AI has its own world to live in, it shocked me!
Can you believe it? AI actually has its own world, where it lives, learns, and explores.
And for ordinary people, it may be a very big opportunity. What's going on?
Recently, Nvidia's chief scientist Jim Fan and others integrated GPT-4 into "Minecraft" (Minecraft) and proposed a new AI agent Voyager.
Not only does it outperform AutoGPT, but it also enables full-scenario lifelong learning in games.
Compared to the previous SOTA, Voyager gets 3.3x more items, travels 2.3x more distance, and unlocks key skill trees 15.3x faster.
This means we are one step closer to AGI.
So, will future games be played by large model-driven NPCs?
After Voyager is connected to GPT-4 for learning, it does not need human manipulation and assistance at all, and can explore, learn, and become talented by itself.
It not only masters basic survival skills such as digging, building houses, collecting, and hunting, but also learns to explore openly by itself.
It will go to different cities by itself, pass by oceans and pyramids, and build portals by itself.
Driven by itself, it constantly explores this magical world and expands its items and equipment.
How big is the potential of digital life?
We only know that Voyager is still exploring non-stop in Minecraft, constantly expanding his territory.
Previously, a major challenge in the field of AI was to build embodied agents with general capabilities, allowing them to independently explore and develop new skills in the open world.
In the past, reinforcement learning and imitation learning were used in the academic circles, but these methods often performed poorly in terms of systematic exploration, interpretability, and generalization.
The emergence of fully intelligent large language models brings new possibilities.
LLM-based agents can use the world knowledge contained in the pre-trained model to generate consistent action plans or executable strategies, which are very suitable for tasks such as games and robots.
However, these agents still cannot get rid of the defects of being unable to learn for life, unable to gradually acquire knowledge and accumulate it over a long time span.
The most important significance of this work is that GPT-4 has opened up a new paradigm: in this process, "training" is performed by code rather than by gradient descent.
The "training model" is Voyager's iterative construction of a skill code library rather than a floating-point matrix. With this approach the team is pushing the limits of gradient-free architectures.
Under such conditions, the trained intelligent AI already has a certain level of lifelong learning ability at the same level as humans.
For example, a Voyager who finds himself in a desert instead of a forest will know that learning to gather sand and cacti is more important than learning to gather iron ore.
It can not only define the most suitable tasks for itself according to the current skill level and the state of the world, but also continuously improve the skills based on feedback and save them in memory for the next call.
So how close are we to silicon-based life?
In short, Voyager is the first LLM-driven embodied agent capable of lifelong learning.
It can use GPT-4 to continuously explore the world to develop increasingly complex skills and always make new discoveries without human intervention.
Voyager is a starting point for general agent development without tuning model parameters.
We can see from this that the emergence of Voyager has brought new possibilities to the field of game development and artificial intelligence.
Therefore, investing in the field of game development and artificial intelligence could be a money-making opportunity.
For the average person, there may be ways to invest in the game development and artificial intelligence fields.
One way is to invest in stocks of companies focused on game development and artificial intelligence.
This can be done by buying shares of these companies or by buying index funds that track these industries.
Also, consider investing in startups that use artificial intelligence for game development.
