Google has made several updates to its free version of the Gemini chatbot, including opening up its low-latency multimodal model Gemini 1.5 Flash to the public and adding more source links to reduce the illusion effect.
Gemini 1.5 Flash was previously only available to developers and is best suited for tasks that require quick responses, such as answering customer queries. Google announced the model at its annual developer conference, Google I/O, in May and has since made it available to the public.
The model has a large context window, referring to how much information or words it can process at once, approximately 1 million tokens. Google states that Gemini 1.5 Flash on the Gemini chatbot will have a context window of 32K tokens. The large context window allows for more complex questions and longer back-and-forth conversations.
To leverage this advantage, Google is updating the free version of Gemini to handle file uploads from Google Drive or devices. This is a feature in Gemini Advanced, the paid version of the chatbot.
When initially launched, Google claimed that Gemini 1.5 Flash is 40% faster than OpenAI's fast model GPT-3.5 Turbo. Gemini 1.5 Flash is not a small model like Google's Gemma series, but rather trained with the same data as Gemini 1.5 Pro.
Gemini 1.5 Flash will be available for both the mobile and desktop versions of Gemini. It can be used in over 230 countries and regions and in 40 languages.
Reducing the illusion through links
The illusion remains an issue for AI models. Google is following suit with other model providers and chatbots by addressing the problem through adding relevant links to prompts requesting information. The idea is to show the AI model that the information is not created out of thin air but has references.
In a blog post, Google stated, "Starting today, in English language prompts in certain countries, you can access additional information about a topic directly in Gemini's response. Just click on the chip at the end of the paragraph to see websites you can dive deeper into for a particular topic."
The company also mentioned that if the information is in an email, Gemini will add links to the relevant email.
Google will also add a fact-checking feature that "uses Google search to validate responses, highlighting which statements have been confirmed or refuted on the web."
Google is not the only company adding links to indicate sources in chatbot responses. ChatGPT and Perplexity also regularly add citations and links to the websites where they found the information.
However, a report from Nieman Labs found that chatbots sometimes fabricate links, attaching them to nonexistent or completely unrelated news stories.
