Introduction
Google’s DeepMind is actively working on a new artificial intelligence (AI) system named Gemini. According to CEO Demis Hassabis, this next-generation AI is set to surpass the capabilities of OpenAI’s ChatGPT, marking a significant milestone in the advancement of AI technology.
Table of Contents
Google DeepMind’s Gemini AI
In a recent interview with Wired, Hassabis revealed that DeepMind is enhancing the capabilities of its large language models using techniques that previously aided in AlphaGo’s victory over a Go champion in 2016. Just like GPT-4, the technology underpinning ChatGPT, Gemini is also a large language model but equipped with additional features inspired by AlphaGo, such as advanced problem-solving and planning abilities.
“You can think of Gemini as combining some of the strengths of AlphaGo-type systems with the amazing language capabilities of large models,” Hassabis detailed.
Gemini’s Development and Cost
Creating Gemini is a mammoth task both in terms of resources and capital. The development cost could range from tens to hundreds of millions, matching the expense of GPT-4, as confirmed by OpenAI CEO Sam Altman earlier this year. This significant investment underscores Gemini’s expected impact on Google’s strategy to counter the competitive threat posed by ChatGPT and other generative AI technologies.
DeepMind’s well-established legacy in reinforcement learning, exemplified through its domination of simple video games, might enable Gemini to possess unique capabilities not seen in other AI models.
DeepMind’s Research Scope and Gemini’s Anticipated Benefits
DeepMind’s diverse research areas, spanning from robotics to neuroscience, provide a distinctive foundation for the development of Gemini. Despite growing concerns about the potential misuse or uncontrollability of advanced AI systems, Hassabis defends the need for further development. He cites the potential benefits in scientific discoveries, especially in health or climate change sectors.
However, the CEO also underscores the necessity of developing safeguards against potential risks, demonstrating the organization’s consciousness of its responsibility in this sphere.
Gemini’s Potential Learning Models
The DeepMind team is exploring innovative ways to enhance Gemini’s learning process. By creating new models and ways to process information, the team hopes to significantly expand its capabilities. Hassabis suggested that fields like robotics and neuroscience, as well as learning from the physical world, could serve as valuable resources to elevate Gemini’s abilities.
Challenges and Way Forward
As AI development accelerates, the primary challenge lies in understanding and controlling more capable AI. Hassabis expressed concern over the necessity to implement more research into testing these advanced systems. He stressed the importance of finding ways to assess their scope and performance. This crucial step aids in mitigating risks associated with the potential development of dangerous threats like malicious software.
The journey towards a more advanced AI is filled with uncertainties. However, plans are underway to improve AI while ensuring its manageable control. For now, all eyes are on Google’s DeepMind as it progresses with its Gemini AI.
Conclusion
As DeepMind continues its work on the Google Gemini AI, anticipation and curiosity in the tech world are growing. While surpassing the capabilities of ChatGPT is no small feat, if any company has the resources and expertise to achieve it, it would be Google’s DeepMind. As we wait for Gemini’s release, the tech world continues to harness the capabilities of current AI systems.
Link : Google Blog
