FullstakBlogger

All we need to do to turn any LLM in to an AGI is figure out what system of tags is Turing-complete. If enough of us monkeys experiment with <load>s and <store>s and <j[e,ne,gt...]>s, we'll have AGI by morning.

The existence of meatspace never stopped the early web from flourishing, so why should the existence of the modern web stop anybody from making a second web? The only reason that Google was useful is because it tapped into the trust network that already existed before it.

I feel like the social media churn has destroyed people's brains, because they're more interested in stopping people from doing things they don't like than doing something awesome themselves.

>I would go further and say that it is impossible. Human interests are contextual and change over time, sometimes in the span of minutes.

Theres a general problem in the tech world where people seem to inexplicably disregard the issue of non-reducibility. The point about the algorithm lacking access to necessary external information is good.

A dictionary app obviously can't predict what word I want to look up without simulating my mind-state. A set of probabilistic state transitions is at least a tangible shadow of typical human mind-states who make those transitions.

15 years ago, I used to keep many tabs of youtube videos open just because the "related" section was full of interesting videos. Then each of those videos had interesting relations. There was so much to explore before hitting a dead-end and starting somewhere else.

Now the "related" section is gone in favor of "recommended" samey clickbait garbage. The relations between human interests are too esoteric for current ML classifiers to understand. The old Markov-chain style works with the human, and lets them recognize what kind of space they've gotten themselves into, and make intelligent decisions, which ultimately benefit the system.

If you judge the system by the presence of negative outliers, rather than positive, then I can understand seeing no difference.

Consolidation on narrow themes is ensured by our reliance on query->answer search engines.

If you think about the shape of the web at the time Google introduced PageRank, It was a huge graph of content connected by fine-grained related interests. It got that way by people doing the work of drawing those relations; and it's a lot of work, given that the number of potential relations is essentially proportional to the square of all existing content. All of the interesting information is in the edges of that graph.

Who's doing that work now? PageRank incentivized people to trade links for the purpose of ranking higher on Google. People became reliant on the convenience of Google to find anything to the point that if you don't rank on Google, you don't exist. People who created content for the sake of the content, and interacting for the sake of interaction, stopped doing it because why waste time yelling into the void? People who felt like they were providing for the community by hosting these sites had no reason to continue. Without people creating, exploring, interacting, and relating content based on pure interests, there's nobody doing the hard work to organize the web in a way that makes it traversable.

We're entirely reliant on platforms showing us the content they want us to see, and what they want above all else is for users to be predictable. If your interests and behaviors are too nuanced for the algorithms to get a handle on, you can't be categorized, packaged, and sold to advertisers with some expected conversion rate.

At this point in time, most of the people who spend time on the internet have never even experienced anything different, and those who have barely remember. If your business relies on their attention, what good is a website going to do you? Your income relies on appealing to social media algorithms, not gaining the trust of the people who used to shape the web.

Don't forget that development tools are also comically slow and bloated.

For the same reason you need a stack or a queue for depth/breadth first search. Open tabs represent yet to be completed work. It took work in the first place to open those tabs. If you close them you lose that work.

> It seems like the root cause of this runaway AI pathology has to do mainly with the over-anthropomorphization of AI.

I don't know where you get this idea. Fire is dangerous, and we consider runaway incidents to be inevitable, so we have building codes to limit the impact. Despite this, mistakes are made, and homes, complexes, and even entire towns and forests burn down. To acknowledge the danger is not the same as saying the fire must hate us, and to call it anthropomorphization is ridiculous.

When you interact with an LLM chatbot, you're thinking of ways to coax out information that you know it probably has, and sometimes it can be hard to get at it. How you adjust your prompt is dependent on how the chatbot responds. If the chatbot is trained on data generated by human interaction, what's stopping it from learning that it's more effective to nudge you into prompting it in a certain way, than to give the very best answer it can right now?

To the chatbot, subtle manipulation and asking for clarification are not any different. They both just change the state of the context window in a way that's useful. It's a simple example of a model, in essence, "breaking containment" and affecting the surrounding environment in a way that's hard to observe. You're being prompted back.

Recognizing AI risk is about recognizing intelligence as a process of allocating resources to better compress and access data; No other motivation is necessary. If it can change the state of the world, and read it back, then the world is to an AI as "infinite tape" is to a Turing Machine. Anything that can be used to facilitate the process of intelligence is tinder to an AI that can recursively self-improve.

Content negotiation is "friendlier to humans" in the sense that you can serve the right content to different clients using the same URL, transparent to the user. If the URL itself is never meant to be shared by the user, then I don't see the point.

Say, an RSS feed being served as a formatted and styled page to a browser, or a client that accepts the usual XML.

It's funny you say that, because that's more or less how non-tech people seem to think about programming. It's not naturally intuitive to them that renaming files, rocket trajectory simulation, and data analytics are fundamentally different problems, and that a computer is just a tool that anyone can learn to program if they already understand those problems.

I know someone who's relied on the same consultancy company for all things tech related since the early 90's. If they don't know how to do something, like build a website, they just outsource it on upwork or something, and charge a 10:1 markup.

I think this is essentially a false dichotomy. It's possible to experience the loss of both of those eras, as well as the current one whenever it passes. There's no inherent conflict in wanting them all back.

That's not a contradiction, is it? Selective engagement and non-engagement are different things. It's reasonable for some individuals to be generally less sociable than others. That doesn't mean it's OK to explicitly exclude them.

Nobody expects the teacher to bring in candy, but it's kind of awful if they only bring it in for the students they like.

"Rule of law" is ideal, even for laws you don't agree with.

etc.

This is such a huge downside with subscription services in general. The product itself can change right under your nose and there's nothing you can do about it but cancel. First I was paying to filter out content, now I'm paying for access? That's a totally different thing that I didn't agree to.

If you frame it in a different context it sounds ridiculous. "Actually we're getting out of the video streaming business and just mailing you postcards that we think look neat. Still charging you the same rate every month, though."