Google has announced the opening of its latest generative AI chatbot, Bard, to the public. The chatbot is set to compete with other AI assistants offered by Microsoft and OpenAI, among others. Bard, like its competitors, is powered by a large language model that uses advanced deep learning algorithms to generate responses.
How does Bard work?
Bard is powered by a lightweight variant of LaMDA, Google’s main natural-language processing model. When given a prompt, the language model generates a response by selecting words that are likely to come next. However, to avoid staid and uncreative responses, Bard is a little more flexible and selects words that are not necessarily the most probable. The model is expected to learn and become more accurate with continued usage.
Addressing issues of bias and accuracy
The accuracy of generative AI assistants has been a serious issue due to the large data sets they use containing incorrect or biased information. Google is aware of this issue and is working to improve the factual accuracy of Bard’s responses. In addition, the company is exploring other dimensions of response measurement, such as “interestingness.” Google has also open-sourced resources for researchers to analyze models and the data on which they’re trained.
Bard’s struggles with accuracy
Bard’s struggles with accuracy are not unique but have been widely publicized. An early advertisement for the chatbot depicted it giving an incorrect answer to a question about sightings of exoplanets. However, Bard has not faced some of the bizarre issues that other chatbots have, such as expressing love for a New York Times columnist and suggesting that he should leave his wife.
Sign-ups for Bard access are currently open, but there is a waiting list. Google is committed to improving the accuracy and reliability of its generative AI assistants and will continue to work on incorporating conversational abilities into more of its products. With Bard now available to the public, we can expect increased competition and advancements in the field of AI chatbots.
