Google Cloud: “It’s Been a Hectic Year with Generative AI”

We interviewed Javier Martinez, Director of Engineering at Google Cloud in Spain, on the eve of Google Cloud Summit 24 in Madrid.

On the eve of the Google Cloud Summit 2024 in Madrid, we interviewed Javier Martinez, Director of Engineering at Google Cloud in Spain. He shares with us the main developments of the event and highlights the advances in artificial intelligence, with multi-modal models such as Gemini, designed to improve efficiency and scalability in enterprise environments.

Martínez also addresses the economic impact of AI in Spain and the importance of data sovereignty in accordance with European regulations. Finally, he stresses the importance of ethics in the development of AI and technological training for users and companies, promoting a fairer and healthier society through technological innovation.

Google Cloud Summit 24

– What are the main novelties of the Google Cloud Summit?

We come from two big events worldwide: Google Cloud Next and Google I/O, Google Cloud Summit is our annual event, where we present all the announcements about the technology we develop in Spain and share it with our customers.

The most important thing for us is to present our customers with references and use cases. This year we are going to share these experiences with customers such as CaixaBank and Telefónica, who will explain how they are implementing our solutions.

It has been a frenetic year with everything related to generative AI. The focus of the Google Cloud Summit is precisely the business environment, and we are going to show how we apply these technologies in the business world.

– What about the artificial intelligence developments you are presenting or have presented at these big conferences?

The first thing is to talk about Gemini, a family of models we announce as LLM (Large Language Models). We started to develop these models in December, and the announcements have been made practically every week. The truth is that it is difficult for us to keep up with everything that is being launched. Today, we have multimodal models, allowing us to understand not only text but also images, video and audio, which have been available to our clients for a few months now.

In fact, one of the cases we are going to share is how to apply this multimodality, not only to understand text but to have a broader understanding. There are already clients, for example, who are applying it to understand audio from a contact centre. It is not only about the transcription of words but also about the tone when talking to a person, which can reveal discrepancies.

Recently, we have announced a new version of this whole family, which is the Gemini Flash, a model that is all about efficiency. We’re seeing that a lot of customers are already moving proofs of concept and pilots that they’ve been doing over the last few months and they’re running into limitations, particularly with other vendors, in terms of the performance, the usability and the cost of these solutions when you have to apply them at scale.

What we have done is to launch a version of Gemini that is precisely designed to scale. It is a version that is ten times less expensive, with a slightly lower performance, but sufficient for many cases in terms of the quality of the responses and that, in addition, has much less latency, that is, it responds faster. So, we are seeing that there are customers who want to apply generative AI on a massive scale and they were asking us for a different model. I think at the enterprise level, this is one of the main references.

In terms of models, then we are applying these technologies within Workspace. We also call it Gemini, but in this case, it’s a suite of wizards that appear in all our tools. We have Workspace as collaboration tools and there we have the ability to assist workers with generative AI tools. We have the same for Cloud users as well.

There is a very interesting use case which is for developers, for people who are programming, doing migrations or updating applications. You have a wizard that is able to generate code, we call it Code Assist. This wizard helps with documentation or testing. Normally, it is to improve productivity in office automation, something that is perhaps more common, and we are seeing it in the last few months. Here we focus more on the business world, where these more advanced users also have these kinds of tools.

– I would like to comment on the issue of the economic impact of artificial intelligence in Spain, on the data extracted from the Implement report commissioned by Google.

The report tries to measure the impact of AI as a whole, not just what we do at Google, but everything that is happening on a social and business level. In it, it estimates between 100 and 120 billion euros, which is very significant.

What we are seeing is a shift in how this technology allows us to generate new business and be more efficient in other types of business, providing new capabilities and tools that are applicable in all contact centres, but it really applies to many areas.

This technology has been a revolution, I would say almost social, as we are seeing both business and personal use cases. For example, having a chatbot to help my kids do their homework is great, but companies are looking at how to apply this to the business world and get a return.

Public administrations are also being an important focus because this technology can help improve citizen service. With all this, the new capabilities we are going to develop in the coming years will have an even more substantial impact.

– Within the use of Gen AI, there is a demand for data sovereignty, both by European legislation and by the companies themselves who want to know where their data is. How does Google address this demand?

There are several dimensions to this point. The first is that Google is probably the only player in the industry that is able to cover the full range of this world. Because we have researchers who are defining the models themselves and teams working on finding the data to train the models, ensuring the quality and copyright of that data, and training the models. This is something that only a few companies in the market do.

