OpenAI’s ChatGPT has revolutionized the AI landscape with its exceptional language processing abilities. Now, Google-owned DeepMind asserts its forthcoming language model, Gemini, could match or even surpass the prowess of ChatGPT, leveraging techniques from its renowned AlphaGo system.
Drawing from AlphaGo’s strengths
DeepMind CEO, Demis Hassabis, highlights the purpose behind Gemini. The design combines the strategic prowess of AlphaGo-type systems and the language abilities of large models like ChatGPT. In a discussion with Wired’s Will Knight, Hassabis hinted at distinct innovations soon to distinguish Gemini.
The conception of Gemini appears to hinge heavily on the principles of reinforcement learning, a technique where an AI system is “rewarded” for specific behaviors while others are discouraged. This method aims to instruct the AI on which actions to undertake given various situations.
The Power of reinforcement learning
Reinforcement learning has already proven instrumental in enhancing the performance of language models. It’s a critical part of how systems like ChatGPT interact with prompts. DeepMind is well-known for its groundbreaking work in reinforcement learning. The company now appears ready to apply these insights to generative AI.
Gemini isn’t DeepMind’s first attempt at language models. Previously, the company launched Sparrow, a chatbot with a specific design to minimize the likelihood of delivering “unsafe” or “inappropriate” responses. Earlier this year, DeepMind discussed plans to roll out a private beta version of Sparrow. However, it’s currently unclear whether these plans are still underway.
A new chapter in generative AI
DeepMind arguably marks Gemini as its boldest venture in this domain to date. According to initial reports, Google’s top-level executives, including the senior AI research executive Jeff Dean, are actively participating in the project. The inception of this initiative seems to stem from the challenges faced by Google’s chatbot project, Bard, in its attempt to keep pace with ChatGPT.
As investor and consumer enthusiasm for generative AI continues to rise, so does the competition to lead this evolving market. According to Grand View Research, the future is promising. It predicts the generative AI market, inclusive of text-analyzing AI like Gemini, could witness an impressive 35.6% surge by 2030, potentially amassing a worth of $109.37 billion.
We are keen to hear your thoughts on DeepMind’s latest venture. Will Gemini successfully rival ChatGPT or even redefine the generative AI market? Please share your views and predictions in the comment section below!