From early innovations to Future Frontiers

AI is not new, and it’s not just chatbots. At Google we’ve been working on AI for over two decades.

September 2025 7 min read

This article is from the FORWARD magazine #2.

In 2001, SpellCheck was our first AI-based tool. Since then, significant advances have proven that AI has the potential to help us make daily life easier, reach big goals and address some of humankind’s most complex and pressing social and economic challenges.

Google’s AI milestones include various technological, scientific and ethical achievements – and even a Nobel Prize. These milestones of the past form the basis for AI-driven tools that now help people and businesses around the world every day. And, of course, the journey continues. Google will keep pursuing the extraordinary and the everyday opportunities of AI, boldly and responsibly, to improve the lives of as many people as possible. In an interview, Lila Ibrahim, COO of GoogleDeepMind, explains how AI can help improve our future.

THE PAST

2001

ML for Spell Check
Good results despite typos in the search window: Google begins using machine learning (ML) with spell check at scale in Google Search.

2006

Google Translate
Launch of Google Translate using ML to automatically translate languages, starting with Arabic-English and English-Arabic. As of 2025, more than 240 languages and varieties are supported.

2015

Google Photos
Launch of a new app that uses AI with search capability to search for and access your photos by the people, places and even things or animals that matter.

2017

Transformer
Google introduces the Transformer architecture, a novel neural network architecture that all large language models are based on today.

2018

AI Principles
Google was one of the first companies to publish AI Principles. We follow these guidelines while developing and using AI. They are designed to maximise AI’s potential while safeguarding against risks.

2020

AlphaFold 2
AlphaFold is recognised as a solution to the 50-year »protein-folding problem«. Accelerating research, more than two million researchers have used it. Received a Nobel Prize in Chemistry in 2024.

2022

PaLM
Google announces PaLM – or the Pathways Language Model. It was Google’s largest language model to date, trained on 540 billion parameters and the predecessor of Gemini.

Learn more on Google’s AI Milestones.

2022

Imagen & Phenaki
Imagen and Phenaki go live, two models that use different techniques to generate photorealistic images and videos from a text description. Both are possible today in Gemini.

THE PRESENT

The Gemini Era

Today, people around the world are creating, building and learning with the most capable and general model we’ve ever built. Gemini is Google’s most advanced AI and the largest science and engineering project we’ve ever undertaken. To keep things simple, we’ve named both the AI model itself and the product you interact with »Gemini«.

Gemini allows people to collaborate directly with generative AI to supercharge their productivity and creativity. Users have turned to Gemini to enhance research and analysis, streamline the way they work, prepare important meetings, but also get more out of their hobbies. Following, you’ll find many examples of how you can use Gemini in your everyday life.

The illustration shows a blue question mark.

Ask complex questions
Want to get an impartial idea of the current challenges of the EU domestic market or understand how energy supply is organised in Europe? Gemini is grounded in Google Search so you can ask it about anything. And, if you’re unsure, the ›Double-check‹ feature allows users to explore supporting and conflicting information and further combat hallucinations.

The illustration shows a person in a green pullover with icons projecting from his hands.

Create images and videos in seconds
In just a few words, generate images to illustrate complex ideas, create unique visuals for social media or explainer videos for when you want to get a point across. Once generated, you can instantly download or share with others.

The illustration shows a person with a laptop and headphones

Your personalised AI research assistant – NotebookLM
Named one of TIME Magazine’s Best Inventions of 2024, think of NotebookLM as a virtual research assistant that is entirely based on the notes and sources you select. This effectively creates a personalised AI that’s versed in the information relevant to you. For example, you could upload different legislative proposals and ask for it to analyse the differences, identify insights or spot common trends. You might also upload your source material on a topic relevant for a specific meeting. NotebookLM can then create a polished presentation outline, complete with key talking points and supporting evidence – and even generate a podcast-style audio discussion on the topic. Because NotebookLM is grounded in information you trust, it minimises the risk of AI hallucination. Furthermore, NotebookLM does not train its model on user-uploaded sources, queries or responses, so your personal data stays private to you.

The illustration shows a man in a yellow jacket holding a notebook in his hands doing research.

Condense hours of searching with Deep Research
Sift through hundreds of websites, analyse the information and create a comprehensive report in minutes. For public sector professionals, this tool can automate the time-consuming aspects of research. For example, when developing policy proposals or preparing responses to public inquiries, you and your team could use Deep Research to quickly compile relevant data and insights. This gives you more time for analysis and decision-making, rather than the manual collection of information. Lastly, you can transform your Deep Research reports into engaging, podcast-style audio and quizzes to test your understanding or infographics.

The illustration shows a woman talking.

