We know there is some powerful technology being developed. However, it appears that things are also going to get very expensive.
Sam Altman is one who is fighting to get regulations passed that will ban open source LLMs. If we did not believe he was after regulatory capture before, this seals the deal.
According to a new report from The Information, OpenAi executives are weighing charging users as much as $2,000 (over an undetermined amount of time) for access to their most advanced AI models. For comparison, ChatGPT premium currently costs $20 per month, a fee that enables the use of GPT-4o, the company's current flagship model.
This is the type of stuff we have discussed over the last couple years. There is a reason why the democratization of data is so important. Here we are seeing it right before our eyes.
If you do not think a world exclusively run by Big Tech is unenviable, just consider what is being proposed here.
Taking Exclusion Even Further
We often discuss having an inclusive system. This is something that makes sense on many different levels. However, as we can see, we are on a path to anything but that if Altman has his way.
Very few individuals or entities can afford $2,000 per month. While this might be the enterprise level, i,e, catering to companies, it is insight into the thinking. Big Tech is investing hundreds of billions into equipment to train these models, money they are looking at recoupling.
Remember, Altman has stated that he believes that OpenAI can be a $100 trillion dollar company. While these technology can forge massive increases in economic productivity, we can see how he seeks to capture much of it for OpenAI.
It is pretty fairly clear why he is emerging as public enemy #1 of the tech world. He is starting to distance himself from Mark Zuckerberg.
Speaking of Zuckerberg, he is on the opposite end of this fight.
His approach is open source. Llama is an open model, at least releasing the weights. Of course, Zuckberg isn't doing this out of kindness. His goal is to place Meta at the epicenter of the AI world, having an ecosystem emerge around his platform.
Think of it as the Android of the LLM world.
While this might not be perfect, it is a better approach. Meta is dumping billions into compute to train its models. This is a cost it is carrying for everyone else.
It also allows for the fragmentation of the technology.
Small Language Models
At this point, I am not sure this is the proper definition but it does capture what we are discussing.
For democratization to occurs, we need to start the process of pushing this out as much as possible. This means platforms embracing this technology and incorporating it wherever they can.
Obviously, processing can still be an issue although that are project such as Qtum which have a 10K GPU cluster. It does not compete with the next generation, which is going to run into the 100K range, but it is a start.
The Qtum development team dropped a bombshell last month, onlining 10,000 GPUs to power an AI image generator called “Qtum Qurator” and an AI chatbot called “Qtum Solstice.” Both of these products are based on open-source models and are the debut product offerings found on the Qtum AI website.
Blockchain is ideal for expanding the open data available. There are two components that are part of the feedback system.
When looking at how this works, we start with data fed into the model. This is where companies are scraping and doing whatever they can to acquire this. Social media has shifted from being the end destination to being a feeder system for the AI.
Once the data is input and the model trained, people can use it. Through prompting, the model delivers information. Remember, this is not simply a mirror output like traditional computing. The model is actually generating something completely new.
Under closed systems, this information is housed on the servers, being used to feed back into the model.
With blockchain, we can take these results and post them. That puts them into the open space, allowing anyone who is developing a model to utilize the data.
Will this compete with OpenAI? Not from the start but over time, as the open source community expands, we can see how the numbers start to shift things.
Democratization Of AI
The ultimate goal has to be the democratization of AI.
Here is where blockchain is valuable. Since many are decentralized, open source networks, anytime place on them is public. This means that AI models and, eventually, agents, that are developed can also be part of the open source realm. While not everything will fit into this category, a good bit of will.
When this occurs, a couple things happen.
To start, things can develop much quicker. As others start to join in, that magnifies what entities such as SingularityNet are doing. Again, it becomes a numbers game.
Another is that people have access to the technologies are near or zero cost. This is where democratization enters. When anyone in the world can use it, we see even more input.
And when we are dealing with a data feedback loop, this will only enhance the development.
Of course, this is what scares Altman. If this does happen, his ability to monopolize what is taking place diminishes. Hence, when he talks about the dangers, what he is really saying is "I am after regulatory capture".
His goal is to have the government help him in making trillions.
Posted Using InLeo Alpha