Google Expands Availability of Gemini Chatbot, Enhances Speed

To keep up with generative AI contenders such as Anthropic and OpenAI, Google has unveiled enhancements to Gemini, its AI-driven chatbot accessible through a free tier. These enhancements aim to boost both efficiency and accessibility across the board.

From Thursday on, Google will extend the availability of Gemini 1.5 Flash — its streamlined multimodal model introduced in May — to users worldwide, offering support in 40 languages and across approximately 230 countries via both web and mobile interfaces. Google boasts significant improvements in the model’s quality and response times, with notable advancements in logical reasoning and image interpretation.

For Google, this update could also mean a reduction in operational costs.

During the launch event for Gemini 1.5 Flash, Google highlighted the model’s design as a streamlined, yet highly efficient, variant of the Gemini 1.5 Pro, optimized for “narrow” and “high-frequency” AI-driven tasks. Considering the financial implications of running a chatbot service similar to Gemini (referencing OpenAI’s ChatGPT operation costs), Google is evidently keen on finding ways to cut costs. These measures are likely to not only save money but possibly also enhance overall performance.

Moreover, Google plans to augment the contextual capacity of Gemini to 32,000 tokens, equivalent to about 24,000 words or 48 pages of text.

The concept of a context window pertains to the information a model examines before generating responses. Expanding this window permits more sophisticated summarizations and inferences over extensive text chunks, and ensures the chatbot remains on-topic during conversations.

Previously, uploading documents for analysis was a feature reserved for Gemini Advanced, accessible through Google’s Google One AI Premium Plan at $20 a month. However, Google intends to soon allow all Gemini users to upload files directly from Google Drive or their devices.

“Imagine uploading your economics study guide and prompting Gemini to formulate practice questions,” Amar Subramanya, Google’s VP of engineering, noted in a blog post disclosed to TechCrunch. “In the near future, Gemini will even assist you in analyzing data files, enabling the discovery of insights and their representation through charts and graphics.”

To address instances where Gemini 1.5 Flash may generate inaccurate content — known as “hallucinations” in AI terminology — Google is testing a function that will append relevant web links below certain responses generated by Gemini. This will initially be available to English-speaking Gemini users in select locations, offering deeper dives through a “chip” icon linking to further online resources or emails, assuming users have permitted Gemini to access their Gmail.

Google is taking these steps following admissions that its AI models can sometimes produce blatantly incorrect outputs, like unsafe cooking advice or fabricated product reviews attributed to real individuals. Earlier, Google introduced a “double check” feature in Gemini to aid in verifying the validity of its statements against other online sources, aiming for greater transparency in the origins of its information.

Still, it remains to be seen how effectively Gemini will present these related links.

Nonetheless, Google is proceeding to deploy Gemini across more platforms.

After initially launching Gemini within Messages for certain devices, Google is expanding this functionality to the EEA, U.K., and Switzerland. It adds support for more languages, including French, Polish, and Spanish, allowing users to engage Gemini directly from the Messages app by selecting it as a chat partner.

Google is also widening the availability of the Gemini mobile app in additional countries and extending access to teenagers globally.

In June, Google crafted a teen-centric version of Gemini, allowing students to register using their school accounts in select countries. This accessibility is set to expand, making Gemini available to teenagers in all countries where it’s typically accessible to adults.

In conjunction with this expansion, Google announces it will implement “extra policies and protections” specifically designed for younger users, though details remain sparse. Additionally, a new onboarding process tailored for teens, along with an “AI literacy guide,” aims to help them engage with AI in a responsible manner.

There’s ongoing debate over whether young individuals use AI tools as intended or misuse them. Google is obviously motivated to dodge claims that Gemini might encourage plagiarism or dispense questionable advice, hence its proactive approach to curbing potential misuse.