech to delve deeper into AI may assist those companies getting deeper into the blockchain realm. A major impediment to onboarding more developers into web3 today is how hard it is to write safe and secure smart contracts because it’s not super intuitive, An said.
“If you want to come and write a smart contract or development application, it’s difficult to do that today,” Shaikh said. But applications like the GitHub Copilot integration, which supports blockchain-based contract development, and Aptos Assistant, an AI chatbot that aims to bridge Web 2.0 and web3, could help non-web3-native companies access smart contracts and other decentralized tech.
Image by: Pexels
