Google just announced its new AI language model called DeepMind. ChatGPT, which is still the undisputed champion of the AI wars, is really far ahead of the competition. While some other companies have already released their own models including Google, they are not that great. On their own, they are promising but compared to ChatGPT, not so much. The main advantage that OpenAI had was time. The latest model, ChatGPT 4 is way better than anything before it. But, can Google DeepMind be something different?
What is Google DeepMind?
DeepMind is a large-scale language model created by Google. It is based on Google’s own AlphaGo. The new AI model will have the ability to plan or solve problems as well as analyze text, according to DeepMind CEO Demis Hassabis. “At a high level you can think of Gemini as combining some of the strengths of AlphaGo-type systems with the amazing language capabilities of the large models,” Hassabis said.
The developers are using the Reinforcement learning technique to teach the AI. This means that the AI will be ‘awarded’ for doing certain things. This will reinforce the model’s behavior. This works exactly like how a pet is trained or how a child is taught.
This way, the learning time of the model can be significantly decreased. If things go according to plan, they just might catch up to ChatGPT.
However, things may or may not be this certain. Google’s last project, Bard, ended up being a failure. Google has also been working on multiple models which don’t really show much promise.