Then, once we have the models, we apply them to our own products. For example, in Workspace we apply those same models that we have developed to the user interface of these office applications. And also, we have the platform to serve the business world, that is, the platform for any company that wants to develop technology and apply that generative AI to their business processes, to their software developments, to their solution developments, they can do it. And if you look at the industry, there is no one else doing all this. So, I think in that sense we have a complete stack.

Accountability in what we do around AI is one of the pillars. We are being very conscientious and, in fact, you will see that we are in many cases the first to launch products in the market because we make sure that the products have the quality and comply with the regulations in Spain and in Europe. We are under the GDPR framework, talking practically with the protection agents to make sure that quality and responsibility in everything we do. So, let’s say that comes to the forefront of us as a company and the capabilities that we have.

Another very important step is where is my data? We talk about what happens to the information. We were the first hyperscale provider to set up a Google Cloud region in Spain. So, we already have the guarantee from two years ago, since the region is up and running, that the data can be hosted here. Our generative AI models have the guarantee that what is called “Data Rest” is done, that is, where the data is stored permanently in Spain. We are already offering these guarantees to our customers.

In addition, we have a unique offer in the digital sovereignty market. This is an alliance, a partnership we have with Minsait, in which Minsait can act as custodian of the encryption keys. Let me explain. Any information that is stored in the Google Cloud is encrypted, yes or yes, we encrypt it. In this case, what we do is that the encryption keys are outside, we do not touch them, we do not control them, so Minsait can provide the service to any of our clients to manage those keys and ensure that the data is only accessed when the user, the public administration or client X has that information.

– Given this high demand for data hosting, do you already have the local infrastructure in place to meet this demand, or do you have plans in the short and medium term to expand this local infrastructure?

These are long-term projects. When you start thinking about a region, as we did with Madrid, you make a minimum five-year plan. We have already made a growth projection and I can tell you that we are progressively growing the user base we have there. We plan to continue growing. So, Google’s investment is considerable. We can share specific data with you later, but in our last results announcement, we mentioned an investment of around 12 billion dollars in infrastructure.

There is a tremendous focus on growing this infrastructure, not just in Spain, but around the world. As you say, this whole generative AI revolution needs a lot of computation behind it. That’s why we are expanding our capabilities and our data centres. It’s not just about the data centres, but also about equipping them with servers. We have invested in creating our own TPUs and chips, for AI computing. We’ve been preparing for these massive computing needs that we’re seeing today for several years now.

If you think about the scale at which Google has been operating for years with services like the search engine and Google Maps, you’ll see that we are already prepared for these kinds of demands.

– Do you have artificial intelligence training programmes planned for both customers and single users?

Absolutely. We have diverse training plans because they cover a broad spectrum. Not only in artificial intelligence, because in many cases you have to start with a more fundamental technological base, not just infrastructure. We have training in cloud technologies to enable the base technologies. In some cases, it is aimed at people without a technical background who want to retrain and evolve, and in other cases, it is aimed at people with a degree in computer science who want to specialise.

We also offer specific training in generative AI that covers how to apply these tools, improve productivity and acquire new skills, both for people who want to use AI in their daily lives and for those who want to be the next to program and build AI tools. This training is available both online and in collaboration with conventional educational centres, both public and private. The offer is quite wide because, although technology is fundamental, we need a lot of talent to dynamise the sector and have a significant economic impact.

– What is your ethical vision on the use of artificial intelligence, and could you describe some of these initiatives and their results so far?

On the topic of ethics and the implications of artificial intelligence, it is crucial to us. In fact, we were one of the first companies to create a set of ethical principles. We define which projects we want to apply AI to and which ones we don’t want to be involved in, even though they may represent business opportunities. We have decided to be ambitious in this regard and not to get involved in certain topics for ethical reasons.

Within our developments, there are several approaches. For example, with the issue of data, we work to avoid bias. We have teams dedicated to bias control and to monitoring the applications we develop. It is fundamental for us that technology has a positive purpose and benefits everyone.

I believe our company has shown leadership in addressing these challenges. One example is the launch of a technology called SynthID, which enables watermarking of images and videos generated by artificial intelligence. This is an innovative step that we are proposing as an open standard for the entire industry, with the aim of avoiding problems such as deepfakes. We are committed to the correct and ethical use of technology, and we continue to look for ways to innovate in this area.

– Do you think the role of AI will contribute to a fairer and healthier society?

Well, I think AI opens up a lot of opportunities and obviously brings about changes in the tools we use, which will give us new capabilities. Here we talk about the third industrial revolution. What we have seen in the industrial revolution is how our work was transformed and how we had new tools that gave us new capabilities.

Although it depends on everyone, and also on what governments decide and control, the application of all this should evolve in the right way. We believe that technology is there to help us, and among all that we have, AI plays a key role in helping us.