An alleged data leak from Google reveals some information about its upcoming multimodal AI model Gemini, as well as a new AI tool called Stubbs. The leak suggests that Gemini will support features such as text and object recognition, image captioning, image understanding, and more. (and possibly audio and video processing). Stubbs is described as an AI platform for building working application prototypes, deploying and sharing them using a simple command prompt and optionally an image of the desired application.
Google Gemini or Gemini AI is an integrated suite of large language models (LLMs) developed by Google. According to Google CEO Sundar Pichai, Gemini’s core models were designed to be multimodal from the ground up. This means users can process and generate text, images, code and audio content from a single interface. Gemini is currently in beta testing by a select group of developers from a small number of companies. Gemini is expected to replace PaLM 2, the language model that currently powers Google Bard, by the end of 2023.
So far, very little information about twins has been filtered out. In August, a report revealed that Gemini is expected to combine GPT-4’s capabilities with training techniques inherited from AlphaGo, such as reinforcement learning and tree search. It would have the potential to displace ChatGPT as the most dominant generative AI solution on the planet. Gemini appears to be the strategic answer to Google ChatGPT. The launch of ChatGPT shook Google to its foundations. The threat was so great that the research giant had to declare a code red and begin investing to catch up with OpenAI.
These efforts resulted in the release of not only Bard, but Gemini as well. While Bard struggles to attract users, Gemini wants to challenge ChatGPT’s dominance in the generative AI market. Zoubin Ghahramani, vice president of Google DeepMind, said Gemini will be available in the same four sizes as PaLM 2: Gecko, Otter, Bison and Unicorn.
- Gecko was intended to be lightweight and ideal for mobile devices;
- Otter is designed to be stronger than Gecko. It should be suitable for a wide range of unimodal tasks;
- The bison is larger and more versatile than the otter. It is expected to be suitable for a limited number of multimodal tasks and compete with the GPT-4 model in terms of market share;
- Unicorn is the largest, most powerful and most versatile of the four Gemini models. This version should be suitable for a wide range of multimodal tasks and far exceed the capabilities of ChatGPT or any of its competitors.
Rumor has it that if all goes well, Gemini will also be integrated into all of the AI-powered cloud services that Google offers to businesses and individuals, including Google Search, Google Translate, and Google Assistant. Once launched, Gemini’s scalability – along with its flexible tooling and API integration capabilities – will allow it to be used in a wide range of real-time desktop and mobile applications. The data leak suggests that Google is expected to release Gemini directly into MakerSuite, a type of studio that enables the quick and easy creation of applications based on generative AI models.
MakerSuite provides an efficient user interface for leveraging some of Google’s latest templates and easily translates prompts into production-ready code that users can integrate into their applications. Google announced in late September that it had removed the waitlist and that users in 179 countries and territories could now use MakerSuite. Additionally, the data leak heralds the impending launch of Makersuite Autosave to enable automatic saving of prompts. It also announces the launch of a new tool: Makersuite’s AI-driven application construction tool Stubbs.
According to the leaker, Stubbs allows you to build working applications on a single site using a single command prompt. Users can create/build, deploy and publish applications from Stubbs. The ability to create applications and launch them from a single site with a simplified interface is perfect. You publish the app and can then share the link. Is this new territory for AI? Probably. There will be a community gallery where you can post your Stubbs for everyone to see. You can also remix Stubbs and have your own version of an idea, notes the author of the leak.
The sequel’s author says Stubbs won’t replace app developers but will provide a “massive” boost to the industry. From what we can see, these will look like AI-generated Figma prototypes and will not create complete code, but prototypes of AI-generated applications that will work. The Stubbs Gallery lets you view Stubbs created by other people. By default, when you build an application from Stubbs, it is not published. The leak says that you have to publish it explicitly, which means the public won’t directly see everything you create.
The data leak provides a glimpse into the future of Gemini and the new Stubbs AI tool, but this information is not precise and it is still difficult to get a clear picture of Google’s true ambitions with these two AI technologies. Stubbs looks like an AI-based no-code tool. Leaked information about this so-called secret Google tool is reigniting debate about the emergence of AI-based no-code tools that could replace developers in the near future.
While waiting for Google to choose Stubbs, the author of the leak announces that more Gemini and Stubbs data leaks could follow. He said: “While the first leak provides an overview, Stubbs’ technical details will be released soon to help you form your own opinion about what he is doing in another leak.”
Sources: blog post, Google
And you ?
What is your opinion on this topic?
What do you think of the AI tool “Stubbs” developed by Google?
In your opinion, can this no-code tool claim to replace developers?
What do you think about the potential features of Google’s Gemini AI model?
With Gemini, could Google overtake ChatGPT and regain the lead in the AI race?
See also
Google’s Gemini AI model could be five times more powerful than OpenAI’s GPT-4 language model and is intended to combine the capabilities of GPT-4 and AI programs such as AlphaGo
Google co-founder Sergey Brin returns to help develop Gemini, ChatGPT’s competitor. Former Alphabet president gets involved in AI strategy and recruiting
How Google DeepMind plans to revolutionize generative AI and overtake ChatGPT with Gemini, its AI system that combines the strengths of AlphaGo with the capabilities of large language models