Microsoft's Gorilla AI model is the closest to reaching AGI
Gorilla AI model is another AI breakthrough from Microsoft.
3 min. read
Published on
Share this article
Read our disclosure page to find out how can you help Windows Report sustain the editorial team Read more
Key notes
- Gorilla was able to conduct tasks on its own.
- The model can seemingly think on its own, without the need for human input.
- You can use it to build and train your own AI model.
The Gorilla AI model is one of the latest AI models funded by Microsoft, and this one looks impressive. Microsoft already funded several other AI models in the past months.
For example, there is Orca 13B an open-source AI language that can help you build your own AI model. The Redmond-based tech giant has invested in AI breakthroughs as well with Kosmos-2 being able to visualize space and form input with its own original perspectives.
Llama 2, arguably the world’s biggest open-source LLM currently, is the AI partnership of Microsoft and Meta. This language will also help individuals and companies build their own AI solutions. And there are many others: phi-1, CoDi, DeepRapper (yes, you read that right), Project Rumi, and all of them come with important features and improvements.
However, the end goal of AI is reaching AGI, which is Artificial General Intelligence. Practically, when reaching AGI, an AI model will have the same thought processes as a human being. And from there, it would be just a matter of time until it will be able to solve humanity’s biggest problems. Or, at least, this is what the majority of AI developers want.
It seems that Microsoft is getting closer to reaching AGI with Gorrila AI. The model, which was funded by Microsoft Research, is the AI research result of Microsoft and UC Berkeley.
How is the Gorilla AI model reaching AGI?
Gorilla AI is a large LLM model that is capable of understanding the tasks at hand and delivering the correct solutions to them. The researchers from UC Berkeley and Microsoft Research conducted a test where Gorilla was put to write an API call.
Compared to GPT-4 and Claude AI, Gorilla managed to write the API calls correctly without hallucinating. Hallucinations on AI models mean that they provide an answer to your task, but that answer is mainly fabricated and incorrect.
However, Gorilla managed not only to write the calls but the model also heavily surpassed GPT-4 on every level. According to the paper, Gorrila is also capable of:
- A strong capability to adapt to test-time document changes.
- Mitigating the issue of hallucination.
So, in other words, Gorilla was able to enable and update the API calls on its own, without any human input. And when encountering the issue of hallucination, the AI model would mitigate it substantially, without actively using it.
Gorilla demonstrates a strong capability to adapt to test-time document changes, enabling flexible API updates and version changes. Gorilla also substantially mitigates the issue of hallucination, commonly encountered when prompting LLMs directly
This means Gorilla relies on its own resources and inputs to resolve tasks and conduct updates on its own. These capabilities make it behave in the same way an independent human being would behave, which brings the AI model closer to reaching AGI than any other so far.
As with many other AI models before, you can use Gorilla to build your own AI model, and the developers have made the code available to download on GitHub, for free. Try it out.
What do you think about it? Let us know in the comments section below.
