Archived: personal AI in the rugpull economy

This is a simplified archive of the page at https://blog.zgp.org/personal-ai-in-the-rugpull-economy/

Use this page embed on your own site:

26 Oct 2024

ReadArchived

26 Oct 2024

Doc Searls writes, in Personal Agentic AI,

Wouldn’t it be good for corporate AI agents to have customer hands to shake that are also equipped with agentic AI? Wouldn’t those customers be better than ones whose agency is merely human, and limited to only what corporate AI agents allow?

The obvious answer for business decision-makers today is: lol, no, a locked-in customer is worth more. If, as a person who likes to watch TV, you had an AI agent, then the agent could keep track of sports seasons and the availability of movies and TV shows, and turn your streaming subscriptions on and off. In the streaming business, like many others, the management consensus is to make things as hard and manual as possible on the customer side, and save the automation for the company side. Just keeping up with watching a National Football League team is hard…even for someone who is ON the team. Automation asymmetry, where the seller gets to reduce service costs while the customer has to do more and more manual work, is seen as a big win by the decision-makers on the high-automation side.

Big company decision-makers don’t want to let smaller companies have their own agentic tools, either. Getting a DMCA Exemption to let McDonald’s franchisees fix their ice cream machines was a big deal that required a lengthy process with the US Copyright Office. Many other small businesses are locked in to the manual, low-information side of a business relationship with a larger one. (Web advertising is another example. Google shoots at everyone’s feet, and agencies, smaller firms, and browser extension developers dance.)Google employees and shareholders would be better off if it were split into two companies that could focus on useful projects for independent customers who had real choices.

The first wave of user reactions to AI is happening, and it’s adversarial. Artists on sites like DeviantArt went first, and now Reddit users are deliberately posting fake answers to feed Google’s AI. On the shopping side, avoiding the output of AI and made-for-AI deceptive crap is becoming a must-have mainstream skill, as covered in How to find helpful content in a sea of made-for-Google BS and How Apple and Microsoft’s trusted brands are being used to scam you. As Baldur Bjarnason writes,

The public has for a while now switched to using AI as a negative—using the term artificial much as you do with artificial flavouring or that smile’s artificial. It’s insincere creativity or deceptive intelligence.

Other news is even worse. In today’s global conflict between evil oligarchs and everyone else, AI is firmly aligned with the evil oligarch side.

But today’s Big AI situation won’t last. Small-scale and underground AI has sustainable advantages over the huge but money-losing contenders. And it sounds like Doc is already thinking post-bubble.

Adversarial now, but what about later?

So how do we get from the AI adversarial situation we have now to the win-win that Doc is looking for? Part of the answer will be resolving the legal issues. Today’s Napster-like free-for-all environment won’t persist, so eventually we will have an AI scene in which companies that want to use your work for training have to get permission and disclose provenance.

The other part of the path from today’s situation—where big companies have AI that enables scam culture and chickenization while individuals and small companies are stuck rowing through funnels and pipelines—is personal, aligned AI that balances automation asymmetries. Whether it’s solving CAPTCHAs, getting data in hard-to-parse formats, or other awkward mazes, automation asymmetries mean that as a customer, you technically have more optionality than you practically have time to use. But AI has a lot more time. If a company gives you user experience grief, with the right tools you can get back to where you would have been if they had applied less obfuscation in the first place. (icymi: Video scraping: extracting JSON data from a 35 second screen capture for less than 1/10th of a cent Not a deliberate obfuscation example, but an approach that can be applied.)

So we’re going to see something like this AI cartoon by Tom Fishburne (thanks to Doc for the link) for privacy labour. Companies are already getting expensive software-as-a-service to make privacy tasks harder for the customers, which means that customers are going to get AI services to make it easier. Eventually some companies will notice the extra layers, pay attention to the research, and get rid of the excess grief on their end so you can stop running de-obfuscation on your end. That will make it work better for everyone. (GPC all the things! Data Rights Protocol)

The biggest win from personal AI will, strangely enough, be in de-personalizing your personal information environment. By doing the privacy labour for you, the agentic AI will limit your addressability and reduce personalization risks. The risks to me from buying the less suitable of two legit brands are much lower than the risk of getting stuck with some awful crap that was personalized to me and not picked up on by norms enforcers like Consumer Reports. Getting more of my privacy labour done for me will not just help me personally do better #mindfulConsumption, but also increase the rewards for win-win moves by sellers. Personalization might be nifty, but filtering out crap and rip-offs is a bigger immediate win: Sunday Internet optimism Doc writes, When you limit what customers can bring to markets, you limit what can happen in those markets. As far as I can tell, the real promise for agentic AI isn’t just in enabling existing processes or making them more efficient. It’s in establishing a credible deterrent to enshittification—if you’re trying to rip me off, don’t talk to me, talk to my bot army.

For just a minute, put yourself in the shoes of a product manager with a proposal for some legit project that they’re trying to get approved. If that proposal is up against a quick win for the company, like one based on creepy surveillance, it’s going to lose. But if the customers have the automation power to lower the ROI from creepy growth hacking, the legit project has a chance. And that pushes up the long-term value of the entire company. An individual locked-in customer is more valuable to the brand than an individual independent customer, but a brand with independent customers is more valuable than a brand with an equal number of locked-in customers.

Anyway, hope to see you at VRM Day.

#