вторник, 3 октября 2023 г.

Should Meta AI be Worried?

Can't read or see images? View this email in a browser
 

After open-sourcing the best AI models, such as Stable Diffusion, Stability AI recently unveiled Stable LM 3B, built for running on-edge devices. Meta AI, considered one of the proponents of open-source AI with the Llama 2 release, now seems to have some real competition besides Falcon and Mistral 7B.

https://stratus.campaign-image.in/images/83238000172927088_8_1696338121531_24a5c941f0f496a5b38e21aa1gif

Meta is possibly heading the closed-source way. In a recent podcast with Lex Fridman, Mark Zuckerberg said that he is still on the fence about deciding whether to open source the next iteration of their LLM, Llama 3. This is similar to how all companies that started as open-source drives, such as OpenAI, become closed-source. Mistral AI also has similar plans.


Developers love open source. That is why Meta with its open source LLaMA and Llama 2 has been getting all the praise. But with Stable LM 3B, the tables might turn against Meta and in favour of Stability AI.


Stable LM 3B is also based on a similar architecture as Llama’s decoder-only transformer architecture. Also, the dataset used for training includes Falcon-refined web and RedPajama, which happens to be the open-source repository for Llama. 


Read the full story here.



Satya is not Smiling 


Of late, Google has been accused of locking up distribution channels to maintain a monopoly in the consumer search engine market. For instance, the Alphabet-owned company pays Apple a handsome price of almost $20 billion annually to keep Google Search as the default browser on its devices. 


Not surprisingly, Apple seems to now favour Google, leaving its fellow Bing hanging. Recently, Microsoft chief Satya Nadella criticised Google's dominance in the search engine market. 


Read the full story here. 



Making GPUs More Accessible in India 


According to reports, the current GPU shortage witnessed by the AI industry could persist until 2025. To address this issue in India, E2E Networks, India’s first AI hyperscaler, is offering cutting-edge NVIDIA H100 GPU (based on Hopper architecture) and the NVIDIA A100 Tensor Core GPU at a price point of just INR 412/hour and INR 226/hour, respectively.

https://stratus.campaign-image.in/images/83238000172927088_7_1696338121301_this-indian-hyperscaler-ijpg

In an exclusive interaction, Tarun Dua, CEO of E2E Networks, told AIM that around 10-15% of the 100 startup unicorns in India were born in the company’s cloud. “We began writing our cloud software around 2013-14. Today, we are very focused on AI/ML. We have been working on the computational side of AI, and heavy-duty processing based on cloud GPUs.”


Read the full interview here.



Did Google Miss the Metaverse Bus? 


Recently, Mark Zuckerberg took everyone by surprise by appearing in an interview with Lex Fridman in the Metaverse. This time, it was for real – nothing sketchy or cartoony. Meanwhile, Apple also plans to create photorealistic avatars for Vision Pro during FaceTime calls. 


Google, however, seems to be missing out on the metaverse adventure. Also, there is little to no noise about its ambitious Project Starline, which is likely to join the Google Graveyard soon. Check out the full story here


     

TAUSIF ALAM & AMIT RAJA NAIK

Tuesday, Oct 3, 2023 | Was this email forwarded to you? Sign up here

     
   

DOWNLOAD OUR MOBILE APP

Stay Connected

info@analyticsindiamag.com

© 2023 Analytics India Magazine

   
Facebook
Twitter
LinkedIn
Youtube
Instagram
   
 
Analytics India Magazine | 280, 2nd floor, 5th Main, 15 A cross, Sector 6, HSR layout Bengaluru, Karnataka 560102

Комментариев нет:

Отправить комментарий

Примечание. Отправлять комментарии могут только участники этого блога.