It is rare that I agree with something that Mark Zuckerberg says. He is a guy that is easy to not only distrust but also dislike.
That said, he is one of the leading technology minds of the last 30 years, something that has resulted in one of the largest technology companies on the planet gives him a lot of street credibility.
Zuckerberg is also a guy who is always looking towards something larger. His deep dive into the metaverse, in spite of costing Meta billions, exemplifies how he is always trying to position the company at the center where technology is going.
When it comes to LLMs, a case could be made that Meta has the leading open model available. Llama3 is one of the players in this field, one not to be overlooked.
Image generated by Ideogram
AI Should Be Open source
Since Zuckerberg took the open approach, he is taking some shots at those who are trying to keep things closed. This could be a slap at Sam Altman, someone who is doing his best to control AI and be the only game.
As an aside, Zuck is not the only technology mind who has questioned Altman's motives. A lot of news is being made around his lack of concern for AI safety.
In a conversation, we gained some insight into what he is thinking.
To start, there will not be "just one AI". This is a premise that I agree with. My view is far too many fall into the mindset of the Terminator scenario and Skynet.
This does not mean, however, that a guy like Altman doesn't have his eyes on being the main (if not sole) player. If LLMs follow what has happened elsewhere, which I do not believe it will, this could end up as a winner take most.
For me, the challenge with LLMs is that they will become commodities. Eventually, most will end up with the same data. Toss enough processing at it, which these major players can, and we get a similar basis.
The difference comes in when we look at more advanced developed along with the applications built on them. Here is where OpenAi could be stepping ahead as it does seem like they are onto some cutting edge stuff.
Getting back to Zuck, he did add this:
Touting the value of open source to put AI tools into many people’s hands, Zuckerberg took a moment to disparage the efforts of unnamed competitors who he sees as less than open, adding that they seem to think they’re “creating God.”
This is something that everyone should concern themselves with. When dealing with something so powerful, we do not want it in the hands of one person or entity.
For Altman, this is paradise. He is even going so far as to petition governments banning open source AI, stating that many players cannot be trusted. Of course, his presumption is that he can be.
On this I disagree.
Decentralization As Much As Possible
I am in the camp of spreading things out as much as possible. Getting data open and into the hands of as many start ups as we can is vital. Those entities can work on their initiatives, moving things in many different directions.
SLMs also offer a great deal of potential. Companies building their own models, utilizing their proprietary data should be encourages. Taking an open source model, adapting it to their needs, and then adding to it will help to offset some of the threat.
Another facet of this discussion has to be the security and safety segments. AI can also be developed to prevent "runaway AI". If everything is closed how is this going to happen? For one, there is no way to know what the entities with the closed models are doing. Hence, generating something to counter potential threats is, at best, very difficult.
To me, AI should be worked on by as many as we can. This has to be a mixture of large and small companies. Big tech is obviously going to be at the center of this. However, they cannot be the only players. Looking at the present state of the Internet shows us how this unfolds.
The problem is we are dealing with technology that keeps getting more powerful. It is dangerous to put it in the hands of just one or two entities.
I think this is the point Zuckerberg was making.
And it is one that I agree with.
Posted Using InLeo Alpha