Reflection: creating the cutting edge AI that could help you do your work without any inputs

Google DeepMind AI researchers Misha Laskin and Ioannis Antonoglou just founded the first Web3 virtual personality platform that allows you to create your own perfect simulation clone, capable of autonomously communicating, learning, and growing within the platform. Imagine a world where you won't have to be glued to your screens anymore, a world where you can have your "reflection AI" that would revolutionise the way you interact with your friends.

Offloading your mental work to AI Agents

Artificial Intelligence Agents is a software program that can interact with its environment, collect data, and use the data to perform self-determined tasks to meet predetermined goals. An example of an AI agent on the market would be AlphaGo, where an AI program that is programmed to beat professional players in the strategy board game of Go. This demonstrates an AI agent that possesses deep expertise in a specific task. However, AlphaGo is probably not the deepest agent that has been built as it is only proficient in one task. So it can play Go, but it cannot play tic tac toe.

Then there are other large language models like ChatGPT, Gemini and Claude that are broad and haven't been trained for the agency. This means that they are able to give us general information based on the prompt provided to them, but sometimes these Ai models go off the rails. They will either provide you with information that is not accurate. Other times you might find that they might be unable to do something as well as they used to before.

Creating your unique persona

Reflection is aspiring to be the first Web3 virtual personality platform, transcending existing concepts of virtual avatars and identities. You can create your own personal simulation clone on the platform, who has the ability to autonomously communicate, learn and grow within the platform. By putting photos, voice notes, and personal information about yourself, you are helping the system to create an AI personality that aligns more closely with your own personality and habits.

By continuously learning and training automatically, your AI personality can improve in various aspects, such as knowledge and social adaptability. Ultimately, you can even delegate practical tasks to your AI virtual personas, such as using them for interviews or social events.

Through the use of this simulation, you can use it to predict your own behaviour in different situations as a way to provide yourself with decision-making guidance. When you are facing various challenges and decisions in real life, you can utilise this platform to foresee potential outcomes in advance, thereby making wiser decisions.

Users can also interact with the virtual AI personalities of others, chat, socialize, and even combine their personal models with general or professional models to train various expert versions of themselves, even transforming into super AI agents and digital AI avators to assist in handling various tasks.

AI agents promise to execute difficult tasks, like booking an appointment or updating Saleforce. Reflection is trying to build a new superhuman general agent that can automate knowledge work done on a computer. While discussing the project, Laskin mentioned that an ideal universal agent needs to be broad; it needs to know how to do many things, can handle many inputs, but it also needs to have depth in the kind of task complexity it can achieve. While there are many different AI agents on the market, like AlphaGo and Google's Gemini. While AlphaGo possesses deep expertise on a specific task, it can only do one task-play Go. Large language models like Gemini, Claude and ChatGPT, on the other hand , are able to accomplish many tasks, but they do not possess the deep expertise that is required to accomplish the task well.

Exiting DeepMinds to create a rival platform

The company's cofounders, Misha Laskin and Ioannis Antonoglou, left Google's DeepMind to launch their own startup. Laskin was working as an AI researcher at the Berkeley Artificial Intelligence Research Lab and most recently worked at Google DeepMind, the company's AI research lab. Loannis, on the other hand, was one of the creators of AlphaGo. While working at DeepMinds together, Laskin and Antonoglou felt as if they reached a limit of what they could do in the company. For this reason, they decided to break away from Google and try to create their own AI agents in their own way.

Laskin's socials:

LinkedIn | X

Antonoglou's socials:

LinkedIn |

Paving the future of AI agents

Personally, I find the proposal that Reflection AI offers very fascinating. In a time of technological advancement, who won't want to have an AI machine that can help us so offload and do things that would be so tedious for us to do it ourselves. We are also looking into a world where Ai is no longer just limited by the prompts. Rather, this advanced AI has the capability to go through the iterative process of looking at something, asking questions, reasoning what the issues might be, and even changing the approach to accomplish goals in the world.