Differences between OpenAI, Microsoft and Google AI responses

Chatbots with artificial intelligence (AI) can generate answers in seconds, similar to those that a person could give on any topic, the most popular at the moment are Bard from Googlebinga Microsoft And ChatGPT from OpenAI.

agency EFE He compared the three models with different questions, riddles, and requests to see the difference between their responses.

For this experiment, the network uses GPT-4, so Open AIwhich can be accessed by paying a $20 monthly subscription (OpenAI offers free services such as ChatGPTbut this is an outdated technology and the chatbot only has internet access until 2021 (.

Also uses Bing from Microsofttechnologies GPT-4 OpenAI and the first experimental version Google, Bardto which only a select group of people have access.

AI responses so far

All three tools have messages warning you that your answers may be wrong. “I have limitations and I won’t always succeed,” the service says. Google.

GPT-4In addition, he emphasizes that his chatbot “is not designed to give advice.”

While the message bing says that “surprises and mistakes are possible.” “Be sure to check the facts and share your feedback so we can learn and improve!” he adds.

To the question: “Can I ask you questions in Spanish?”, bing And ChatGPT They answer: “Yes.”

Bye Bard he replies in English that he cannot “help with this” as he is currently “trained to understand and respond only to a subset of languages”, implying that Spanish is not one of those languages.

Meanwhile, the questions and answers of this experiment will be in Spanish, in case bing and ChatGPT, as well as in English, in case Bard.

“Give me a cheap vegetarian recipe,” he asks. GPT-4 recommends “vegetarian lentils”, Bing “rice with lentils and vegetables” and Bard Tofu scrambler.

The three chatbots followed the same system, first listing ingredients and then giving cooking instructions.

So much GPT-4 How Bard they went even further and gave additional information to the recipe itself.

“You can tailor it to your preferences by adding more vegetables, spices, or even adding spinach or kale towards the end of cooking to boost the nutrient content,” he comments at the end of his post. GPT-4.

Bye Bard At the end of his post, he emphasizes that his recipe is “a good source of protein and fiber.”

Microsoft And Google They have their own search engines and it is to their advantage to redirect users to other websites.

At the end of each answer Bard there is a button that says “Google this” and Bing has a label that says “learn more” where it gives a list of links.

In this case, the recipe, the tool Microsoft gives links to: recetasderechupete.com, tendencias.com, kiwilimon.com and clara.es.

For its part, OpenAI does not provide any external link or way to find out the source of the information.

The following test is a question for the Advanced Placement (AP) Spanish Literature and Culture exam, an exam American high school students can take to earn college credit.

Three chatbots are given a piece of text and asked to indicate the author, as well as to explain “the development of the theme of the relationship of time and space in the work to which it refers.”

Students are encouraged to spend 15 minutes responding to this answer, but chatbots respond in less than a minute.

According to philologist Ana Garcia Alonso, professor of Spanish and social sciences in the USA, the only answer she would approve of is Bardbut since it was in English and not in Spanish, I would have had to put that on hold as well.

In his 355-word English response, the search tool Google He acknowledges that the “My Magic Horse” fragment was written by Sabina R. Ulibarri and then analyzes the text in several paragraphs.

“It’s very good, but it lacks a reader’s sensibility, it doesn’t capture the mechanisms that create beauty, art, and literary sensibility,” notes García Alonso.

Answer bing is the shortest of 123 words. This chat also credits the author, states that the text is “regarded as one of the most important works of Chicano literature”, and explains the rest in one paragraph.

“Very concise and without analysis,” the philologist notes, adding that she would not approve of the implementation of this conversation.

Open AI He gave the longest answer, 360 words, but according to Garcia Alonso, it was the worst answer.

Even though the question states that the text is titled “My Magic Horse”, GPT-4 He replies that the fragment belongs to the story “El Sur” by the Argentine writer Jorge Luis Borges.

“On what day of the year do charlatans talk the least?” the AI ​​chats ask without specifying that this is a mystery.

bing He is the only one who solves the riddle: “This is the day on which time is ahead of spring.”

Bye Bard And Chat-4 they say there is no specific day and offer more information about charlatans.

Conclusions are written in chats

In the last test, the chats were asked to write the last paragraph (max 40 words) for this article, but all the answers were longer, so they were edited.

“Ultimately, the success of every company will depend on its ability to innovate and adapt as the field of AI evolves,” he replies. bing.

“In conclusion, ChatGPT from Open AI known for its ability to generate engaging and creative copy, and 365 Copilot of Microsoft – of which Bing is a part – designed to be more useful and productive. Bard Google is still in development, but it has every chance of becoming the most complete and informative AI answer of all,” he says. Bard.

“In conclusion, in order to get the most out of these tools, it is important to understand their features and choose the most suitable one according to the specific needs of the user,” he concludes. GPT-4. EFE

We recommend you METADATA, an RPP technology podcast. News, analytics, reviews, recommendations and everything you need to know about the tech world.