from Hacker News

Meta to release commercial AI model in effort to catch rivals

by mgreg on 7/13/23, 4:45 PM with 7 comments

  • by mgreg on 7/13/23, 5:08 PM

    "Meta released its own language model, known as LLaMA, to researchers and academics earlier this year, but the new version will be more widely available and customisable by companies, three people familiar with the plans said. The release is expected imminently, one of the people said."

    Will be curious how they've improved on LLaMA (which is only about 4 months old) for their next release and what they've gained from the open source adoption of the model.

  • by ren_engineer on 7/13/23, 6:35 PM

    I always thought Facebook was dumb for not branching out more to diversify their income from purely social ads. I always thought their own search engine made sense, although that would also be ad based revenue.

    Monetizing their infrastructure and AI stuff via cloud services makes sense to me, although they may not have the cultural DNA to do it well. Google's struggles compared to AWS and Azure show it's not just the quality of tech that matters for B2B

  • by world2vec on 7/13/23, 11:04 PM

    Mark Zuckerberg said in Lex Fridman podcast that they learned a lot from the community by open sourcing LLaMA, even name-checked llama.cpp and that they're now using their own version of it internally. Looking forward to see what they've learned and improved in the past few months. This is quite exciting approach when compared to what OpenAI and Anthropic are doing.

  • by Oras on 7/13/23, 6:57 PM

    Would be interesting to see how companies will use these AI tools from Meta.

    Meta does not have the best reputation when it comes to data privacy

  • by archo on 7/13/23, 8:25 PM

  • by Havoc on 7/13/23, 6:07 PM

    Amazing. I really hope they release multiple size too like they did with llama