Could ChatGPT be a threat to Google? In any case, the American company will respond: the group is preparing a response based on much more recent data and which could be offered directly on its homepage.
Details are beginning to emerge on how Google will retaliate against OpenAI and its ubiquitous ChatGPT. According to information obtained by CNBC on Jan. 31, the search giant is currently testing a competing chatbot dubbed Apprentice Bard, which can formulate answers equivalent to those provided by ChatGPT today.
There is, however, a notable difference in the description reported by the American TV channel. Apprentice Bard is able to work on much fresher data, where ChatGPT is based on a body of information stopping at 2021. CNBC was thus able to get Google’s chatbot to react to the group’s layoffs, announced in January 2023.
If this ability is retained in a hypothetical release of this chatbot, this would give it a substantial advantage over ChatGPT, if it does not evolve accordingly. Of course, the freshness of the data will not be everything: the relevance of the answers will be crucial. We saw it with ChatGPT. It’s good to make texts that are articulated and inspired by natural language. It’s better when it’s accurate.
A chatbot powered by LaMDA, the language model made in Google
To feed Apprentice Bard – this name will certainly change if the product is released one day – Google uses the LaMDA language model (Language Model for Dialog Applications), an in-house solution developed by the American company. Here too, it is a question of simulating a natural language, as if we were conversing with a real person.
It was in 2021 that LaMDA was first presented to the public. On stage, the boss of Google had fun staging a discussion with a paper airplane on the one hand and with the dwarf planet Pluto on the other, to demonstrate the progress of the Mountain View firm in the treatment automatic language and in the ability to follow an exchange.
It should be noted that LaMDA was again talked about a year later, when a company engineer started claiming that LaMDA had evolved to become a “real” artificial intelligence, with a conscience. The employee, dismissed by his employer, was however contradicted by his peers and AI specialists, believing that he was on the wrong track.
Another reflection of Google around its chatbot would be to replace the “I’m lucky” button on its homepage with a shortcut allowing you to immediately send a command (a “prompt”) to obtain an answer personalized. The search engine would then become a “response engine”. But it is still necessary, here again, that the answers are good.
It’s unclear if and when these leads will come to fruition, but CNBC’s material attests to the fighting momentum at Google over the past few months. Another recent example was the discovery of a research paper around a “musical ChatGPT” made in google. On paper, it can be made to generate music using words.
However, announcements are expected soon. In January, the New York Times reported that more than twenty services and products based on artificial intelligence algorithms are in the pipeline. No doubt Apprentice Bard is one of them, as is the prospect of a button to launch prompts directly on the engine’s home page.
As such, two meetings are precisely to be followed with great attention: on February 8, a conference around AI is planned – we should have a small overview of the state of research at Google. The most spectacular revelations, however, should take place in May, during the annual Google I/O conference, at the group’s headquarters.
If you liked this article, you will like the following ones: do not miss them by subscribing to Numerama on Google News.