Jeroen Baten
Sometimes I have an idea. I may not be the first, nor do I claim to be. But when I think it is useful I put it down in writing in a blog post. And so is it that this short blog post was written:
https://www.jeroenbaten.nl/thinking-about-ai/
Basically it says that for all the people using AI to do their work, it should be possible to migrate your history easily to another AI vendor. Technically it is very easy. But I'm afraid we are going to need a law to make it happen.
Thinking about AI - Jeroen Baten

Thinking about AI and both the success of it and the competition in that market I got an idea.So I wrote done the bullet points of my thoughts and ask Claude to write a coherent text about it. AI Data Portability: The Case for Conversation Portability Artificial intelligence is arguably the most transformative technology of […]

Jeroen Baten
Mar 19 at 8:58amWeb
Show threadHans van Zijst5d ago
@Jeroen Baten You're not the first to think about this, indeed, and it is indeed a very important thing to get right from the start.

During the Industrial Revolution the means of production got highly concentrated, which resulted in much more wealth and power for those who owned the machines, and a life in the margins for those who did the actual labour.

This AI revolution could very well start a similar concentration of production means, only this time those production means are knowledge.

We've already seen how powerful an enterprise like Google can become by pointing people at knowledge that's "out there", imagine the impact if a handful of enterprises make the "out there" their own?

Those few enterprises would become the worldwide oligarchs of knowledge and information, they would have control over who gets which information.

Thinking only about elections, this would have disastrous effects. We've already seen how social media and their algorithms have caused polarisation, this would have a much bigger effect.

And can you trust software any longer if you know that (most of) it was written by an AI that knows who you are and what you do? Can I trust my banking software when I know it was largely written by an AI in an America that is very hostile to Europe?

Who guarantees that these AI providers aren't controlled by the government of the country they operate in, who can assure that we're getting clean, honest information and not some Trump-approved cock-and-bull story?
Show threadJeroen Baten5d ago
@hans I agree to some extent with this. But I think not everyone or everything is evil. Yes, we need to be alert and I think we are in an industrial revolution kind of transformation period. Where the first industrial revolution replaced hand labor with machines, this revolution will replace cognitive work by AI. Seeing how the first revolution changed society, this does not bode well....
Show threadHans van Zijst4d ago
@Jeroen Baten No, not everybody is evil. In fact, I think the vast majority of people aren't.

But wealth and power have a very predictable effect on people. What do people with money want? More money! The richer or more powerful someone becomes, the more they start living in their own bubble with its own reality. That's a well-documented fact.

And if knowledge and information get concentrated in the hands of a few, there is a 100% chance that they will abuse their position. Only very slightly to begin with, probably even for the greater good. But it's a slippery slope, that will eventually end in widespread corruption.

The only way to prevent that, is to prevent this concentration. A capitalist world will always favour concentration, so we would need laws to prevent, or at least limit that.

Allowing people to leave one provider for the next and taking all their prompts plus answers with them, would mean a reasonably level playing field.

But we need laws for that, somewhere along the lines of the DMA and DSA.
Show threadRonny Lam5d ago
@JeroenBaten this is difficult, but not impossible. Copying (part of) a neural network depends on how similar the neural networks are, otherwise you can not do it on this level. Then you would fall back to relearning your old conversations to the new AI. But with a different neural net, the outcome might be different. So in the end you are asking for a difficult thing in my opinion. But what you are saying is right. The lock-in is real!
Show threadJeroen Baten4d ago
@ronnylam of course products differ and are not compatible in their output. But they all use prompts and send back "stuff" (tm). But what I want to achive is that the new system knows about the history of a "chat" or "project". It is very easy for AI companies to build, but they will no doubt also be very reluctant.Even a migration toll can be build that scans former dialogs and feeds them into another AI.
Show threadJeroen Baten4d ago
@ronnylam Hm, if only there was a way to fund development of such an open source tool... :-)
Show threadHans van Zijst4d ago
@Jeroen Baten Call me a dirty commie, but government should have a firm hand in this.

@Ronny Lam
Show threadRonny Lam4d ago
@hans @JeroenBaten dirty commie!