In Brief
Ada, an AI agent developed by DeepMind, is a recent innovation that has nearly equaled human intelligence.
Like people, Ada is capable of picking up new skills fast and adapting to changes in her environment.
AdA can “live in the actual environment,” albeit imperfectly, and can learn new embodied three-dimensional jobs just as quickly as people.
The Trust Project is a worldwide group of news organizations working to establish transparency standards.
In a recent breakthrough, DeepMind has created an AI agent, Ada, that is nearly as intelligent as a human. Ada can learn new tasks quickly and adapt to environmental changes, just as humans can. This is a fundamental breakthrough that could have far-reaching implications for the field of AI.
DeepMind has achieved this breakthrough by making seemingly insignificant changes to its previous AI agents. First, it increased the size of the AI agent model. Second, the memory depth of AI agents was increased so that AI could rely more on its previous experiences. Finally, DeepMind increased the number of environments in which AI agents train. These small changes resulted in a big leap forward in Ada’s intelligence.
| Related article: Columbia University: AI will be able to think in the next two years |
Ada is not only smarter than previous AI agents but closer to humans in terms of its capabilities. This is a major breakthrough that could pave the way for even more advances in AI in the future.
Ada’s advancements are crucial as it is the first AI agent to learn when faced with new tasks in some complex 3D worlds, quickly learn a new task, and then figure out how to do it. For example, it can do it just by watching another intelligent agent complete such a task. And all of this happens around the same time for people.
To understand why this is such a big deal, let’s first look at how AI agents are typically very good at only one thing. For example, Google’s AlphaGo AI agent was specifically designed to play the game of Go. And it does so very well, even beating the world’s best Go player in 2016. However, if you asked AlphaGo to do anything else, it would be at a complete loss.
| Related article: Stanford Observatory: AI will lead to ubiquitous lies and irresistible propaganda in 2023 |
In contrast, Ada is not nearly as good as AlphaGo at playing Go. However, Ada is much better than AlphaGo at rapidly learning new tasks, and that’s because Ada was designed from the ground up to be an AI agent.
It won’t be a big stretch to say that Ada has reached the level of ChatGPT in terms of its capabilities. But:
- If ChatGPT’s “cleverly chatting” is becoming almost indistinguishable from people.
- AdA can, albeit imperfectly, “live in the real world,” adapting to new embodied three-dimensional tasks as quickly as humans.
Now there is only one question. To what extent will the ability to live in a 3D simulation be applicable to the life of an AI agent in our physical world? And if this skill is well tolerated by AI agents in our world, people will have to make room here.
Read more about AI:
Disclaimer
Any data, text, or other content on this page is provided as general market information and not as investment advice. Past performance is not necessarily an indicator of future results.
