I think itt's really disappointing that we don't have a good large language model that people can tinker with and that you can build upon, and businesses can change for their use cases. This is what's bothering with me. This kind of thing where we will have this centralized system, this kind of priestly class of intellectuals who will sit and theorize in San Francisco or Oxford and they will figure out the optimal paths for civilization and they will be the Philosopher Kings to make decisions for the future of humanity and impose it through their brilliance or the assumption of state power which they keep pushing for-- when I hear that spiel, I think immediately oh yeah that's the communist impulse. If there's a culture of torrenting and building things, sharing weights, I think that's really awesome. Every kind of communist professor in the past 100 years-- the professors are commnunist because if you're not communist you're off building things instead.

When it's low variance, you know exactly how to put pressure on things and control it. Beff: I would say cultural selective pressure. It's kind of like companies got woke over time because they could afford to, they were just printing so much money that they can let their culture fester because there's no selective pressure. Summer 2020 we all saw that with our colleagues and coworkers and people who were previously reasonable just going nuts. A lot of people became enlightment pilled with Sydney. Amjad: There were glimmers of that with the Sydney mania. Maybe the mainline thing to do is to delete your Microsoft operating system if Sydney seems evil maybe that's how people will protest. I think that's missing now. The top-down control for each hyper parameter setting of the architecture will have a control against prompt attacking, like a counter-prompt, but you know, kind of like a vaccine or something- but the underlying thing just mutates its hyper parameters and now your control strategy doesn't apply.

If you give certain types of doomercel people the control of information pipes like search or social media or chatgpt plus deutsch or other proto-googles then you give them the keys to control the flow of information and the psyche of all the people. With an evolutionary algorithm, like in the space of companies, cultures or products, there are certain bounds that show that if you want to go fast you have to embrace variance. In reality, with the no free lunch theorem, you can't have one model that is best at everything. As long as there is a healthy ever-evolving landscape- and because it has exponential compounding because of virality, then the rate of improvement will be faster and it can converge on the best model much faster than big centralized orgs. China has a monopoly on precious metals right now. There has been a great stagnation there since 1970s is heartbreaking how little progress we have made since then.

That's the thing that the decels and authoritarians-- they want to psyop you into believing you have far less agency than you actually have. I think it's so amazing that the way AI played out was not the symbolic AI- it was the connectionist AI and that's inehrently networked and parallelized AI. Super paraphrasing, his recommendation for maintaining civilization stability is to impose a global panopticon with predictive policing and a world government. We live in a society that is super saturated with that. This was a big deal because Lennin in 1917 applied communism to the pre-industrialied agrarian society of Russia at the time, and therefore it would never industrialize and it would lead to chaos and death. This is a good question. In the broader public, it's a really good question. I think you could read it as a rightwing endorsement really. All this rationalization about how nobody really had anything to do with him, which wasn't true up until that exact moment.

If you adored this article and you also would like to collect more info about chatgpt Auf deutsch umstellen i implore you to visit our own web site.