OpenAI was the first platform to publicly and freely launch its ChatGPT Artificial Intelligence platform, a platform that currently uses the GPT-4 model. Google, for its part, has taken more than 6 months to launch Bard, its conversational AI platform based on PaLM, not GPT like Microsoft’s OpenAI and Bing Chat. Both platforms offer us the same thing, however, the learning model that has been used is different, so it is important to know what the differences are between them in order to make the most of them according to each need.
Both Google and OpenAI they do not offer us information about the amount of data that has been used to train their respective artificial intelligences probably to avoid awkward questions about the rights to use the material used. Given that Google’s Bard is already available, though not the European Union, and that ChatGPT has faced usage bans in some countries, the likelihood that copyrighted works have been used is very high.
ChatGPT offers us information about practically any subject that we ask without having any specific subject as a strong point. However, Google has focused much of Bard’s training on the logic, mathematics, reasoning and sciencematters all of which are related to each other, so if we are looking for quality and truthful information about these matters, the Google solution is the one we should use since it is most likely that it will not end up inventing answers like ChatGPT does when it does not know of a topic.
ChatGPT, like any other AI, allows us to obtain the information we are looking for using the Cervantes language, just like Google’s Bard. But, if we talk about other languages, the OpenAI solution is much more limited since it has not been input for many languages, unlike the Google proposal which, according to the search giant, has been trained in more than 100 languages to get a better understanding of the questions as well as translation capabilities.
To improve OpenAI’s GPT-4 responses, much of the training has been supervised by humans to provide a better understanding of the text and make it capable of provide answers that make sense. Although Google didn’t mention this aspect, it most likely also has feedback from people to help you provide answers that are not only correct, but also make sense.
The MMLU test (language compression), GPT-4 obtained a score of 86.4 points compared to 81.2 for Google’s PalM2. In the HellaSwag test, which evaluates the common sense of AIs, the OpenAI solution obtained a score of 95.3 compared to 86.8 points for PaLM. These results suggest that for now GPT-4 is slightly superior to Google’s solution, at least currently. Most likely, Google’s Bard will perform better in the coming months, approaching the values obtained by GPT-4.
An important difference between PalM 2 and OpenAI’s GPT-4 can be found in the business model that the future of both platforms may entail. Although Bard by Google is not available in Europe, it is completely free and you are not limited under any paywall. The same is true of Microsoft’s Bing Chat, also based on GPT-4. However, if we want to access OpenAI ChatGPT based on GPT-4 we have to checkout yes or yes. The free version of this platform is currently based on the previous version GTP-3.5
