sign up log in
Want to go ad-free? Find out how, here.

AI search, virtual assistant, generative video and music, chatbots and new Gemini model unveiled by Google

Technology / news
AI search, virtual assistant, generative video and music, chatbots and new Gemini model unveiled by Google
Dall-E-3 Google AI overviews
Google's Gemini refused to generate images with people for us, so we used OpenAI's Dall-E 3 for this AI impression of searching the Internet with artificial intelligence.

Google looks set to infuse just about all of its products and services with artificial intelligence, built on its Gemini models, as per the slew of announcements at its I/O 2024 developer conference show.

Some, such as the Veo video generator from Google subsidiary DeepMind, look like solid catchup efforts, following the lead of Microsoft-backed competitor OpenAI and its Sora technology. With Veo, you can type in text prompts to the AI and it will generate video in 1080p resolution. Google said Veo builds on its researchers' past work on video generation and no, you can't try it out quite yet and have to join the waiting list for access.

How extensive that access will be, and what safeguards will apply, remains to be seen, but Google said it will bring generative video capabilities to YouTube Shorts clips for users. This contrasts with OpenAI which has not yet announced any intentions of providing public access to Sora, out of concerns that the realistic videos the model can produce will be used for abusive purposes.

The Imagen image generator has also been updated with version 3, promising "an incredible level of detail, producing photorealistic, lifelike images, with far fewer distracting visual artifacts than our prior models." Again, Imagen 3 is waitlisted. Google has also put together what it says is a set of tools for musicians, prosaically named the Music AI Sandbox. There's no public waitlist for the Music AI sandbox, and it looks like you have to be at the level of Wyclef Jean fame to have access to the tools.

Google is the mastodon in the Internet room, and it has become so on the back of the content Internet users themselves generate, creating attractive tools to publish the material and helping others to find it - and selling vast amounts of advertising in return.

Where the next step in that process, which is Google moving up the content generation ladder with AI, will take the Internet is anyone's guess, but concerns are being raised that it won't be to a good place. For the above technologies, Google is seeking to allay fears of what it calls are "challenges", by which the company may or may not mean "flooding the Internet with vast amounts of plagiaristic AI generated slop", similar to what Facebook is now being overrun with.

We’re mindful about not only advancing the state of the art, but doing so responsibly. So we’re taking measures to address the challenges raised by generative technologies and helping enable people and organizations to responsibly work with AI-generated content. - Google.


Similarly, Google has been experimenting with generative AI for search, the key business that underpins its advertising sales. At I/O 2024, the expanded AI Overviews that summarise and organise search results was touted by Google. Since Google Search has a market share north of 80% worldwide, AI that summarises web content for users who don't have to visit the sites the material is actually hosted on would seem guaranteed to sink much of the Internet publishing economy. Maybe much of the web too, as looking at AI summarised search results is a one-way street.

Google disagrees, and said:

With AI Overviews, people are visiting a greater diversity of websites for help with more complex questions. And we see that the links included in AI Overviews get more clicks than if the page had appeared as a traditional web listing for that query. As we expand this experience, we’ll continue to focus on sending valuable traffic to publishers and creators. As always, ads will continue to appear in dedicated slots throughout the page, with clear labelling to distinguish between organic and sponsored results.

Overviews will be rolled out in the US first, then the rest of the world.

Less menacing, Google's Project Astra which is a multi-modal (that is, it can take text, audio and video input) AI looks quite impressive, judging by the demos. The virtual assistant can observe the world through your handheld device and take action accordingly; which kind of brings us to Google Lens that can now search video as well as images. 

On the Gemini side of things, the AI is coming to Google Workspace next month. If you're a Gemini subscriber that is. On Android devices that has the Gemini Nano AI built in (Google Pixel and Samsung smartphones) scam detection warnings for calls looks like an interesting and potentially very useful security measure. 

Gemini Nano, the smallest large language model in the range, is also coming to Chrome and can generate social media posts and product reviews. How that'll work in practice will be interesting to see, given the Internet's propensity to warp and corrupt well-intended efforts.

There's much more AI related news from I/O 2024, but the educational LearnLM stands out. Again, this is Gemini-based tech, with Google researchers talking about how "recent advances in generative AI (gen AI) have created excitement about the potential of new technologies to offer a personal tutor for every learner and a teaching assistant for every teacher".

Except that excitement has been sad-tromboned "due to the difficulties with verbalising pedagogical intuitions into gen AI prompts and the lack of good evaluation practices, reinforced by the challenges in defining excellent pedagogy." LearnLM hopes to fix that by fine-tuning datasets for Gemini and other tweaks for the AI. 

We welcome your comments below. If you are not already registered, please register to comment.

Remember we welcome robust, respectful and insightful debate. We don't welcome abusive or defamatory comments and will de-register those repeatedly making such comments. Our current comment policy is here.


Well yeah. The goal here is to pack the knowledge genie back into its bottle.

The Internet was established with the goal of a vast scale of knowledge-sharing. People could inform themselves about anything, mostly freely. Google capitalised on that by controlling the most convenient gateway for access to that knowledge.

"AI", in reality much of the A but none of the I, has set about destroying free and open search, driving people instead towards regurgitated drivel churned out by algorithms which gives the semblance of depth and wisdom but is in fact buzzphrases and high-level vagueness. When "assistants" dominate the landscape at our fingertips, and the trends set by early adopters have become pathetic dependencies, quality knowledge will again be gated and the ticket will be clipped. This is why "Pro" and "Plus" models exist.


Before I submitted the above, Copilot offered to rewrite it for me. Unsurprisingly the output was neutered and trite.

I'm not opposed to this technology, as trained algorithms have many incredibly desirable applications. But generative output for human consumption corrupts our collective wisdom, and I loathe its monetisation and models of control. I pray that the bottom drops out of it spectacularly.