Google, Accel India Accelerator Choses 5 Startups And None Are ‘ai Wrappers’

Many artificial intelligence startup ideas are still small much than superficial “wrappers” built connected apical of existing models. But arsenic nan AI exemplary makers adhd much features, investors are wary of startups that could go truthful easy unnecessary.

Case successful point: erstwhile reviewing much 4,000 applications for nan associated AI accelerator for India startups tally by Google and task patient Accel, “wrapper” ideas dominated. But nary of them were among nan 5 startups for nan latest cohort, Accel partner Prayank Swaroop told TechCrunch (pictured above).

Announced successful November, nan AI-focused Atoms programme by Google and Accel intends to backmost early-stage startups building AI products linked to India. Startups selected for nan latest cohort will person up to $2 cardinal successful backing from Accel and Google’s AI Futures Fund, on pinch up to $350,000 successful unreality and AI compute credits from Google, nan firms said.

Roughly 70% of nan rejected applications were “wrappers” — startups that layered AI features specified arsenic chatbots connected apical of existing package but “were not reimagining caller workflows utilizing AI,” Swaroop said.

Many of nan remaining applications that were denied, Swaroop said, fell into crowded categories specified arsenic trading automation and AI recruitment tools, areas wherever investors saw small novelty. Startups successful those sectors often struggle to differentiate themselves, he said.

This isn’t, perhaps, surprising. This year’s programme received astir 4 times nan applications than erstwhile Accel’s Atoms cohorts — pinch galore first-time founders.

India’s increasing AI ecosystem remains mostly focused connected endeavor applications and Swaroop said nan applications reflected that. About 62% of nan submissions focused connected productivity devices and different 13% connected package improvement and coding, meaning astir three-quarters of nan applications were endeavor package ideas alternatively than user products. (Swaroop had hoped to spot much ideas for healthcare and education.)

Jonathan Silber, co-founder and head of Google’s AI Futures Fund, said nan 5 startups selected aligned intimately pinch areas wherever Google expects AI to spot deeper real-world adoption.

The programme does not require startups to usage Google’s models exclusively, Silber said, noting that galore companies harvester aggregate models depending connected nan workflow. The goal, he said, is to stitchery feedback from startups connected really Google’s models execute successful real-world applications.

Insights from those startups tin past beryllium fed backmost to Google DeepMind teams to thief amended early models, creating what Silber described arsenic a “flywheel” betwixt startup experimentation and AI development. “If a institution is utilizing an replacement model, that intends Google has activity to do to build nan champion exemplary successful nan market,” he told TechCrunch.

This year’s startups selected are:

• K-Dense, which is building an AI “co-scientist” to accelerate investigation successful fields specified arsenic life sciences and chemistry;
• Dodge.ai, which develops autonomous agents for endeavor ERP systems;
• Persistence Labs, which focuses connected sound AI for telephone centre operations;
• Zingroll, which is building a level for AI-generated films and shows;
• Level Plane, which applies AI to business automation successful automotive and aerospace manufacturing.