Google is testing new chatbots – the integration of AI into the company’s search engine cannot be ruled out in the future. Active work is underway within Alphabet’s cloud department, it is expected that its own version will be presented, capable of competing with the ChatGPT project, to which the public gained access at the end of last year.
Google is also testing the Apprentice Bard chatbot, employees can ask it questions and get detailed answers, just like in the case of ChatGPT. A special department is testing a version of the search engine for desktop computers – the search can be in the form of questions and answers. Management encourages more employees to use new developments and give feedback to the company.
Product testing has intensified since a recent company meeting where employees expressed concern about the emergence of ChatGPT and the possibility that Google could lose the AI lead. Google’s head of AI, Jeff Dean, said Google is being forced to pay more attention to reputational risk than a startup and, accordingly, adopt more conservative policies. However, the head of the company Sundar Pichai (Sundar Pichai) announced the possible release of Google-like products this year.
One of the test products is the Apprentice Bard chatbot, which uses Google’s LaMDA (Language Model for Dialogue Applications) technology. As a result of the launch of ChatGPT, the company’s LaMDA team was asked to prioritize the development of a competitor for OpenAI.
Apprentice Bard is similar to ChatGPT in many ways: It is able to answer questions in a dialog box. It will replace the earlier version of Meena, which was once introduced within the company but the project was soon shelved. Staff note that the Apprentice Bard has become a more advanced tool in recent weeks.
Additionally, the company is known for testing alternative search sites that can work in a question-and-answer format. For example, a template is already being tested in which “more human” answers are displayed directly below the search bar, followed by topic-related questions and then the usual search results including links and headings. It is not yet known which of several options Google will choose.
There is also evidence that the Google team has already compared the quality of the beta version of LaMDA with ChatGPT. He is known to have received a Level 3 pre-qualification for the position of programmer during ChatGPT’s “interview” which usually indicates the employee’s ability to coordinate the actions of Level 1 and Level 2 colleagues . It is not known what level the LaMDA tool has received.
When asked by bots whether solutions like ChatGPT and AlphaCode could replace programmers, OpenAI like LaMDA confidently and competently answered that this would not happen, although such tools could make the job easier. When asked to describe the shoplifting scene for the LaMDA film, a script was offered, while ChatGPT went for a longer and deeper story. Finally, when solving a logical-arithmetic problem, ChatGPT gave the wrong answer, while LaMDA presented the correct solution. However, the data are rather superficial. Recently, for example, there were reports that ChatGPT was taught to look for errors in the program code. This can be an important benefit when comparing the two systems.