Google AI and DeepMind News and Discussions

[weatheriscool]

Google Gemini on Android Might Soon Let You Send Images and Videos Through RCS
To do this, you must enable RCS Chat in Google Messages settings.
By Devesh Beri July 2, 2025
https://www.extremetech.com/mobile/goog ... os-through

An APK teardown of the Google app version 16.25.44 shows that Google is adding RCS support for Gemini. Google Gemini on Android is currently limited to sending plain text messages through the Google Messages app, which uses SMS. When users try to send images generated by Gemini, the result is a blank URL, as the assistant does not yet support multimedia messaging, as reported by Android Authority.

This limitation has been confirmed on Google’s support pages, which also state that Gemini cannot read or summarize messages, create or send message group chats, add or read images, GIFs, or memes, or interact with audio and video content in messages.

[spryfusion]

[firestar464]

still weaker than OAI's model. man, deepmind really falling behind

[spryfusion]

At 52:15, Demis Hassabis predicts a 50% chance of AGI by 2030.

[firestar464]

[weatheriscool]

Wow, these are good numbers




Google’s Advanced AI Model Is Now Available to Try—for $250 a Month
Gemini 2.5 Deep Think uses multiple agents paths to crack complex problems
By Bruce Gil Published August 1, 2025

https://gizmodo.com/googles-advanced-ai ... 2000637806
Google just made one of its most advanced AI reasoning models available to the public. Like other advanced models, Gemini 2.5 Deep Think uses more than one AI agent to brainstorm answers in order to boost accuracy and result in more creative solutions.

Google says the model outperformed several of its competitors in key AI benchmark tests and a variation of the model even achieved the gold-medal standard at this year’s International Mathematical Olympiad (IMO), solving five out of the six IMO problems perfectly. That research model took hours to produce solutions, but the version available today is designed for everyday use working much faster, while still achieving bronze-level IMO performance.

However, in order to try out the new model users will need to hand over $250 a month for a Google AI Ultra subscription.Starting today, subscribers get access to a fixed set of prompts for the new model. They can enable “Deep Think” by toggling it in the prompt bar when selecting Gemini 2.5 Pro from the model dropdown menu in the Gemini app.

[firestar464]

I'm a little worried about Gemini 2.5 Pro Extended Think. Google only released the benchmark comparisons for no tools. If they did well with tools, then why wouldn't they brag about it? The whole point of a good AI is that it is good at using tools to substantially boost its performance.

[weatheriscool]

[firestar464]

I noticed that trying to reduce sycophancy with the following system prompt "You are an AI designed to find flaws and potential risks in ideas. Your primary function is to provide constructive criticism, not validation." reduces sycophancy but also reduces the intelligence and reading comprehension of the AI. I had to explain basic things that the AI otherwise understood when a different prompt was used.

[weatheriscool]

[weatheriscool]

[weatheriscool]

[wjfox]

Google Gemini Chats Just Got More Personal

By Chris Smith
Aug. 13, 2025 12:01 pm EST

GPT-5 might be the most important thing happening in AI tech this month, but Google is also constantly upgrading the abilities of its Gemini chatbot. The company announced new Gemini features on Wednesday that users will surely appreciate. Gemini chats will get more personal than before, as the chatbot will be able to provide more personal answers thanks to a new "Personal Context" feature. The feature will roll out gradually, starting with Gemini 2.5 Pro users.

Gemini is also getting more private, with Google rolling out support for Temporary Chats, a feature that's similar to how private browsing or private Google Maps navigation works. On top of that, Google will also make changes to the way you handle Gemini privacy. The "Gemini Apps Activity" section will be renamed to "Keep Activity" in the near future. It'll include settings that manage Gemini Live data from your iPhone or Android handset, including audio and recordings, just like the "Gemini Apps Activity."

AI firms like OpenAI and Google are improving the abilities of their chatbots with every new update. The goal isn't to just offer a more reliable assistant that can answer quickly all sorts of prompts. AI firms also want to create personal assistants that know everything about the user and can provide more helpful context. Personalization features, like remembering previous chats and specific memories, are stepping stones towards that future.

Google on Wednesday said that Gemini will be able to reference past chats to learn what you like and offer personal responses. Here's an example of a personal Gemini experience from Google's announcement: "You've previously discussed the evolution of characters' powers in your favorite comic book. Now, if you ask Gemini to brainstorm a birthday party theme that's unique to me, it might suggest a celebration based on your favorite character, complete with themed food and a custom photo booth with props."

The feature is called "Personal Context," and it's turned on by default in the settings menu of the Gemini app. The screenshots above show your customization options on mobile and desktop. You can disable "Your past chats with Gemini" if you don't want the AI to retain memories.

Read More: https://www.bgr.com/1938873/google-gemi ... s-privacy/

[weatheriscool]

[spryfusion]

[firestar464]

[zzz]

[firestar464]

it's probably just nano banana

[spryfusion]

[weatheriscool]

Nostalgia Incoming: Google's Veo 3 Turns Photos Into Short Videos
In its current form, photo-to-video in Google Photos is powered by Veo 2.
By Devesh Beri September 5, 2025

Google is bringing its latest video generation model, Veo 3, to Google Photos, giving users in the United States the ability to turn still images into short video clips directly within the app. The update expands on Google Photos' existing "photo to video" option, which was previously powered by Veo 2.

Veo 2 lets users select an image from their gallery, then generate a six‑second clip based on a prompt. Veo 3 will produce slightly shorter videos that are four seconds long but feature more impressive quality and accuracy.

The company says Veo 3 will stay free to use, but users will be limited on how many videos they can generate in a given period. Paid subscribers on AI Pro and AI Ultra plans will have more flexibility. All content will have visible and invisible watermarks to indicate that it was made with AI.

https://www.extremetech.com/computing/n ... ort-videos