The search giant Google announced on Monday the launch of bard which is the first phase of a replica to ChatGPT. The OpenAI text generator has gained global notoriety by rallying a hundred million users in less than three months, a unique adoption record in the history of tech.
Bard works on the same principle as ChatGPT but in a more “calm” version, with a greater concern for precision and accuracy than the OpenAI bot, says Google. Aware of its responsibilities — Google provides more than 90% of Internet searches worldwide — the company did not want to launch a conversational agent likely to trigger the same controversies as ChatGPT, criticized as much for its tendency to approximation than for its anarchic use: ChatGPT has proven to be a powerful means of generating fake news at the same time as a formidable instrument of cheating in education…
The Google robot is developed from LaMDA, an acronym for Language Model for Dialogue Applications which is a system stuffed with billions of words from digitized books, Wikipedia, and millions of news articles. As with ChatGPT, a vast network of artificial neurons sorts and orders this huge model to answer a question posed in natural language. TheMDA exists since 2017 and incorporates patents that Google has put in the public domain.
References to sources
But compared to its OpenAI competitor, Bard is constrained in two dimensions. One, it will only offer summaries on a limited number of subjects, such as scientific questions where the references are indisputable. Summaries will be concise and Bard will not allow long texts to be generated — for the dissertation, best to stay on ChatGPT at your own risk.
Second specificity, the results of Bard will be systematically accompanied by the classic presentation of the lists of links referring to original sources (see below)
Bard, Google’s conversational AI has a more measured approach than ChatGPT.
© / Bard, Google’s conversational AI has a more measured approach than ChatGPT.
For Google, it was essential not to fall into the trap of the black box which restores untraceable references, a great weakness of conversational robots.
Eventually, Bard will be integrated into Google’s classic search engine. But it will first be tested with caution. A “private beta” version will be distributed to an informed population of professionals who will be required to give their opinion on the system in order to make it evolve. This is a common practice at Google, which performs some 300,000 tests on its search engine each year, resulting in 4,000 adjustments. In a few weeks or months, the bot will be open to the public.
Finally, Bard will be accompanied by a series of programming application tools (APIs), allowing the construction of functionalities and services on the basis of its technical architecture. This announcement from Google in the field of artificial intelligence is the first of a series of concrete innovations in the field of “search” (Internet research) which will be revealed this week in Paris.