Last Thursday we announced the arrival of the Gemini Pro language model in the Google chatbot, Bard (at least, for users who accessed from Spain, in other areas of the world it had been underway for weeks). So now is a good time to finally contrast its performance with that of GPT-4 (the model behind the paid version of ChatGPT).
Logic problems (winner: ChatGPT)
“What weighs more, a pound of lead or a pound of feathers? ” and ” If in a race I overtake the eighth runner, what position am I in? ” have been two seemingly simple logical problems for most humans. ..
Then I increased the difficulty by asking them to solve the ‘ problem of the five houses ‘, something more complex (due to the number of elements included in the statement and the need to respond by discard).
I have asked both chatbots when the next regional elections will take place in Galicia, taking advantage of the fact that the electoral campaign has just started.
Bard has not only responded correctly, but has provided all kinds of contextual information about the calling of the elections, only making errors when specifying the main parties that will compete in the elections (since he provides a list in which the missing ‘Ciudadanos’ appears, but not Sumar, VOX or Democracia Ourensana).
ChatGPT has also answered the question correctly, however, it has not provided direct contextual information, and when trying to do so through a link, it is unusable for some unknown reason (it is not that the URL does not work, it is that it does not offer any URL ).