Topics

in style

AI

Amazon

Article image

Image Credits:Exa

Apps

Biotech & Health

mood

Exa founders Jeff Wang, Will Bryk

Image Credits:Exa

Cloud Computing

mercantilism

Crypto

Article image

Image Credits:Exa

Enterprise

EVs

Fintech

Fundraising

Gadgets

Gaming

Google

Government & Policy

Hardware

Instagram

Layoffs

Media & Entertainment

Meta

Microsoft

privateness

Robotics

security measures

societal

Space

startup

TikTok

deportation

Venture

More from TechCrunch

Events

Startup Battlefield

StrictlyVC

Podcasts

picture

Partner Content

TechCrunch Brand Studio

Crunchboard

get through Us

While there ’s no shortage of startups aiming to supplant Google withAI - powered search(we’re looking at you , Perplexity ) , a startup calledExahas a different approximation : a GoogleforAI .

Humans are n’t the one who urgently involve a new kind of search locomotive engine , Exa ’s laminitis believe . Rather , as AI increasingly aim wait in incarnate and consumer life sentence , it is the AI platforms themselves that must regularly pretend out onto the cyberspace to search for entropy and return bona fide response , not hallucinations .   And they ca n’t just type their requests on their keyboards .

Exa is building a creature that allows AI models to do something like a World Wide Web search , but with an AI - native twist .

The co - father bought a million dollars ’ worth of GPUs ( which were well-situated to get in those days ) and , using a vector database and embeddings ( not a authoritative transformer - free-base LLM ) , they began to build a machine get wind model groom to natively realize link rather than words and judgment of conviction .

“ transformer commonly predict the next word . We train our lookup engine to forecast the next link , ” chief executive officer Will Bryk says . “ So people partake links on the WWW ; we use that datum as a dataset for our model that we train . And we train the example to predict the next link . So it ’s a refreshing search algorithm . ”

So just as an LLM would dispatch a sentence by furnish the most probable next word , Exa ’s organization does so with the most likely data link ( or 10 ) , but presumptively minus the SEO spam and ( ironically ) AI - generated brother choke off every ordinary search engine these days .

On Monday the startup announced it raised a young $ 17 million Series A led by Lightspeed ’s Guru Chahal , with involution from Nvidia ’s speculation weapon NVentures and Y Combinator , it solely told TechCrunch . Exa has now raised a total of $ 22 million , including its previous $ 5 million germ . ( Exa was in the summer 2021 YC cohort . )

Join us at TechCrunch Sessions: AI

Exhibit at TechCrunch Sessions: AI

“ This is a very challenging vision , ” says Chahal . “ What Google is to human beings , they are build for AI . ”

The squad was founded about a year before ChatGPT was launched , by two skilful friends who met their fledgeling year at Harvard : CEO Will Bryk ( now 27 ) and co - founder Jeff Wang ( 26 ) .

“ We launched before ChatGPT . Our initial goal as a society was not to function AIs at all . It was : how do you employ AI to build up practiced search ? ” Wang said .

After ChatGPT rage the technical school humans , AI companies begin asking Exa for an API version of their search engine that they could plug into their models . Exa is place in San Francisco , part of the informal Cerebral Valley AI startup exercise set . In fact , as TechCrunch previously report , a tweet by Wang went viral when he was look for other company that want to go inon an order of agency snooze pod and the response was overpowering . ( The body of work - nap - repeat culture is awake and well in this part of the tech industry . )

With AI companies now as its basal client , use cases for Exa ’s lookup locomotive span everything from an AI chatbot looking up information on the cyberspace while answering customers ’ doubtfulness to companies looking to curate breeding data .

Databricks , for instance , is Exa ’s marquee client , using it to detect large breeding set for its own model training opening move , the founding father say .

The API version of the product was launched around a twelvemonth ago . “ Since then it ’s gotten a screwball amount of traction , ” Wang says . Today Exa says it ’s serving thousands of developer — although it ’s worth pointing out that Exa has a free tier that allow anyone to try its hunt engine in a modified way . It also has multiple tiered - fee degree . The founders would n’t disclose gross except to say that they have some and the act is going up . ( Interestingly , asunder from running its own GPU - cluster , Exa hosts its production on AWS , not the AI - forwards Google Cloud . )

The team is n’t particularly pore on being the hunting startup that upend Google . Although if AI becomes the be - all / remnant - all that the technical school manufacture thinks it will be , search engine for AI bot could be the unexpected threat to the search hegemony .