Leo's amazing predictions

[AerisVahnEphelia]

29 minutes ago, Leo Gura said:

I would argue that calling it intelligence is disingenuous because it has memorized 2000 years of human knowledge. That allows it to fake intelligence very very well.

It is like you trained a computer to memorize the answer to every multiple choice test that has ever been written. And now it sure seems like it can pass every test. But that is disingenuous because it doesn't actually know how to pass a test.

There is a fundamental disconnect between intelligence and cramming LLMs with more data. These are not the same thing at all. No amount of data makes you intelligent. But if you memorize all of Wikipedia, you sure can pass as intelligent in most circles. But not in my circle.

I disagree. Everything is data, but LLMs are just one kind of it. Words must lead to practices and actions.

There is no fundamental difference between listening to your video and talking to ChatGPT. Both are interactions with pure data.

Doing psychedelics, though, is a form of practice that opens you up to different inputs and forms of data.

It does not matter whether LLMs are intelligent or not. The question assumes that they possess a self, but they do not.

The real question is whether they make me cognitively more “efficient,” and within the framework of human language, the answer is yes.

[Leo Gura]

8 minutes ago, Andrey said:

Here is the mindfuck, by trying to avoid losses you will actually lose money 

That possibility is certainly not lost on me.

You're damned if you do and damned if you don't.

But buying into a historic hype cycle is crazy-stupid if you worked hard for your money.

Investing requires sober, conservative thinking. Not hype, hope, and greed.

If I miss the AI boom, that's not going to hurt as much as crashing out. So I err on the side of caution. The upside now is too low and the downside too high.

Edited 6 hours ago by Leo Gura

[Andrey]

Microsoft's price-to-earnings ratio is at 23 (last time it was lower was 2018) and net earnings per share is at record high. A company with practically zero debt.
https://www.macrotrends.net/stocks/charts/MSFT/microsoft/pe-ratio

[AerisVahnEphelia]

12 minutes ago, Leo Gura said:

That possibility is certainly not lost on me.

You're damned if you do and damned if you don't.

But buying into a historic hype cycle is crazy-stupid if you worked hard for your money.

Investing requires sober, conservative thinking. Not hype, hope, and greed.

If I miss the AI boom, that's not going to hurt as much as crashing out. So I err on the side of caution. The upside now is too low and the downside too high.

right now the train has left the station that's for sure.

Maybe robotic.

[something_else]

1 hour ago, Leo Gura said:

I would argue that calling it intelligence is disingenuous because it has memorized 2000 years of human knowledge. That allows it to fake intelligence very very well.

It is like you trained a computer to memorize the answer to every multiple choice test that has ever been written. And now it sure seems like it can pass every test. But that is disingenuous because it doesn't actually know how to pass a test.

There is a fundamental disconnect between intelligence and cramming LLMs with more data. These are not the same thing at all. No amount of data makes you intelligent. But if you memorize all of Wikipedia, you sure can pass as intelligent in most circles. But not in my circle.

I disagree. They are able to strategise, plan, answer questions, figure out where to get the data they need from, and come up with new solutions to problems in completely novel scenarios, that is more than memorisation. If LLMs were relying purely on memorisation then they would not function very well as reasoning engines, yet they do.

The core concept they are relying upon is actually pattern recognition, not memorisation. Wikipedia is more akin to memorisation than an LLM is. 

Here is why their pattern recognition is so good:

The entire internet (which at this point is comparable to what Anthropic models are being trained on) is estimated to be around 200 zettabytes. So lets say all human knowledge from the past 2000 years comes out at 200 zettabytes as a very crude estimate.

Claude Opus is around 1TB in size. 200 billion times smaller. Yet somehow it is still able to make a pretty good attempt at giving you information from that entire 200ZB corpus. The only way this is possible is with intelligent abstraction of all of the core concepts and patterns contained within that 200ZB until it can model most of them pretty well with only 0.0000000005% of the amount of data.

It is this abstraction of concepts that I feel most people would call 'intelligence'. Most human measures consider the ability to identify abstract patterns to be a core component of intelligence as it is one of the fundamental building blocks of doing anything useful with intelligence.

Edited 5 hours ago by something_else

[Leo Gura]

I am not saying that it's literally memorization. But it is made on data. If data is missing it cannot perform that function.

An AI cannot draw a swordfish unless it has data for a swordfish. It cannot code a language it wasn't train on. It is not actually intelligent. But it has e ough data to make you think that it is.

Edited 5 hours ago by Leo Gura

[something_else]

14 minutes ago, Leo Gura said:

It cannot code a language it wasn't train on

They can. It will translate the core concepts from languages it was trained on into the syntax of another language. It's better when it has more examples of the unseen language's syntax, but the same can be said of human intelligence.

Quote

An AI cannot draw a swordfish unless it has data for a swordfish

If it had modelled the concepts of fish, sword, head and ocean then it could make a pretty decent effort at combining all of that together into a swordfish.

Image models are still quite a bit behind text models in capability but they can still combine two known concepts into a novel one to a degree, just not as well

Edited 5 hours ago by something_else

[Leo Gura]

8 minutes ago, something_else said:

Image models are still quite a bit behind text models

That is only because text data is so much more than images.

[something_else]

3 minutes ago, Leo Gura said:

That is only because text data is so much more than images.

I mean yes, I think where we disagree is why this matters. More data, more training and more experience makes you smarter, the same is true for humans.

[Leo Gura]

2 minutes ago, something_else said:

I mean yes, I think where we disagree is why this matters. More data, more training and more experience makes you smarter, the same is true for humans.

No! It is not the same at all.

Humans are intelligent without data.

[Leo Gura]

1 hour ago, Andrey said:

Microsoft's price-to-earnings ratio is at 23 (last time it was lower was 2018) and net earnings per share is at record high. A company with practically zero debt.
https://www.macrotrends.net/stocks/charts/MSFT/microsoft/pe-ratio

Yeah, but Microsoft is such a disgusting, evil company.

Even if Microsoft is under-priced, the rest of the stocks aren't. And when the bubble pops, your Microsoft will drop 50%.

What will be its P/E once the market falls out from AI? Once the chatbots prove unprofitable? Earning in tech will drop a lot. Entire customers will go bust.

Edited 4 hours ago by Leo Gura

[something_else]

2 hours ago, Leo Gura said:

No! It is not the same at all.

Humans are intelligent without data.

Are they? A newborn isn’t that intelligent and that’s basically a human without any data 

[Leo Gura]

Just now, something_else said:

A newborn isn’t that intelligent

You are insane.

[Stick]

😂😂😂

[Leo Gura]

This is long but good:

 

[Leo Gura]

Feel free to post counter-arguments. I don't want an echochamber here.

[AerisVahnEphelia]

I m bullish on AI, but I think it's so over leveraged for the current reality of it.

We already see how it's stagnating right now & the price of it.
 

I still wouldn't invest in AI right now, crash or correction is very very certain unless some magic happen.

Edited 2 hours ago by AerisVahnEphelia

[Basman]

Good to know that Leo got some butter in his house. Las Vegas right?

Leo, I'm proud to announce I finally started contemplating shit. I'm going to be rich soon.