Post by account_disabled on Mar 10, 2024 7:18:40 GMT
Everyone talks about ChatGPT but not everyone has understood how it works and what the real implications of this technological leap called generative artificial intelligence could be. So I thought I'd have an in-depth conversation with Vittorio Di Tomaso, former founder of Blogmeter and currently partner in H-FARM Innovation, as well as an expert in computational linguistics. I had a good explanation of how these Large Linguistic Models work and how we got to this point. The turning point was the convergence of three factors: the existence of enormous quantities of information available, the increase in computational power and the invention of a specific neural network architecture called "Transformer" (by Google) .
In the specific case of ChatGPT, the merit of OpenAI was to be able to create a product that can India Mobile Number Data be used by everyone, bringing together data scientists, engineers and designers. The big question is: are these new GenAI software simply “stochastic parrots” that simply guess words to complete sentences or are they something more? The reference is to some unexpected and inexplicable behaviors, called "emergent properties", that these systems sometimes display. Finally we tried to understand how these models can be used to solve business problems.
Of course, says Vittorio, these systems promise to be capable of digesting all written corporate knowledge and making it accessible through a conversational system, but that's not enough. There is a control problem. We need to make sure that these corporate chatbots don't start talking about competitors or things that have nothing to do with the question. And this is done by implementing layers of control that require time and human work.
In the specific case of ChatGPT, the merit of OpenAI was to be able to create a product that can India Mobile Number Data be used by everyone, bringing together data scientists, engineers and designers. The big question is: are these new GenAI software simply “stochastic parrots” that simply guess words to complete sentences or are they something more? The reference is to some unexpected and inexplicable behaviors, called "emergent properties", that these systems sometimes display. Finally we tried to understand how these models can be used to solve business problems.
Of course, says Vittorio, these systems promise to be capable of digesting all written corporate knowledge and making it accessible through a conversational system, but that's not enough. There is a control problem. We need to make sure that these corporate chatbots don't start talking about competitors or things that have nothing to do with the question. And this is done by implementing layers of control that require time and human work.