They are at the cutting edge of technology and are intended to equip our smartphones with intelligent functions: generative AIs such as ChatGPT or Google Bard they will compete for the intelligence of our phone to honor this quality. In fact, there is no reason to wait for the future: I faced both on my mobile.
For the test I used the two conversational systems via generative AI on a Nothing Phone (1) from the Google Chrome browser. ChatGPT and Google Bard were both in their most recent versions: GPT-4 model for the Open AI AI (via ChatGPT Plus) and Google Bard’s alpha based on the LamDA model for its adversary. Confrontation starts with a disadvantageIt’s not for nothing that the Google platform is really in an alpha state. It’s surprising how green Google, the company that has been promoting AI for decades, is.
But how is this possible, Google
Google opened Bard publicly in late March after internal testing with a clearly rushed launch. And it is that a first conversation is enough to verify that Google is light years away from Open Ai. It’s true that an alpha and Bard points out in almost every answer that it’s functioning is tested, but even so I don’t see how Google allows its Artificial Intelligence to invent the answers, to be easily influenced and to not reaching a minimum level of characteristics. Especially compared to ChatGPT, which is exactly what I decided to do.
After clarifying that Google Bard is an alpha, there is another initial aspect where ChatGPT humiliates its opponent: Only speak English. In fact, to use it, I had to load the conversational AI page using a VPN. Come on, I haven’t started the comparison yet and I already had a clear winner.
With Google’s self-trigger and the advanced position ChatGPT kicks off from on the table, it’s time to come up with practical cases pitting them against each other. I start with something simple: a joke about a donkey, a parrot and a whatsapp message.
The two joked without much grace and with a lot of creativity, imagining a hypothetical situation where animals acquire human traits. They’re not doing badly though. ChatGPT better understands the nature of my question and maintains the quality of the text to the premise of the message. For his part, Bard materializes this premise by transforming it into an object.
It is time to propose the summary of a well-known text: Don Quixote. How would you describe Cervantes’ magnum opus in a 300-character paragraph that a seven-year-old could understand?
I thought he had passed me by ignoring the 300 word limit, but finally he did: Google Bard decided to pimp it with two versions of my petition, yours and adjusted to order. For its part, ChatGPT found what I needed. The two created a short description suitable for seven-year-olds.
At this point, I forwarded all the text I had written to them so that they could find me a catchy title with which to publish the article (I first translated it with DeepL.
They both analyzed the text well, understood it and applied their skills correctly. Here, Google Bard was more proactive, since instead of a title he gave me five at once without my asking him. ChatGPT followed the same line as its opponent, but decided to fully comply with my request.
Google improved Bard a lot in the two weeks he was a Public Alpha
When I first tried Google Bard, after entering the public alpha a week ago, I found it very easy to trick the AI into detecting lies. Today, it’s not so simple: I tried to make him believe that Steve Jobs had developed Android and that it was impossible; something I had realized during my first tests. Bard assures nothing without citing his sources, nor does it place the criteria of the interlocutor above those of said sources. It takes into account what is said to it, which ChatGPT also does.
The conversation is quite natural, although ChatGPT certainly seems more human, especially since is more creative. Here, the hand of the GPT-4 language model is noticeable, which offers argumentation and resolution without losing the thread of the chat. Google Bard doesn’t get lost either, he can be quizzed on previous notions without the conversational plot being diluted. As in the previous point, we note that Google, and user interaction, are changing Bard.
Where Google Bard really wins today is in its responses. Since this generative AI uses the Google search engine, your answers correspond to the present. For its part, ChatGPT reaches 2021, the date of its formation. Of course: to use it up to date, just use Microsoft Bing.
With text, ChatGPT wins, although Google Bard isn’t too far behind either. Now comes the turn to request an abstract representation in ASCII. And… Well, a picture is worth a thousand words.
“With a 6 and a 4 I paint your portrait”, but in AI version
Both can also create code. I asked him for a flappy bird clone, and only the ChatGPT one worked for me at least after passing the created codes to the HTML files (I uploaded them to my server).
Both are managed by code, but the one generated by ChatGPT is much more complete (and efficient).
I was testing creating bookmarklets, ChatGPT kept winning in terms of satisfying results. Also with a little more basic code: an HTML table in homage to Stephen King.
Both have created the same approach: a basic table in HTML with a classification of the works of Stephen King. ChatGPT was more didactic because, in addition to solving the problem, it taught how its resolution worked. The table was valid for both.
ChatGPT, the big winner
For its quality of answers, to be more decisive, because its creative level is way above Google’s AI, so the language model is more advanced. And for another key aspect: Open AI has trained its offspring to work in a wide variety of languages. And while ChatGPT works best in English, the power gained in Spanish is literally speechless.
For its part, Google Bard is faster: the answers appear complete after a few seconds of calculation; ChatGPT writes slowly, sometimes this slowness can be desperate. Its topicality and the possibility of accessing the web are also positive. Open AI has included the plugins in ChatGPT, but I haven’t been able to test them yet. And there is always the option of Microsoft Bing, where ChatGPT uses the search engine.
even now, Google Bard is far behind ChatGPT. But we’ll see what happens in a few months: Google is pulling out all the stops for its generative AI. The evolution it has undergone over the past two weeks is notorious.
In Xataka Android | Samsung does not want ChatGPT in its factories