Talk it out with Gemini Live 
Brainstorm ideas out loud or practise interview questions with this conversational feature. For instance, Gemini Live can serve as a stand-in practice journalist – having come up with a list of relevant, hard-hitting questions that you might face.

The illustration shows different colorful apps on a smartphone.

Get help with tasks in multiple apps at once
Gemini can connect to other Google apps like Gmail, Google Calendar, Google Maps and YouTube to help you find or do what you need without switching between apps. Like getting summaries from your Gmail or organising your Calendar.

A person in a green jacket sits on a bench with a laptop on his knees.

Write in less time
Go from a blank page to a finished product faster. You can use Gemini to generate first draft reports, white papers, emails, translate or summarise text and even to get feedback on things you’ve already written.

The illustration shows different symbols - like a book, a pen or a light bulb - in a small bag.

Build custom experts
Gems let you save highly detailed prompt instructions and upload files for your most repeatable tasks so you can save time and focus on deeper, more creative collaboration. Gems can range from an interview coach or brainstorm partner to a coding helper.

The illustration shows a person in a purple pullover trying to reach different symbols around her.

Dive into large files
With a long context window of 1 million tokens (tokens are a unit of text or code that LLMs process), Gemini in Google AI Pro can understand and analyse whole books, lengthy reports and more with uploads of up to 1,500 pages or 30,000 lines of code, all at once.

THE FUTURE

Lila Ibrahim on the Future of AI

From personalised education to helping solve global challenges – Lila Ibrahim, COO of Google DeepMind, explains how AI can help improve our future.

What »magical,« unimagined ways do you see people using future generative AI products in their daily lives, work and creative pursuits?

Lila Ibrahim: What inspires me most is how AI can help personalise education. I’ve witnessed firsthand with my teenage daughters the positive impact of AI tools that explain challenging concepts using text, audio, video and graphics. Now imagine every teacher having an AI assistant and every student working with a personalised tutor that helps them to learn about a subject or master a new skill by adapting to their pace and style. I’m particularly passionate about our role in responsibly building that future.

What is Google’s core philosophy for designing AI products that empower users and enhance their experiences?

Lila Ibrahim: We believe AI systems should be capable and reliable tools designed to partner with people to solve real-world problems. A powerful example is WeatherNext, our family of AI models that produce state-of-the-art weather forecasts that can help meteorologists warn people of major storms up to 15 days in advance and explore impact paths. This offers communities critical time to prepare. WeatherNext embodies our approach to building AI that could empower people to make better, faster decisions.

What’s the next transformative frontier for generative AI that will directly impact how users create and achieve, and what product advancements excite you most?

Lila Ibrahim: We’re entering the »agentic era« where advanced AI systems move beyond being creative tools to become collaborative partners that can take action on behalf of – and with – people. These tools will soon handle complex but mundane tasks, like planning travel or grocery shopping, giving us back our most valuable resource: time to pursue more creative endeavours or to spend with our friends and families. This is why our work on Project Astra is so exciting — it represents our vision for a universal AI agent that empowers people in their daily lives.

A woman sitting at a table, smiling into the camera.

What role do open-source initiatives and collaboration with academia play in the further development of AI at Google? 

Lila Ibrahim: One thing I’m proud of is how our team builds meaningful relationships beyond traditional stakeholder management. AlphaFold is one of our best examples of open-source technology making science more accessible: more than 2.5 million people across 190 countries use AlphaFold in their research work. As part of ensuring equitable access to AI in science, we hosted international researchers with EMBL-EBI, a leading research organisation that provides freely available public biological data resources and bioinformatics services to empower scientists to advance their research and address local, regional and global scientific challenges.

What role do international collaboration and regulation play in shaping the future of AI?

Lila Ibrahim: AI is an inherently borderless technology that no one company, organisation or country can manage alone. That’s why we are very engaged in international dialogues to establish shared principles rooted in democratic values and a deep commitment to safety. Ultimately, our goal is to create a framework that builds public trust and ensures this powerful technology is developed and deployed responsibly for the benefit of everyone. That requires a global effort from the start.

If you were to send a message to the world about the promise of AI and Google’s role in realising it, what would be the single most important thing you’d want people to feel excited and optimistic about for the coming decade?

Lila Ibrahim: I’m most excited about how AI can accelerate scientific discovery. It is the modern-day telescope or microscope – a powerful new tool that gives us a fundamentally different view of the world. We saw this with AlphaFold, which is helping scientists to co-create solutions for everything from vaccine development to plastic pollution. Our mission is to continue building these tools and placing them in the hands of the world’s brightest minds to unlock our understanding of the universe and solve humanity’s biggest challenges, responsibly.

Illustrations: Bratislav Milenkovic; Photo: Google

FORWARD - European Perspectives on Tech & Society - No. 02 (EU)