Topics
Latest
AI
Amazon
Image Credits:Peresmeh / Getty Images
Apps
Biotech & Health
Climate
Image Credits:Peresmeh / Getty Images
Cloud Computing
Commerce
Crypto
endeavor
EVs
Fintech
fundraise
Gadgets
bet on
Government & Policy
ironware
Layoffs
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
protection
Social
Space
Startups
TikTok
Department of Transportation
speculation
More from TechCrunch
result
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
Magic , an AI inauguration creating mannequin to generate computer code and automatise a reach of software program ontogenesis tasks , has kick upstairs a tumid tranche of hard currency from investor , include ex - Google CEO Eric Schmidt .
In ablog poston Thursday , Magic said that it closed a $ 320 million fundraising around with contributions from Schmidt , as well as Alphabet ’s CapitalG , Atlassian , Elad Gil , Jane Street , Nat Friedman and Daniel Gross , Sequoia and others . The financial support impart the company ’s total raised to nearly half a billion buck ( $ 465 million ) , catapulting it into a age group of better - funded AI put on startups whose appendage includeCodeium , Cognition , Poolside , AnysphereandAugment . ( Interestingly , Schmidt is back up Augment , too . )
In July , Reutersreportedthat Magic was seek to raise over $ 200 million at a $ 1.5 billion valuation . manifestly , the round came in above expectations , although the inauguration ’s current evaluation could n’t be ascertained ; Magic was valued at $ 500 million in February .
Magic also on Thursday announced apartnershipwith Google Cloud to build two “ supercomputers ” on Google Cloud Platform . The Magic - G4 will be made up of Nvidia H100 GPUs , and the Magic G5 will practice Nvidia ’s next - gen Blackwell chips scheduled to come online next year . ( GPUs , thanks to their ability to run many computations in analog , are commonly used to train and suffice generative AI models . )
Magic says it aims to scale the latter clustering to “ X of M ” of GPUs over time , and that together , the cluster will be capable to achieve 160 exaflops , where one exaflop is equal to one quintillion calculator operation per secondly .
“ We are excited to partner with Google and Nvidia to progress our next - gen AI supercomputer on Google Cloud , ” Magic co - laminitis and CEO Eric Steinberger said in a financial statement . “ Nvidia ’s [ Blackwell ] system will greatly improve illation and preparation efficiency for our models , and Google Cloud pop the question us the fastest timeline to scale of measurement , and a rich ecosystem of cloud services . ”
Steinberger and Sebastian De Ro co - set up Magic in 2022 . In aprevious audience , Steinberger told TechCrunch that he was inspired by the potential of AI at a young age ; in high shoal , he and his friends wire up the school ’s computers for motorcar - learning algorithm training .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
That experience plant the seeds for Steinberger ’s computer science Bachelor ’s program at Cambridge ( he dropped out after a year ) and , later , his job at Meta as an AI researcher . De Ro come from German business process direction house FireStart , where he worked his elbow room up to the role of CTO . Steinberger and De Ro meet at the environmental volunteer organisation Steinberger co - created , ClimateScience.org .
Magic develops AI - driven tool ( not yet for sales event ) designed to aid software program applied scientist write , reexamination , debug and design codification change . The peter operate like an automatise duad computer programmer , attempting to understand and continuously see more about the context of various coding projects .
Lots of platforms do the same , including the elephant in the roomGitHub co-pilot . But one of Magic ’s invention lie down in its models ’ ultra - long setting window . It calls the theoretical account ’ computer architecture “ longsighted - term Memory connection , ” or “ LTM ” for forgetful .
A model ’s circumstance , or context of use windowpane , refer to stimulation datum ( e.g. code ) that the model considers before generating output signal ( e.g. additional code ) . A bare question — “ Who won the 2020 U.S. presidential election ? ” — can serve as context , as can a movie handwriting , show or audio magazine .
As context windows grow , so does the size of the documents — or codebases , as the case may be — being fit into them . Long context of use can forbid models from “ forgetting ” the message of recent docs and information , and from veering off subject and extrapolating wrong .
Magic claims its late modelling , LTM-2 - mini , has a 100 million - token context windowpane . ( Tokens are subdivide bits of raw information , like the syllables “ devotee , ” “ tantalum ” and “ tic ” in the Logos “ fantastic . ” ) One hundred million tokens is equivalent to around 10 million lines of computer code , or 750 novels . And it ’s by far the largest context windowpane of any commercial model ; the next - large areGoogle ’s Gemini flagship modelsat 2 million tokens .
Magic says that thanks to its prospicient circumstance , LTM-2 - mini was able to implement a countersign strength meter for an open source undertaking and produce a calculator using a customs UI framework pretty much autonomously .
The company ’s now in the cognitive operation of training a larger version of that model .
Magic has a modest squad — around two dozen multitude — and no revenue to verbalize of . But it ’s move after a market that could be worth $ 27.17 billion by 2032,accordingto an estimation by Polaris Research , and investors comprehend that to be a worthwhile ( and possibly quite remunerative ) endeavor .
Despite the security , copyright and reliableness concern around AI - powered assistive encrypt tools , developers have express enthusiasm for them , with thevast majorityof responder in GitHub ’s latest poll saying that they ’ve take on AI tools in some form . Microsoft reported in April that co-pilot hadover 1.8 million paying usersand more than 50,000 business customer .
And Magic ’s ambitions are grander than automating mundane software package developing task . On the fellowship ’s website , it verbalise of a track to AGI — AI that can solve problems more reliably than humans can alone .
Toward such AI , San Francisco - found Magic recently hired Ben Chess , a former lead on OpenAI ’s supercomputing squad , and plan to expand its cybersecurity , engineering , research and organisation technology teams .
