Google, Accel India accelerator chooses 5 startups and none are ‘AI wrappers’

Many synthetic intelligence startup concepts are nonetheless little greater than superficial “wrappers” constructed on high of present fashions. However because the AI mannequin makers add extra options, traders are cautious of startups that might develop into so simply pointless.

Working example: when reviewing extra 4,000 functions for the joint AI accelerator for India startups run by Google and enterprise agency Accel, “wrapper” concepts dominated. However none of them have been among the many 5 startups for the newest cohort, Accel accomplice Prayank Swaroop advised TechCrunch (pictured above).

Introduced in November, the AI-focused Atoms program by Google and Accel goals to again early-stage startups constructing AI merchandise linked to India. Startups chosen for the newest cohort will obtain as much as $2 million in funding from Accel and Google’s AI Futures Fund, together with as much as $350,000 in cloud and AI compute credit from Google, the corporations stated.

Roughly 70% of the rejected functions have been “wrappers” — startups that layered AI options akin to chatbots on high of present software program however “weren’t reimagining new workflows utilizing AI,” Swaroop stated.

Lots of the remaining functions that have been denied, Swaroop stated, fell into crowded classes akin to advertising automation and AI recruitment instruments, areas the place traders noticed little novelty. Startups in these sectors usually battle to distinguish themselves, he stated.

This isn’t, maybe, stunning. This 12 months’s program obtained almost 4 instances the functions than earlier Accel’s Atoms cohorts — with many first-time founders.

India’s rising AI ecosystem stays largely centered on enterprise functions and Swaroop stated the functions mirrored that. About 62% of the submissions centered on productiveness instruments and one other 13% on software program growth and coding, that means round three-quarters of the functions have been enterprise software program concepts quite than client merchandise. (Swaroop had hoped to see extra concepts for healthcare and schooling.)

Jonathan Silber, co-founder and director of Google’s AI Futures Fund, stated the 5 startups chosen aligned carefully with areas the place Google expects AI to see deeper real-world adoption.

This system doesn’t require startups to make use of Google’s fashions completely, Silber stated, noting that many firms mix a number of fashions relying on the workflow. The objective, he stated, is to assemble suggestions from startups on how Google’s fashions carry out in real-world functions.

Insights from these startups can then be fed again to Google DeepMind groups to assist enhance future fashions, creating what Silber described as a “flywheel” between startup experimentation and AI growth. “If an organization is utilizing another mannequin, meaning Google has work to do to construct one of the best mannequin out there,” he advised TechCrunch.

This 12 months’s startups chosen are:

• K-Dense, which is constructing an AI “co-scientist” to speed up analysis in fields akin to life sciences and chemistry;
• Dodge.ai, which develops autonomous brokers for enterprise ERP programs;
• Persistence Labs, which focuses on voice AI for name centre operations;
• Zingroll, which is constructing a platform for AI-generated movies and exhibits;
• Level Plane, which applies AI to industrial automation in automotive and aerospace manufacturing.