The Seal article: Researchers develop AI who can think-and never stops learning to learn first appeared at the online magazine Basic Thinking. You can start the day well every morning via our newsletter update.
Researchers of the MIT have developed a self -learning AI that should be able to adapt to their surroundings. The so-called Seal model is able to learn continuously and even think yourself.
AI models can now write more than just texts. You can generate pictures and videos, program code and even compose music. But so far they have been denied a crucial thing: the ability to learn from experience. But that’s exactly what Researchers at the Massachusetts Institute of Technology (with) change.
With a new approach called Seal (Self-Adapting Language Models), you want to create a voice model that is supposed to learn constantly-without the disclosure of new training data. The researchers’ idea: a AI who takes notes, gains new knowledge and adapts their “memory” at any time.
It should be able to generate its own training data based on what she has just learned. This is comparable to me to a student or student who writes a summary in order to better understand the content.
AI model Seal should never stop learning
An example: If the AI model Seal receives information about the Apollo room trip program, it should generate new connections. You could then use this self -generated information to “train” yourself. Then the model should check whether the new knowledge can do better answers.
If that works, she keeps the change. The whole thing is reminiscent of the principle of “learning by doing” – in the digital version. The system is not yet perfect. The so -called “catastrophic forgetting” remains a major challenge. The background: If the AI learns something new, it can happen that it loses old knowledge.
This is one reason why machines and people still differ significantly in their way. Seal also needs a lot of computing power. How exactly the researchers want to plan learning at sensible intervals is also not entirely clear.
Self -learning AI with future potential
Nevertheless: the direction seems to be right. Because if language models can continue to learn independently in the future, this could improve many applications. From personal chatbot, which remembers user preferences, to AI assistants who are constantly adapting to new tasks.
The idea that machines become smarter over time than humans is moving closer to a little bit. Nevertheless, it remains open to what extent the technology can be fed with false information – or could feed yourself. Because the AI simply passes on what it learns from humans, knowledge could arise that does not correspond to the truth.
Also interesting:
- AI distortions: What does bias mean?
- Prompts: You shouldn’t reveal these things Chatgpt
- Chatgpt: Deactivate AI training with your content
- Edit pictures with chatt-this is how the AI function works
The Seal article: Researchers develop AI who can think – and never stops learning to learn first on basic thinking. Follow us too Google News and Flipboard Or subscribe to our update newsletter.
As a tech industry expert, I am both excited and cautious about the development of an AI that can think and never stops learning. On one hand, the potential for such technology is immense, as it could revolutionize various industries and solve complex problems that humans may struggle with. The ability for AI to continuously learn and adapt could lead to groundbreaking advancements in fields such as healthcare, finance, and transportation.
However, there are also significant ethical and societal implications to consider. The idea of an AI that never stops learning raises concerns about control and oversight, as well as the potential for unintended consequences. It is crucial that we establish clear guidelines and regulations to ensure that this technology is used responsibly and ethically.
Overall, the development of an AI that can think and continuously learn is a major milestone in the field of artificial intelligence. It will be important for researchers, policymakers, and industry leaders to work together to harness the potential of this technology while also mitigating any risks that may arise.
Credits