Join daily and weekly newsletters to obtain the latest updates and exclusive content to cover the leading artificial intelligence in the industry. Learn more
Slide in the TV news program at CBS Weekly 60 minutes Yesterday night (I also participated in YouTube hereI showed an inner look at Google DeepMind and the vision of its co -founder Nobel Prize CEO, legendary artificial intelligence researcher Dimis Hasabis.
The interview tracked the rapid progress of DeepMind in artificial intelligence and its ambition to achieve artificial general intelligence (AGI)-automated intelligence with a diversity that resembles a person and a supermarket.
Hassabis described the path of artificial intelligence today as “the curve of the ASIs of improvement”, and feeds it increasing the attention, talents and resources that enter the field.
Two years later 60 minutes Chatbot, Hassabis and DeepMind are now following more capable systems that are not designed not only to understand the language, but also the material world around them.
The interview came yet The cloud conference in the next Google 2025 earlier this monthWhere the search giant presented a set of artificial intelligence models and new features centered around the multimedia AI Gemini 2.5 model family. Google came out of this conference, which appears to have taken the initiative Compared to other technology companies in providing strong artificial intelligence for institutions use in the most affordable price points, exceeding Openai.
More details about Google DeepMind’s ‘Project Astra’
One of the focal points of the part was the ASTRA project, the next generation of Chatbot from DeepMind that exceeds the text. Astra is designed to interpret the visual world in actual time.
In one of the experimental, the paintings were identified, the emotional states were concluded, and created a story about the Qadous plate with the line: “Only the flow of ideas that move forward.”
When asked if he was bored, ASTRA answered carefully, and revealed a degree of sensitivity to the tone and personal differences.
Product Manager Bibbo Shu has confirmed the unique design of ASTRA: AI can “see, hear and chat about anything” – a noticeable step towards the embodied AI systems.
Gemini: Towards AI, which can be implemented
The broadcast also appeared twinDeepMind is trained from artificial intelligence not only on the interpretation of the world but also to dispose of it – coordination tasks such as tickets and online shopping.
Hasabis said that Gemini is a step towards AGI: Amnesty International with a human -like ability to move and work in complex environments.
the 60 minutes The team tried a preliminary model included in the glasses, indicating visual recognition in actual time and vocal responses. You can also hint on a coming back from the leading but eventually outside early reinforced reality glasses known as Google Glass, That appeared for the first time in 2012 before Retired in 2015?
While like Gemini Model versions such as Gemini 2.5 Pro or Flash Not mentioned in this sector, the broader ecosystem of Google’s artificial intelligence has recently presented these models to use institutions, which may reflect parallel development efforts.
This integration supports Google’s increasing ambitions in Applied Ai, although it is located outside the scope of what has been covered directly in the interview.
AGI once 2030?
When asked about a schedule, Hassabis Agi can arrive after 2030, with systems whose environments understand “in very accurate and deep ways.” He suggested that such systems can be included smoothly in daily life, from wearable devices to home assistants.
The interview also covered the possibility of self -awareness in artificial intelligence. Hasabis said the current systems are not conscious, but future models can show signs of self -understanding. However, he emphasized the philosophical and biological gap: even if the machines mimic conscious behavior, they are not made of the same “issue of carbon” as human beings.
He also predicted the major developments in robots, saying that breakthroughs may come in the next few years. Part guarantee robots that complement tasks with mysterious instructions – such as identifying a green mass that are formed by mixing the increasing yellow and blue capabilities – in physical systems.
Safety achievements and fears
This sector has reviewed the historical DeepMind achievement with Alphafold, an artificial intelligence model that expected more than 200 million protein.
Hassabis and his colleague John Jumper won the 2024 Nobel Prize in Chemistry for this work. Hassabis emphasized that this progress can accelerate the development of drugs, and perhaps reducing time tables from decade to weeks only. “I think one day, we may be able to treat all the disease with the help of artificial intelligence,” he said.
Despite optimism, a account expressed clear concerns. He pointed to major risks: The misuse of Amnesty International by bad and increasingly independent actors of systems outside human control. He stressed the importance of building in handrails and value systems – which leads to teaching artificial intelligence as a child. He also called for international cooperation, noting that the impact of artificial intelligence will touch every country and culture.
He said: “One of my great fears is that the race for artificial intelligence dominates it may become a race for safety.” He stressed the need for senior players and national countries to coordinate ethical development and oversight.
The part ended with meditation in the future: a world in which artificial intelligence tools can transform almost every human endeavor – and in the end the reshaping of how we think about knowledge, awareness and even the meaning of life. “We need wonderful philosophers to come … to understand the effects of this system.”
Source link