THE BATTLE FOR LARGE LANGUAGE MODELS
While lots of people are sharing the ChatGPT output here on LinkedIn and OpenAI has been better on the marketing front, it is easy to forget that there are more competitors on this front. Primarily META has been very big on this front. Also Google has been very active and there is even an Open Source model already on the market.
META is very active:
– Their OPT model is a direct competitor to OpenAI GPT-3 model with similar accuracy levels
– Their SPHERE model is meant to be a Google Search killer
– Their GALACTICA model is focused and trained on scientific papers
GOOGLE is bringing the topic to the next level:
– Their PALM model is able to use multi-modal information at the same time (video, audio, text) while GPT models are text-only
– Reinforcement Learning is included that can mimick the ChatGPT self-learning capability
– BLOOM is the first open-source competitor to GPT-3
– While its accuracy is not as high yet, I am sure the open source community and anti Big-Tech community will iterate fast
Personally, I think META’s A.I. team is one of the best in the market right now. They have surpassed Google’s A.I. teams. As a result, the use of their Pytorch library for example has been growing as compared to Google’s Tensorflow/Keras frameworks.
Microsoft is presumably trying to increase its equity in OpenAI to also get a foothold in this space. Noteworthy is the absence of other big Tech Players like IBM and Amazon who both seem to already be too far behind in the A.I. race in general.
However, all of these Big Tech players might be disrupted themselves by open-source alternatives once they reach a similar maturity level.
#genzai #artificialintelligence #gpt #openai #opt #meta #palm #google #bloom #sphere #bloom #galactica