Artificial intelligence will be a danger to everyone if it gets into the wrong hands: Google co-founder Deepmind

Co-founder and CEO of Google DeepMind, Demis Hassabis, recognized as one of the pioneers in the field of artificial intelligence, shared a promising vision for the near future of this technology.

According to Hassabis, what has yet been seen in terms of language models in AI It is nothing but the beginning. Soon these systems will be able to “plan” and “act” in real-world situations.

“We are just beginning to discover what will be possible in next years” he declared, anticipating an era of innovation that could radically change how we interact with technology.

In the long term, Hassabis speculates that within ten years, AI systems will be able to generate their own hypotheses and solve problems without human intervention.

Although This can lead to humans not fully understanding the processes of AI models.Hassabis believes it will be possible to interrogate these systems about their solution methods, much like one would with another person.

– Medicine: An expert predicts that within two years, clinical trials for drugs designed by artificial intelligence will begin to develop.

These drugs will be particularly effective because they will be able to create compounds that specifically act on the desired protein.without affecting other molecules, which reduces possible side effects.

– Materials Science and Meteorology: Hassabis also projects important advances in materials science and expresses his desire to discover a superconductor that works at room temperature.

In addition, it highlights improvements in the accuracy of weather forecasts, with AI systems now able to perform calculations in seconds that would take hours on traditional supercomputers.

– mathematics: In the field of mathematics, Hassabis believes that artificial intelligence will contribute to the resolution of great conundrums that still have no answers, and will open the door to revolutionary discoveries in various fields.

A British scientist who started his career in the world of video games before earning a doctorate in neuroscience, believes that the development towards artificial intelligence capable of thinking like humans will take place graduallyand not by sudden advance.

Since 2016, Hassabis has been a key figure in the recent rise of artificial intelligence, starting with the AlphaGo project, which managed to win the complicated game of Go, challenging traditional algorithms.

Later, in 2020, he introduced AlphaFold, a tool that was able to solve the complex question of how proteins fold, a major scientific challenge.

AlphaFold can predict the three-dimensional structure of a protein based on its amino acid sequence in seconds, a task that could previously take up to five years of doctoral research.

“We managed to assemble almost 200 million known proteins,” Hassabis said, highlighting how his software solved a problem that had puzzled biologists for decades in one year.

In addition, Hassabis commented on the rapid public adoption of language models such as ChatGPT for activities such as programming or text summarization, even though these models are not yet perfectly accurate. He was surprised that this adoption took place “two years earlier than expected”.

The creator of Google DeepMind shares with other specialists in this field cautions about possible future risks of artificial intelligence.

Currently, he believes they are not an immediate problem given that AI systems are in the early stages of development.

However, he predicts that in the medium term, within three to five years, when AI systems acquire advanced planning skills and the ability to interact with the environment and solve complex problems with defined goals, the risk could increase significantly if said systems fall into the wrong hands .