Databricks co-founder wins prestigious ACM award, says ‘AGI is here already’

Databricks co-founder and CTO Matei Zaharia astir missed the email telling him that helium was the 2026 recipient of the ACM Prize successful Computing. “Yeah, it was a surprise,” helium told TechCrunch.

Back successful 2009, the tech Zaharia developed for his PhD astatine UC Berkeley, nether the tutelage of famed prof Ion Stoica, was launched into Databricks.

Zaharia had created a mode to dramatically velocity the results of slow, clunky, large information projects and released it arsenic an unfastened root task called Spark. Big information was successful those days what AI is contiguous and Spark turned the tech manufacture connected its ear. The 28-year-old Zaharia became a tech celeb.

Since then, helium has helmed the engineering astatine Databricks, increasing it into a unreality retention elephantine and present a information instauration for AI and agents. Along the mode the institution has raised implicit $20 cardinal — valuing it astatine $134 billion — and deed $5.4 cardinal successful revenue. The Silicon Valley dream.

On Wednesday, the Association for Computing Machinery issued him the grant for his corporate contributions. The grant comes with a $250,000 currency prize that helium is donating to an as-yet-to-be-determined charity.

Zaharia, who successful summation to his CTO duties is besides an subordinate prof astatine UC Berkeley, is looking forward, not back. Like everyone other successful the Valley, the aboriginal helium sees is filled with AI.

“AGI is present already. It’s conscionable not successful a signifier that we appreciate,” helium told TechCrunch. “I deliberation the bigger constituent of it is: we should halt trying to use quality standards to these AI models.”

A person, for instance, tin lone walk the barroom exam to beryllium a lawyer if they’ve integrated immense amounts of knowledge. But an AI tin ingest immense amounts of facts easily. If it answers cognition questions correctly, that doesn’t equate to wide knowledge.

This tendancy to dainty AI similar a quality tin person immoderate profoundly antagonistic impacts. He offers the illustration of the fashionable AI cause OpenClaw.

“On the 1 hand, it’s awesome. You tin bash truthful galore things with it. It conscionable does them automatically,” helium said. But it’s besides “a information nightmare” due to the fact that its designed to mimic a quality adjunct that you spot with things similar passwords. That leads to the hazard of being hacked, oregon the cause spending unauthorized wealth from your slope due to the fact that your browser is logged in.

“Yeah, it’s not a small quality there,” helium says.

As a prof and merchandise engineer, Zaharia is astir excited astir however AI tin assistance automate probe connected everything from biology experiments to information compilation.

Just similar however vibe coding made prototyping and programming accessible to anyone, helium thinks that accurate, no-hallucinations AI-powered probe volition someday go universal.

“Not that galore radical request to physique applications, but tons of radical request to recognize information,” helium said. Eventually we’ll marque AI enactment amended for america by having it thin into its strengths: telling america what each rattle successful our car means, oregon scanning beyond substance and images to see vigor and microwaves, or, what he’s seeing students bash now, simulate molecular-level changes and foretell their effectiveness.

“The happening that I’m astir excited astir is what I’d telephone AI for search, but specifically for probe oregon engineering,” helium said.