David Sanz, responsible for data analytics, AI and emerging technologies of KPMG in Spain, attends the revolution that artificial intelligence (AI) is causing organizations. Given the prodigious jump that this technology supposes, the expert proposes an adequate management of the risks and ethical implications of AI, beyond normative compliance. With a view to the future, as explained in an interview for the election.
AI agents begin to talk in organizations. What will be the following?
In KPMG we have been pioneers to implement this type of architecture of AI agents within our house, with more than 500 professionals overturned in the matter. And since the middle of last year we are also sharing it with our clients. After the agents we can expect the so -called LCM, (Large Conceptual Models, large models of concepts). Faced with the best known LLM (large language models), the American Meta and the China Manus have already recently launched their LCM. That is, instead of using languages based on concepts, they are based on concepts. It is a step that approaches us to general artificial intelligence (AIG).
When can we expect general artificial intelligence?
It is the question we all ask ourselves. I find it pretentious to estimate a date. I do not see it in the short term and it is hard for me to intuit it in the medium term. However, given the speed in the construction and creation of new models, and in innovation around artificial intelligence, that AIG is closer to what was thought three years ago, when it was located at 2050. The deadlines have now been shortened and it is estimated for three or four years.
Will it depend much on the evolution of the processors?
I think that computing capacity is a problem for the current approaches we have of artificial intelligence. These have to change radically, because they cannot be traditionally based on current mathematics. The approaches have to be different. And the Facebook and Manus movement brings us a little more.
It should be followed closely to Manus ..
No one expected the impact of Manus, a Chinese model close to the capabilities OpenAi offers with Operator. It is a startup that flew under the radar and, suddenly, it has launched a surprising. Few suspected that a startup could offer hundreds of use cases.
How do you imagine that general artificial intelligence?
I set an example. When an employee has to make a report, nobody tells him how he has to do it. Infrastructure and tools are provided, as well as training. And you ask him what and the employee is already in charge of reasoning autonomously to execute it. That is not the agents of AI. These are able to reason, plan and execute tasks, but need human supervision. They also need clear rules, where an ethical framework and a risk control scenario are established. For its part, the general AI learns autonomously, is capable of having some conscience and distinguishes the ethically correct or incorrect. You can make decisions and modular just like humans do …
How does quantum computing participate in that equation?
With quantum computing we are jumping several decades. We are working with customers around Quantum Computing. In Spain we have executed several projects, and we have an international hub with abundant ‘Expertise’ on the matter. Without a doubt, the ability of quantum computers opens a world of possibilities to which we can now not aspire. And not only because of the greater processing capacity, but because of the reality that this technology imposes. That will carry risks associated with cybersecurity. We need to define postcuantic cryptography. That will help us address issues such as parallel the resolution of the different problems to discover what is the optimal response within many possibilities.
But the great technological ones have the quantum challenge in their respective agendas …
From the quantum perspective, advances are significant. A few days ago, Microsoft presented a new chip around Quantum that is a significant impact on the approximation we handle. But I think there is still space to identify use cases. The works we have done in Spain, through quantum computers simulators, helps us identify the impact and potential. We have a lot to learn.
Could China’s artificial intelligence overcome that of the US?
Until recently, that game was played only in the United States, which was where the greatest knowledge was generated. And it is logical, because hyperscareres such as Google, Amazon or Azure, offer the availability of these capabilities within their clouds. And the fact is that no one was looking at China, where similar models are being generated, within their cloud of Alibaba and we had few references. Also, for a more geopolitical theme, in the world of business it is less afraid to incorporate American technology than China. There are reluctance, although it is true that you can make your own model based on China.
How do you explain Depsheek?
China was at a time where I could use – and used American microchips. When microprocessor export restrictions were incorporated, China was in a situation that has allowed it to turn significant events. First he tried to develop his own chip, certainly with little success. And then he did things in a more intelligent way. They turned the need by virtue and developed a new way of doing things. And so Deepseek appears, which offers a different way of generating these types of models and also with permeability. Unlike the United States, China opens the code to all and also to everything around, the entire ecosystem that is needed to optimize training. China changed the rules of the game and positioned itself as a new player. But new Chinese models that are flying under the radar have also emerged.
And what about Europe and AI?
In Europe we are strong in the Risk Management of AI, because we are very guarantee. We have a very detailed regulation around how artificial intelligence has to be managed, something that does not exist anywhere else in the world.
#David #Sanz #KPMG #Chinas #offers #level #similar #elements #exceeds
