19/01/2023 Recurso de agente de diálogo POLITICA INVESTIGACIÓN Y TECNOLOGÍA DEEPMIND

The co-founder and CEO of DeepMind, Demis Hassabis,has warned of the dangers to humanity that a technology as powerful as the one he himself has helped develop can entail, and has announced that they plan to start facilitating their own ChatGPT-style chatbot this year.

DeepMind is a subsidiary of Alphabet, parent company of Google, focused on the development of artificial intelligence technologies,with the aim of “solving intelligence problems, to advance science and benefit humanity”, as she herself explains on her website.

This company is responsible for MuZero, who learned to play chess, shogi, and Go, as well as 57 different Atari games without first knowing the rules; the Pythia algorithm, which deciphers texts from ancient Greece that have been damaged and are almost unreadable; and AlphaGo, who went on to defeat one of Go’s professional players with a 5-0 result in 2016.

In 2021, AlphaFold2, also from DeepMind, was highlighted by Science for having managed to solve the structures of 350,000 human proteins, 44 percent of all known human proteins. The company currently directs its machine learning technologies at nuclear fusion research.

Hassabis’s greatest ambition, however, is to create an artificial general intelligence (AGI), with an intelligence equal to that of human beings, with the capacity to think, learn and plan for the future, which contributes to solving the most difficult problems of humanity. But this technology, which for the researcher “will define an era”, not without risk.

These revolve around privacy violations, biases, identity theft, the distribution of false or erroneous news, among others, which have motivated, for example, the demand for specific ethical mechanisms for AI or in Spain, the creation of an Artificial Intelligence Supervisory Agency.

“I would advocate not moving fast and breaking things,” Hassabis says in an interview with Time. “When it comes to very powerful technologies, and obviously AI will be one of the most powerful in historyWe have to be careful,” he adds.

This concern was shared by the founder of DeepMind with former Google senior vice president Alan Eustace in 2013 when the tech giant took an interest in his artificial intelligence prototype. “He was thoughtful enough to understand that technology had long-term social implications,and I wanted to understand them before the technology was invented, not after it was implemented,” Eustace said.

His position contrasts with that of one of his rivals, OpenAI, co-founded by magnate Elon Musk, which in 2020 stood out for its language model GPT-3considered “the largest and most advanced in the world”, with 175 billion parameters, as rated by Microsoft.

OpenAI is also known for the generative art model Give her,which creates realistic images from a text description, and more recently by ChatGPT, who has surprised by the naturalness of his responses and by his ability to generate and link ideas and remember previous conversations. And the new challenges that he poses in education and cybersecurity, among other sectors.

DeepMind also has recent work including Chinchilla, for training computational models, and Sparrow, un chatbot similar a ChatGPT,which the company considers launching in the form of a private beta sometime this year.

Leave a Reply

Your email address will not be published. Required fields are marked *

error: Content is protected !!