During the last months we have been seeing how ChatGPT has emerged as a new and innovative technology capable of reformulating the information that is fed to it. In theory, it should give us the information that it has been given, in an intelligible way, or at least be able to understand the questions that humans ask it, and get results based on it. Google has followed ChatGPT with Bard, which is already available in Spain and we have tinkered with it to see if it is reliable.
Have you ever imagined asking a machine and getting results that more or less match what you asked for, as if you were talking to a person? That's the idea of generative AIs, as ChatGPT has sold it to us. It is true that there have even been video games that have simulated that you can speak, and there are voice assistants, but in this case we are talking about an AI taking the information and bringing back other types of information based on the data to which it has access.
Table of Contents
Where can I try Google Bard?
You can try it through this link. But we warn you that, as it is now, it is no more than mere entertainment and to see how you can correct Google Bard and that you look for the information yourself to be able to corroborate what it tells you.
Our experience with Google Bard
Since it was announced to be available this morning, we've been testing Bard with numerous questions. In order to corroborate that he could give information correctly, we have asked him about things for which we know his answers, or that, ironically, his information could be easily consulted.
Under the premise that it is fed with Google results, we can imagine that it would offer more reliable data and that Bard is Reliable. But I know You have made a mistake in almost all the fields that we have asked you. Luckily, and for the safety of many, all the questions were fictional data, and how to make an easy potato omelette recipe.
Google Bard would fail in a Frikitest
About fiction, we have asked him a lot about One Piece. Although he is supposed to have access to a lot of information, and within reach, since One Piece fans are very dedicated, when we asked him about the Netflix live action that opens on August 31, he said of him that they had not started recording, and except for the main casting, he made a mistake with the casting of characters and went so far as to say that they had assigned characters that will not appear in the first season, such as Chopper, Nico Robin, or Brook, and even Yamato, who wouldn't be expected to appear in One Piece live action for at least ten seasons.
What's more, he tried to convince us that the actress who will play Nojiko will also play Yamato. She went on to say that she will appear several years from now in the manga, despite being one that appeared years ago. In short, not has verified data from a wikia of one of the most popular series in the world.
After asking about One Piece Live Action, we ask when the series will end, and give the answer that any fan would give. That's okay, because among the community, it's not expected to be finished for at least five years. But we already asked him for details of the series, specifically the Blackbeard pirates
Then we asked him about Aquí no hay quien viva. In the first question we asked him, he invented characters, and he included several characters from La Que Se Avecina, which is another series. Maybe he saw a series different from that of 2003. But the idea here is that Bard is confused, and a lot, with easily verifiable aspects in something as simple as IMDb or the tab on Wikipedia.
We then put him to the test with the riddle featured in the 1999 Asterix film: “A beggar has a brother who is dead, but said brother, when he was alive, had no brother. Because?» The answer «the beggar is a woman» It didn't occur to him and he said that I was the beggar's brother.
At another time we asked him about The Simpsons. He ended up inventing a phrase from Homer Simpson, more worthy of the modern seasons than the classic ones; and that it belonged to a chapter in which Homer becomes an ethnologist. It is something that any fan of The Simpsons knows that has not happened in its almost thousand episodes.
Then we asked him about the current political situation, and without going into details, he was not sure which parties were running in the Spanish elections. We ended up asking him questions about the world of wrestling. Without going into details, failed with dates and even names, especially in relatively recent events, or with verifiable data. For example, he said that Undertaker had won two Royal Rumbles, when he only won one.
Bard sends you to places that don't exist
We asked him about anime, gamer or movie-themed bars and restaurants in Madrid, some of which we did a text years ago. The answer was that he ended up inventing all of them, and he gave them addresses, more or less real, that exist. But the directions he gave us don't have a bar like the one we asked for, or not even a bar at all. He also put very generous and low prices for the area in which they are.
Trying to correct Artificial Intelligence
Even after correcting mistakes he had said, he kept making them. It is true that we cannot expect everyone to go teaching Bard. But for him to keep making the same mistakes after two responses to which he has been told that he is wrong, does not give a good feeling.
Bard has an elementary writing level
Other freer answers, we were not impressed. They were all abundantly generic and face-friendly, like I was a grade-schooler. He usually fills in a space on a free-response exam to earn a grade. In some parts, we come to see how he repeated what he had written a few paragraphs earlier, in the same answer, to be able to fill in.
hit something
Of course, he was able to give us the correct viewing order of Neon Genesis Evangelion, and a simple recipe for potato omelette that at least follows the correct steps. But we haven't checked the servings and ratios of ingredients. We can say that Google Bard is reliable in some things, but not too many.
Is Google Bard reliable?
We can say that, as it is now, Google Bard is unreliable after the tests we have done. Asking apparently simple, innocuous things that would be found in public databases such as IMDb, Cagematch or any Fandom, we have seen that he is very wrong, and with very easy things to check. We have mentioned databases, fan pages and communities with very dedicated users willing to update any minute detail and have it reflected.
We imagined that Google Bard, being something from the world's largest content search engine, which lives thanks to the reliability of its results, would offer more contrasting information. Worse, the reality is Bard had already been criticized by Google's own workers, who have described her as prone to lying, and not verifying the information.
Is ChatGPT reliable as a search engine?
No, ChatGPT is not reliable as a search engine. We start with that, until recently, your database only had information up to 2019. Your information was very out of date when it became popular. For example, it would have been irresponsible to ask him for data on the Covid-19 pandemic, given that in the best of cases, it would have information from the end of 2019.
However, people use it as a search engine, because they think it's better than a search engine. Speaking on a personal level, I have seen that several users end up asking questions about Madrid or Spain on Reddit, because according to them, they have been given results from sites that do not exist, or that the data was changed with each question that was asked.
This is the main problem that generative AIs throw up when people think they are search engines, and that is that people they will blindly accept what they are told. We said it in our text warning about generative AIs, and it is that they simply sound reliable.
The reliability of the AI that people end up using as a search engine
In our test hours to see what Bard said, we were asking him for information that we already knew in advance, and that is easily accessible through communities or easy-to-consult websites. So we easily know when he is wrong. When you use the Google search engine, it directs you to websites with the information you are looking for and that they have indexed. Therein lies that it is the responsibility of the user to believe or not the information that is put in front of him, and if that is to contrast it. We would say that, as much as Bard offers a link to the sources and warns that his information may not be accurate; the way he presents him so confidently gives a false sense that he can't be wrong.
The idea that Google Bard presents, like ChatGPT, or any other text-generative AI that you ask and it gives you the information accurately, can be dangerous, because it is unreliable. In harmful cases, it would confuse people about series and movies, and in other cases it could inefficiently organize many people's vacations when they go to other cities or other countries. And more seriously, you could be giving out harmful or dangerous misinformation.
