If you are old enough to remember “vendor lock in” you will certainly be sensitive to the new challenge – putting your sensitive corporate data into someone else’s LLM – and a stranger forever owns what they intuit from your data.
Any clever query builder can likely access the sensitive information you put up into that LLM while you were thinking you were building a competitive advantage.
Welcome to the data sovereignty challenge – the immovable object up against the irresistible force of your board of directors making you do artificial intelligence – now, regardless of the cost.
Companies and governments have been sold the nonsensical, false narrative that the way to do A.I. is with unbounded large language models – using oceans of data – to which you gleefully contribute yours – to train an A.I system.
That is no longer the case.
The first issue with the ungainly LLMs is they are generally unbounded.
They try to answer almost any question for any person under any circumstance.
If you believe there is enough data in the universe which, when put into an LLM, answers the question “why is man” you need to stay away from A.I. budgets.
More leading firms are now adopting the concept of constrained, models – which solve, or address, a single problem with massive scale.
Once you determine or agree that your business or government program needs to answer problems in a domain, bounded by certain information types, you open the door to small language models.
The problem, the LLM guys will tell you, is that you need the massive data of a META or Twitter/X to get the best training for those models.
Aha! Sounds good but isn’t true – and we have the proof.
The issue of data sovereignty is exactly the one you are about to face with these LLMs – putting your data in someone else’s cloud – which can be in a country you abhor, run by your enemies – or competitors – wait until your board finds that out!
The alternative is taking lots of SMALL language models that run in Fractal instances that you control and aggregating them – with your data never leaving your control – so you get the benefits of scale, the benefits of far lower costs, while retaining data sovereignty.
This is where Fractal, designed from the bottom up to be fully distributed – with no cloud, no central point of control, enables scale far beyond any LLM, while you retain data sovereignty.
Let’s do an exercise.
You set up a Fractal system with 1,000 nodes, or Fractals.
Maybe 100,000 or millions, but unless you are Apple Computer doing a proof of concept, let’s go with the 1,000.
You allocate your data to 200 Fractals, each under your sole control.
Some are in the cloud, some are in your data center, some are at a customer site or all are behind your firewall – your configuration choice.
You fire up the training models and run them – in this case, you can even choose to do so without a data center at all – something Fractal is pretty much known for.
Regardless, you now have security and control on every piece of your data.
Nobody but you controls your data – no one else can see it or touch it.
When your board member calls to see how it’s going, you report that you built a huge training model, with over a thousand nodes – but you protected your data sovereignty at every step.
You received all the benefits of a state-of-the art training model, yet, unlike the LLM guys, nobody but your firm or agency saw, touched or received your sensitive data.
40 years ago every CIO, back then called VP of IT, sweated vendor lock in.
They feared a DEC, IBM, Burroughs – would tie their systems to vendor hardware and they would be owned by that vendor.
Today, with A.I. the CIO and board are learning when they participate in someone else’s LLM they have something worse than vendor lock in.
They lost the ownership of their data – forever.
They may have given up their most valuable intellectual property - to an LLM run in a distant place, out of their control.
So adopt data sovereignty; data is your most important asset.
Fractal distributed models – brought together – are more powerful than any LLM, and you get a good night’s sleep.
Without an data center - with sustainable computing - available today.
FractalComputing Substack is a newsletter about the journey of taking a massively disruptive technology to market. We envision a book about our journey so each post is a way to capture some fun events.
Subscribe at FractalComputing.Substack.com
Fractal Website: Fractal-Computing.com
Fractal Utility Site: TheFractalUtility.com
Fractal Government Site: TheFractalGovernment.com
Fractal Sustainable Computing: TheSustainableComputinginitiative.com
@FractalCompute
Portions of our revenue are given to animal rescue charities.
Here is a worry for every sovereign individual in our nation. Data sovereignty is the central issue when it comes to your personal information. Palantir is sucking up every citizens information and feeding their LLM with it. That’s information on your health, social security information, IRS information, insurance information, and likely whatever the NSA stores on you. In other words, they are centralizing all government data about all citizens. What could go wrong? This is a grave danger to our liberty which every citizen needs to understand. Fractal is the answer to gain the benefits of AI without the risks of loss of data sovereignty.
Data sovereignty in the 21st-century is equivalent to individual sovereignty and parallel to national sovereignty. The world is at war whether people acknowledge it or not: Globalism vs. Nationalism. This war is an information war and data is its weaponized currency. The globalist elite’s campaign for world domination is revolution without bullets. "Companies and governments have been sold the nonsensical, false narrative that the way to do A.I. is with unbounded large language models - using oceans of data - to which you gleefully contribute yours - to train an A.I. system." But it is YOUR individual sovereignty that is the spoils of this war! The enemy does not need bullets to control you––it only needs your data!