Grok is now more open than OpenAi.
Was this out of spite? We will never know the answer although the motivation doesn't matter. What we know is Elon Musk challenged OpenAI to the point where he is suing the company. We will see what the courts rule.
Regardless, we see the public pressure being applied.
It seems like one step in this process is to open Grok.
Is this the right move?
This cannot be answered since there are valid arguments on both sides. Some believe that AI should be all open source, removing the full control from single companies. Others argue that this is a mistake and a larger threat to humanity since one with nefarious intentions can cause massive damage.
Grok Opened Up For Anyone To Use
We got a few of the specifications of what was released. These are making the rounds of YouTube already.
What we have:
- 314 billion paramaters
- 8 experts
- 86 billion active parameters
- Apache 2.0 license
Here is the post announcing the release.
This is not the first open source LLM. Meta has Llama out there so this is another one for people to play with.
The key is it is general purpose without having any specific application. We will likely see this generated over the time.
Performance
This was the first version that was released back in the Fall. We were told that version 1.5 is on its way. This is not what was released so the performance will trail the other models. Of course, we will have to see what 1.5 is like, most likely being available on X via subscription.
In terms of the parameters, the will place it second behind what many project ChatGPT 4.0 at. The difference is stark meaning, when it comes to the general purpose models, OpenAI is still leading the race.
What will be interesting it see is how Grok is adapted once developers start utilizing it. There are some specialized models that were created from Llama2. It is likely we will see a similar path followed with Grok.
Why This Is Good News
To me, it is better if more is open source. We are dealing with some powerful technology and having it open means that we are not completely beholden to mega-technology companies.
Personally, I believe open source is going to win this one. The innovation that comes from the open source community is likely to outpace what the individual companies can created.
That said, open source is still trailing. One reason is they started a bit later to the game. Fortunately, they are catching up.
Another benefit is the specialization. As these LLMs are out there, specific use cases can be developed. This is likely to surpass the capabilities of the general models, at least in those fields.
My view is that the road to AGi is going to be paved by many ANI. When creating an AI for a specific purpose, the potential threat is reduced.
It seems like decentralization is a defense against vulnerabilities. Naturally, we still have the problem of the amount of compute required along with the data. At least with Grok, all that is done on X is going to be open since it will eventually be fed into Grok. Over time, with each new generation, the open model is going to be upgraded, remaining a version behind.
We are still in the early days with all this. It was about 500 days ago that ChatGPT was released. Grok is simply another step in this evolution.
Here is the present major players:
- ChatGPT
- Gemini
- Llama2
- Claude
- Grok
There are an assortment of other smaller models which are showing promise.
The next 24 months are going to be very interesting.
Posted Using InLeo Alpha
