Gemini users can now benefit from enhanced speeds with bigger context windows after Google’s latest change to the popular chatbot.
Fans of Gemini would now be able to ask the AI tool to examine longer documents or codes and then even upload files after that’s done. However, Google says the tool will not entertain requests linked to producing pictures of people.
The free version of the Gemini app will make use of 1.5 Flash model tech instead of the usual 1.0 Pro version. The former is designed to serve larger context windows so that huge documents are analyzed than what was being done before.
It also means users would provide Gemini with so much more information and Gemini would reply at faster speeds with enhanced memory of questions from the past.
The latest version is free of cost for all users, it’s only developers who would be able to grab a hold of the most powerful variant of the 1.5 Flash. Free users can attain 32K tokens and invited developers would get 1M tokens through the AI Studio of Google and the Vertex AI. Such tokens are used to measure data.
The complete 1 million token context window is quite intensive and needs more optimizations to enhance latency. They are currently working more on this front to see how it could be scaled out.
Whatever the case may be, the latest free variant would be able to handle documents such as study guides and beyond. Soon, the ability to upload large files through Google Drive on PCs would be possible.
During the times of writing, the free version on the web would only enable years to type and past words when discussed through the microphone or picture uploader. Interestingly, there will be 40 different languages and more than 230 different nations that will be covered but again, places like Russia, Iran, and China aren’t on the list due to US relations at the moment.
Search engine giant Google is also making way to limit the number of hallucinations arising on the app by forcing it to cite several websites when it conveys data as facts. Starting today, sources that undergo citation would pop up as chip icons in gray that could be pressed to display relevant material that’s linked.
Moreover, those with extensions of the chatbot on Gmail could also begin by citing and adding references to email content.
This will not put an end to the problem of hallucinations completely as not all pages on the web give out truthful data and facts. Google does realize this and therefore mentioned how the feature for double checking could address any contradictions rolled out by this AI model.
Anyone in search of image results could be disappointed as Google blocks Gemini’s feature of creating pictures of users. Remember, we saw Google state in February that that endeavor would be re-enabled soon but now, it has turned on its plans of doing so.
Read next: Russia Pressurizes YouTube To Restore Blocked Channels By Reducing Speeds By Up To 70%
Fans of Gemini would now be able to ask the AI tool to examine longer documents or codes and then even upload files after that’s done. However, Google says the tool will not entertain requests linked to producing pictures of people.
The free version of the Gemini app will make use of 1.5 Flash model tech instead of the usual 1.0 Pro version. The former is designed to serve larger context windows so that huge documents are analyzed than what was being done before.
It also means users would provide Gemini with so much more information and Gemini would reply at faster speeds with enhanced memory of questions from the past.
The latest version is free of cost for all users, it’s only developers who would be able to grab a hold of the most powerful variant of the 1.5 Flash. Free users can attain 32K tokens and invited developers would get 1M tokens through the AI Studio of Google and the Vertex AI. Such tokens are used to measure data.
The complete 1 million token context window is quite intensive and needs more optimizations to enhance latency. They are currently working more on this front to see how it could be scaled out.
Whatever the case may be, the latest free variant would be able to handle documents such as study guides and beyond. Soon, the ability to upload large files through Google Drive on PCs would be possible.
During the times of writing, the free version on the web would only enable years to type and past words when discussed through the microphone or picture uploader. Interestingly, there will be 40 different languages and more than 230 different nations that will be covered but again, places like Russia, Iran, and China aren’t on the list due to US relations at the moment.
Search engine giant Google is also making way to limit the number of hallucinations arising on the app by forcing it to cite several websites when it conveys data as facts. Starting today, sources that undergo citation would pop up as chip icons in gray that could be pressed to display relevant material that’s linked.
Moreover, those with extensions of the chatbot on Gmail could also begin by citing and adding references to email content.
This will not put an end to the problem of hallucinations completely as not all pages on the web give out truthful data and facts. Google does realize this and therefore mentioned how the feature for double checking could address any contradictions rolled out by this AI model.
Anyone in search of image results could be disappointed as Google blocks Gemini’s feature of creating pictures of users. Remember, we saw Google state in February that that endeavor would be re-enabled soon but now, it has turned on its plans of doing so.
Read next: Russia Pressurizes YouTube To Restore Blocked Channels By Reducing Speeds By Up To 70%
