AI Still Has a Lot to Learn—and Unlearn: Ava Labs CEO

By Stacy Elliott

For all the buzz artificial intelligence has generated across the crypto industry, Ava Labs CEO Emin Gün Sirer isn’t impressed—yet.

“We can’t all go into AI,” he said during a recent interview for the gm from Decrypt podcast. “The AI valuations are sky high. There’ll be some stuff there, maybe, but $30 billion valuations for things that barely hold a conversation and can’t do arithmetic?”

“If you want to buy into it, fine. That’s great. But you can only do so much of that,” he added.

Despite his skepticism, Sirer isn’t trying to create FUD (fear, uncertainty, and doubt) around the wave of AI-driven projects starting to flourish in the blockchain industry.

The category, which includes AI tokens like Fetch and Singularity, has ballooned to account for almost $5 billion worth of the $1 trillion crypto global market capitalization, according to CoinMarketCap.

For context, that number has almost doubled since late January, when the category accounted for $2.7 billion worth of roughly the same global market cap.

“Imagine what you could do with an intelligent bot that is able to digitize assets, send them, buy them, trade them, using a uniform interface,” Sirer said. “That is going to be an amazing new thing, where I think asset pricing is going to be determined, not so much by people anymore, but also by bots—by algorithms.”

But first, the tech must overcome two big obstacles: Computer scientists need to find a way to stop AI from replicating gender or race biases, he said. And Sirer would like to see bots, specifically the ones that trade AVAX, learn to recognize black swan events.

He recalled that AVAX, the native token of the Avalanche network, took some collateral damage from the news that crypto exchange FTX filed for bankruptcy in November. That happened despite the fact that Avalanche has been pretty well insulated from direct exposure to Sam Bankman-Fried’s collapsed empire.

Sirer explained that for a long time, AVAX and SOL prices mirrored one another. So trading bots that use past data to predict future prices inferred that if one of the tokens moved, the other would follow.

Sirer hypothesized that when SOL’s price floundered on news of its financial ties with FTX and Alameda Research, the algorithms behind trading bots didn’t account for the fact that Ava Labs and AVAX weren’t similarly entangled with Bankman-Fried’s companies.

“If you have the training data that you did up until the FTX crash, then that’s fine. You’ve got that training data and that correlation is reinforced,”  he said. “And then FTX happens and that should only affect Solana—it’s the Sam coin.

“But because the bots have learned this [correlation],” he continues, “you have to now wait for a whole training cycle before the bots can unlearn the correlation between Solana and Avalanche.”

Listen to the full episode and subscribe to the gm podcast.

Recommended News