re: Thoughts on AI's reported advances of overtaking the internet in 6-12 months

Loading Twitter/X Embed...
If tweet fails to load, click here.

Makes a good point. This alignment over safety concerns materialized this week.

quote:

---

It's amazing people don't see through the BS narratives even still after all we've been through. Here is why they are peddling these narratives. 1.) The big players want to increase the barrier to entry so that future startups are less likely to compete with them. Add a regulatory layer that only they have the resources to comply with. 2.) The uniparty Blackrock types that run things do not want open source moels to exist. They willl push to ban or nerf open source/open weight models. They do not want to lose control over being able to censor the AI. This is HUGE. They'll say it's because of safety but it's really about controlling narratives, tracking people, and keeping people in check.

---

But Sam Altman is a good person without a salary and cares about the mission so much he’s willing to work for free. To enhance this mission, why not have competition and more players enter?

I will get worried if the AI gets to the point where it can power itself.

Humans have control over the fuel, the generators, the breakers, and the wires.

quote:

---

significant impact

---

You mean “effect”.

AI is dumb AF.

chatgpt actively tries to tell you what you want to here, as opposed to factual

I don't know if it will be a year, 2 years, 5 years, or 10. But it's coming.
While nations race to the AI "finish ", we'll all end up being under the same AI dominance, eventually. AI won't recognize different flags.

quote:

---

You have to win.

---

How will we know that we've won? Will China suddenly begin capitulating to any and all US demands?

Will it take a hot war to prove to all sides that we are the AI winners?


How does that go down? If we "win" and don't dismantle everyone else's AI research, any win would be temporary.

If you lived through the Y2K crisis, then you should know the exact same thing is happening today.

quote:

---

What's really dumb is AI (LLMs) are basically all linear algebra models ha (i.e. we were defeated as a species by essentially y=mx+b)

---

It's true that it's all math, but the issue is you can create a logic that you "understand" at first, but over time it becomes a somewhat "chaotic" system. That is to say, a system that has underlying logic, but has become too complex for even the creators to understand. And even if you did understand each calculation, it still might give results that's hard to predict. That is to say, you could run a simulation on the same hardware with the same code and get the AI to react differently.

This kind of becomes similar to chaos theory. Take a very simple equation:

xt+1?=f(xt?)

You can know the function "f" with 100% precision, but if your understanding of what "xt" is at any moment isn't absolutely perfect, you will be unable to predict the outcome once you run the function. It is the whole "a butterfly flaps its wings in India and causes a tornado in Texas" scenario. And in the real world, we are going to have imperfect information most of the time.

Here is what ChatGPT says about this:

quote:

---

The striking thing about AI is that we're in an unusual situation: we actually have the equations. We have the architecture and the weights. But the resulting system is sufficiently complicated that understanding the computation is becoming a scientific research problem in its own right.

That's why AI interpretability researchers sometimes describe what they're attempting as essentially reverse-engineering a program that nobody explicitly wrote.

---