Generative Artificial Intelligence continues to remain at the center of the technological discussion: it is the topic of the moment. Above all because several companies are investing and experimenting in a public way, to convince us of the potential of their solution. So we tried to compare ChatGPT, Bing e Google Bard to understand how these AI chatbots work and what they are the main differences between them. To try and predict which AI might become the winner in this already exciting race.
ChatGPT vs Bing vs Bard: le differenze fra i chatbot AI
Prova Bitdefender. Leader in Cybersecurity
ChatGPT e Bing use the same language model: GPT-4, developed by OpenAI. A novelty that has recently arrived, improving the performance of GPT-3.5: having tested both models more than once on both platforms, the change in accuracy seems visible.
Instead, Google only recently opened Bard to the public. And the results for the moment seem to demonstrate that it is still needed some experience to be able to catch up. Google’s artificial intelligence, based on LaMDA, currently seems to be struggling to produce results at the level of the competition – although we’ll see how much potential there is. Bard’s Trial, however, it is not possible in Italy at the moment: Google has only opened testing in North America. We will therefore use as examples the tests of various American sources (such as this article by The Verge) to evaluate the functionality of the service, which we will test as soon as it is legally possible to do so in Italy.
We also point out that there are several other companies that have developed AI-based chatbots, including the recent ones DuckAssist di DuckDuckGo, Ernie About Baidu and it seems that even Apple is working on something for Siri. The university of Stanford she even managed to create one for just $600. But at the moment, the main competition seems to be between the three protagonists of this article: ChatGPT, Bing and Bard – all of which have noticeable differences in the way they respond.
The main differences, especially between ChatGPT compared to the other two, concern the features. Bing and Google want to use AI to improve their search platform. ChatGPT, on the other hand, is used to train the a compose texts in the most “human” way possible.
This means, for example, that ChatGPT does not use recent information to answer our questions. If you were to ask him to talk about the best smartphone out there, for example, he would compile a list of features to look for, while Bing and Google would search the web to give us direct advice. Unfortunately, all three platforms still have the tendency to provide inaccurate answers if there is little material available online: for example, if we ask about a newly announced product, they use invented specifications. But if online information has only been available for a few weeks, accuracy (especially from Bing) is good – but be warned, not perfect.
Images and sources
A difference recently brought in Bing is about being able to generate images, taking advantage of generation models that however go beyond GPT and use other OpenAI solutions. But the main difference lies in the possibility of citing sources.
In fact, every time you ask for information at Bing Chat, it will provide links to show the sources from which it got the answers. Google has promised that Bard will do the same when fully operational, but for now it is limited to offering a button to “Goog” the question and thus find the sources used. Instead, OpenAI recently announced the use of plugins to do the same. But for now, Bing Chat has an edge for those who need to research and prove the sources used.
The right tool for every question
artificial intelligence university
From the various tests we’ve attempted in person and the ones they report from the US on the Bard, there seem to be different results depending on what you’re asking. The differences between ChatGPT, Bing and Google Bard mean that there is no clear winner. Although for the moment there seems to be a Clear underdog: Google Bard.
ChatGPT works very well when asked to be creative – although he does it better in English than in Italian, where for example he has trouble creating rhyming poems. But if you ask to compose texts with a particular style or following well-defined steps, ChatGPT seems to be the winner. Even when it comes to using logic to understand text and give contextual responses, it seems to have an edge.
Bing on the other hand works great for giving contextual answers and providing sources of what it reports. Likewise, it allows better to create texts on recent topical subjects – even if a control of the information remains a must.
At the moment therefore we recommend to use ChatGPT if you want to look for something evergreen (recipes, advice on what to see in a city visited by tourists, etc.) e Bing for more recent and detailed information (newly released products and comparisons, recent news). And to use Google without AI, until it gets a little better.
But any evaluation can only remain partial: these models “feed” on data and when Google will bring Bard around the world it could quickly recover ground. We will keep you posted.