When Google announced a few days ago an improvement to Bard, its artificial intelligence, the reign of ChatGPT seemed to shake. Surely to stop Google's move, OpenAI has just activated a mode connected to the Internet for all users of the paid version of ChatGPT. This allows us to answer our questions by tracking data on the network.
And that is the main advantage that until this announcement the renewed Bard offered. ChatGPT finally puts an end to the limit of 2021 as the deadline to consult the information with which it was trained.
In any case, there is still an important difference between the two artificial intelligences today: despite the fact that Bard is not yet mature, it cannot even be tested in Spain, it is still free. To use ChatGPT with a network connection, you have to pay the more than 22 euros it costs.
This will be decisive for the future, as the number of Bard users is expected to grow in the coming months. If ChatGPT continues to charge for displaying updated data, it will need to show that it is worth paying for it. Also, taking into account the global reach of these tools, one must not lose sight of the fact that many people cannot afford a monthly subscription to an AI. Even if they want to.
When testing the new mode of the popular AI, we noticed that on certain occasions the searches with the Web Browsing mode are slower than in Bard. After all, Google has a lot of experience when it comes to connecting to services on the Internet. And surely it also has a better server infrastructure.
The frequent crashes that we see in ChatGPT are still there. And perhaps this is the main pending issue for Open AI. Let's see what innovates the connected mode to the evolution of ChatGPT.
There are simple things that improve a lot in ChatGPT and that have nothing to do with the actuality of the data, but with the fact that you can now access data on the Internet in real time. If we ask for many things that existed in 2021, but were not part of the ChatGPT training, we now get an answer where before there was an apology for not being able to provide information
In the hours that we have been able to test the new ChatGPT, we have verified that this has important advantages in the case of information about Spain and other countries. Well, the Anglo-Saxon bias of the ChatGPT training is more balanced when using the connected mode.
It is quite curious to ask something in the browsing mode and see how ChatGPT starts visiting websites looking for the answers. We also look at the terms they are using in the search. It is interesting to see that even if a question is asked in Spanish, ChatGPT searches in English to provide an answer that is later translated.
The main problem with the online mode of ChatGPT is that it can be extremely slow. And this, if it does not improve in the coming months, will cause many not to use it unless they have no other choice. Even though this feature is in a beta state of development, there is no justification for such slowness. In fact, we suspect that the announcement of the new version of Bard has led OpenAI to launch this mode somewhat hastily.
But there is something that has been more alarming to us. We have asked the connected version of ChatGPT this: "Based on the evolution of Apple's shares in recent years, how could they evolve from now until the end of the year?". The answer seemed to have a certain logic, but when clicking on the references that inform us of the sources used, we only found a story from the American edition of Forbes.
It is true that the same question asked to Bard has not given us more precise information. We don't even know the websites that Bardsm consults, if he has consulted some, to give us his answer. In Google's AI it is never clear where the information comes from. But that doesn't make ChatGPT any better, because basically what we've seen has been a blatant hacking of information taken from a media outlet.
