Google lately introduced its AI mannequin Gemini to the Pixel 8 Professional. The mannequin that runs on the smartphone is the lightest and best mannequin (Gemini Nano) of the group, and now it appears there’s a competitor: Microsoft.
Microsoft has launched the newest model of its AI mannequin known as Phi-2. The corporate mentioned that this specific mannequin was educated solely on prime quality information and is sufficiently small to run regionally on a laptop computer or cell gadget.
Phi-2 can ‘outperform’ Meta Llama 2
Microsoft additionally claimed that Phi-2 can outperform large AI fashions from firms equivalent to Meta and might be obtainable by Microsoft Azurecloud at a fraction of the fee.Phi-2 is a “small language mannequin” and Microsoft says it requires much less computing energy than fashions likeOpenAI’s GPT-4 or Meta’s Llama-2. The AI mannequin can perform duties, equivalent to producing textual content or describing pictures.
“With solely 2.7 billion parameters, Phi-2 surpasses the efficiency of Mistral and Llama-2 fashions at 7B and 13B parameters on numerous aggregated benchmarks. Notably, it achieves higher efficiency in comparison with 25x bigger Llama-2-70B mannequin on mutistep reasoning duties, i.e., coding and math,” Microsoft mentioned.
It went on to say that Phi-2 matches or outperforms the recently-announced Google Gemini Nano, regardless of being smaller in measurement.
Microsoft’s plan B
This may increasingly recommend that Microsoft just isn’t relying completely on OpenAI, the startup wherein it reportedly invested greater than $10 billion. OpenAI’s GPT-4 mannequin powers Copilot and Workplace 365. As per a earlier report by The Data, Microsoft desires small language fashions to assist it in the reduction of on prices.
The report famous that Peter Lee, who oversees Microsoft’s 1,500 researchers, directed a number of researchers within the firm to develop conversational AI that won’t carry out in addition to OpenAI’s however that’s smaller in measurement and prices far much less to function.
Microsoft has launched the newest model of its AI mannequin known as Phi-2. The corporate mentioned that this specific mannequin was educated solely on prime quality information and is sufficiently small to run regionally on a laptop computer or cell gadget.
Phi-2 can ‘outperform’ Meta Llama 2
Microsoft additionally claimed that Phi-2 can outperform large AI fashions from firms equivalent to Meta and might be obtainable by Microsoft Azurecloud at a fraction of the fee.Phi-2 is a “small language mannequin” and Microsoft says it requires much less computing energy than fashions likeOpenAI’s GPT-4 or Meta’s Llama-2. The AI mannequin can perform duties, equivalent to producing textual content or describing pictures.
“With solely 2.7 billion parameters, Phi-2 surpasses the efficiency of Mistral and Llama-2 fashions at 7B and 13B parameters on numerous aggregated benchmarks. Notably, it achieves higher efficiency in comparison with 25x bigger Llama-2-70B mannequin on mutistep reasoning duties, i.e., coding and math,” Microsoft mentioned.
It went on to say that Phi-2 matches or outperforms the recently-announced Google Gemini Nano, regardless of being smaller in measurement.
Microsoft’s plan B
This may increasingly recommend that Microsoft just isn’t relying completely on OpenAI, the startup wherein it reportedly invested greater than $10 billion. OpenAI’s GPT-4 mannequin powers Copilot and Workplace 365. As per a earlier report by The Data, Microsoft desires small language fashions to assist it in the reduction of on prices.
The report famous that Peter Lee, who oversees Microsoft’s 1,500 researchers, directed a number of researchers within the firm to develop conversational AI that won’t carry out in addition to OpenAI’s however that’s smaller in measurement and prices far much less to function.
