Google is opening up to non-English speaking users of its tools. From February 1, 2024, Gemini Pro - the most advanced AI model of the company - is available in the Bard chatbot in most languages of the world, including Polish. But these are not all the novelties.
Free access to this article is provided to you by MU Group.
"From today, Gemini Pro in Bard is available in all languages, countries and regions where Bard is available" - announced on February 1, 2024 in a post on the Google blog Jack Krawczyk, product lead at Bard. This specifically refers to over two hundred and thirty countries and over forty languages of the world, including Polish.
In practice, this means above all faster and more contextually matched statements given in languages other than English, but also advanced responses about coding, better content summaries and more advanced chatbot reasoning. Users will also receive greater possibilities for verifying content generated by AI.
"[...] we are making our checking feature available in over forty languages. When you click the »G« icon, Bard will check if there are contents on the internet confirming his answer. If you find such content, you can click on highlighted phrases to display similar or conflicting information found in the search engine" - we read in Krawczyk's post.
This is all the more impressive as, according to the Large Model Systems Organisation report, which Google refers to in its blog post, Bard AI with Gemini Pro is one of the most popular chatbots in the world available for free.
Jakub Rusak
In addition to expanding the language availability of Gemini Pro in Bard, Google also announced the update of its chatbot with the Imagen 2 model. This engine is used to generate high-quality photorealistic images and has so far been available exclusively to developers - from February 1, however, ordinary users will also be able to use it.
Imagen 2 - unlike other models of this type - is supposed to be good at realistically generating human hands and avoiding artifacts. Google also assures that the tool has been trained and programmed to prioritize data security. It uses, among others, SynthID to place digital watermarks used to distinguish AI products from human ones.
"We have also added technical safeguards to limit problematic results, such as content containing violence, offensive or explicitly sexual in nature. In accordance with our AI policies, we have also conducted extensive adversarial tests and red teaming attacks to identify and minimize the number of potentially harmful and problematic content. We also use filters to avoid generating images of specific people" - assures Jack Krawczyk in the post.
Currently, the image generation function in Bard is available only in English. Google does not provide information on when this may change.
Blindly following trends, lack of a clearly defined identity and vision, or relying on too narrow measures of success. How not to fall into the most popular traps that lurk for brands on social media?
Many indications suggest that the year 2024 will be crucial for the further development of generative AI models trained on publicly available materials. Subsequent courts will decide whether and to what extent this technology infringes copyright.
Inclusive language is a way of communication that aims to avoid discrimination and promote equality by taking into account differences. Unfortunately, something that should unite, more and more often divides. What does this result in and how to change it?
The company Creandum published a report on the salaries of startup founders. The study was conducted among 650 entrepreneurs from 47 European countries, and the aim was to provide data on the average earnings of founders at various stages of their companies' development.
English