Demis Hassabis, CEO of Google's DeepMind AI unit, announced the development of Gemini, an advanced AI system. Gemini's following algorithm is expected to outperform OpenAI's ChatGPT by employing approaches developed from DeepMind's historic victory in Go. 

This discovery is a watershed moment in AI, promising more outstanding capabilities and novel developments. Join us as we go into the specifics of this breakthrough advancement and its possible implications for the future of AI.

AlphaGo

For those unaware, DeepMind created a computer programme called AlphaGo that bested the human world champion at the strategy board game Go. AlphaGo also employed a Monte Carlo tree search (MCTS) algorithm to find and select its moves in a game of Go. You can see how the consumer-facing solution developed by applying these methods to Gemini could compete with ChatGPT.

AlphaGo employed a technique known as reinforcement learning, which allows the software to learn via trial and error while receiving feedback on its performance. Gemini can plan and solve difficulties more effectively using this strategy. The next significant hurdle for language models is to communicate more independently with the internet and computers, and DeepMind is focusing on that with Gemini.

Gemini AI

Google introduced Gemini during their I/O developer conference. Google stated at the ceremony, "Gemini was created from the ground up to be multimodal, highly efficient at the tool and API integrations, and built to enable future innovations." Google claims its Gemini AI will be available in various shapes and capacities. It demonstrates its adaptability and allows for fine-tuning to specific requirements. Google is sure that Gemini AI will win ChatGPT for reasons other than financial gain. Instead, the heart of the AI itself is being scrutinised. 

Why Gemini AI?

Google has no desire to create something that already exists. That is why Gemini AI is focused on something other than matching the GPT 4's performance. AlphaGo is the name given to the artificial intelligence that defeated the world Go champion. Google is also powering Gemini AI with the same algorithms that powered AlphaGo. As a result, Gemini will be a master planner and problem solver in the long run. AlphaGo was created using the reinforcement learning process. This strategy causes the programme to attempt to perform a task several times. It can improve based on the feedback it receives.

Problem-solving functionalities

Gemini researchers are pushing the boundaries by including problem-solving functionalities inspired by their expertise in robotics and neuroscience rather than focusing on the typical LLM method of feeding models with data. DeepMind has had an uphill battle to catch up to ChatGPT's rapid growth in popularity since its inception. Hassabis, though, seems undeterred and insists that DeepMind and Google's primary AI unit, Brain, will merge successfully.

Concerned consumer protection groups in the EU and the US are encouraging regulators to take preventative action to limit the risks from AI misuse, given the threats AI poses to Web3, finance, media, and health.

Conclusion

Despite ChatGPT's dominance, Demis Hassabis, CEO of Google DeepMind, recently told Wired that Google's Gemini AI system would eventually outperform GPT-4. Google intends to combine Gemini's text-response capabilities with those of AlphaGo.

Hassabis, on the other hand, is not advocating for AI development lightly. He was one of several professionals and public personalities who signed a statement warning that AI could represent a hazard similar to nuclear war or a pandemic. He also warned in the article that there is little time left to establish protections if AI continues at its current rate.

Image source: Unsplash