Amazon invests $4B in Anthropic AI startup

Ama­zon and arti­fi­cial intel­li­gence (AI) start­up Anthrop­ic have announced a new invest­ment agree­ment to aid the research and devel­op­ment of new high-per­form­ing foun­da­tion models. 

In a post on X (for­mer­ly Twit­ter), Anthrop­ic revealed that Ama­zon will invest $4 bil­lion in its work and give it access to Ama­zon Web Ser­vices Traini­um and Infer­en­tia chips. 

Anthrop­ic said it will, in return, offer “enhanced sup­port” of Ama­zon Bedrock, which pro­duces foun­da­tion mod­els, with “secure mod­el cus­tomiza­tion and fine-tun­ing” for busi­ness­es. Ama­zon teams will also be able to use Bedrock to build on Anthropic’s models. 

Our agree­ment pro­vides Anthrop­ic with access to @awscloud Traini­um and Infer­en­tia chips. We’ll also offer enhanced sup­port of Ama­zon Bedrock with secure mod­el cus­tomiza­tion and fine-tun­ing for businesses.

— Anthrop­ic (@AnthropicAI) Sep­tem­ber 25, 2023

Addi­tion­al­ly, through the deal, Ama­zon is report­ed­ly tak­ing a “minor­i­ty stake” in Anthrop­ic. The lat­ter said this has not dis­rupt­ed its governance. 

“As out­lined in this pol­i­cy, we will con­duct pre-deploy­ment tests of new mod­els to help us man­age the risks of increas­ing­ly capa­ble AI systems.”

The AI start­up was formed by for­mer mem­bers of Microsoft-backed Ope­nAI, the cre­ator of the viral AI chat­bot ChatGPT.

Relat­ed: Anthrop­ic cracks open the black box to see how AI comes up with the stuff it says

This lat­est devel­op­ment comes short­ly after Anthrop­ic announced an invest­ment of $100 mil­lion from the South Kore­an telecom­mu­ni­ca­tions giant SK Tele­com back in August. 

That invest­ment fol­lowed a col­lab­o­ra­tion between Anthrop­ic and SK Tele­com to devel­op a mul­ti­lin­gual large lan­guage mod­el (LLM) for the latter’s Tel­co AI Platform.

Anthrop­ic has been a part of major move­ments with­in the AI com­mu­ni­ty. In July, it joined Google, Ope­nAI, Microsoft and oth­ers in form­ing the “Fron­tier Mod­el Forum,” which was cre­at­ed to self-reg­u­late AI devel­op­ment from the inside. 

It has also been a part of ini­tia­tives led by the Unit­ed States gov­ern­ment relat­ing to AI devel­op­ment and reg­u­la­tion, includ­ing a cyber­se­cu­ri­ty chal­lenge to help strength­en its “crit­i­cal infrastructure.” 

Mag­a­zine: ‘AI has killed the indus­try’: Easy­Trans­late boss on adapt­ing to change