AI Hype vs. Reality: Lessons from Zuckerberg’s $50B Metaverse Bust

Remember the Metaverse? It was super-over-hyped. Hysteria, really. Mark Zuckerberg put more than 50 billion USD into “the metaverse”.

Let me repeat that: 50+ billion USD.

And that was just five years ago… Zuckerberg poured all of that money into nothing. Where is Meta’s Metaverse today?

Nowhere.

But now, it’s AI that is supposed to “change everything”?

Suddenly, the former metaverse expert Mark Zuckerberg has become an AI expert? The same Mark Zuckerberg who just recently wasted 50+ billion USD on “metaverse”?

And, do you remember “agents”? According to Sam Altman, people were going to pay thousands of dollars per month for an “agent”.

And that was just last year.

As an example: Salesforce announced its transformation into an “AI Agentic” company last year.

Now, we hear news about Salesforce’s latest market projections; the company is not expecting *any* growth from AI this year…

What happened with all those amazing agents that were going to increase sales and profits?

Yeah, nothing. Just like “Metaverse”.

Google’s Sundar Pichai and OpenAI’s Sam Altman were both straight up lying about what agents will be able to do, when they talked about it, only last year. Today, they are rather quiet on “agents”. They had no clue about the future last year. But they pretended like they did. Maybe they even believed themselves?

Tech bros have no special insights

So, we can conclude that VCs have no special insights about the future, or what to invest in. Including Zuckerberg and Musk. They just throw around money and see what sticks.

Because even if they already call it “AI”, there is no Artificial Intelligence. Yet. There are only pieces of software and web servers hosting static neural networks, like LLMs.

There are snake oil salesmen everywhere, however.

How about thinking for yourself for a moment? Yes, “metaverse”, “blockchain” and “AI” can be useful for certain specific things, but these buzzwords are not “changing everything”.

Stablecoins are nice. Virtual environments are nice. An LLM googling and summarizing for you is nice. But it does not “change everything”. Computers (AI) helping out with programming and math problems are making coding work more efficient, they are not “changing everything”.

Computer graphics actually looked better in Hollywood 20 years ago than it does in today’s so-called “generative AI”, like Sora.

And Jony Ive is not going to “solve AI” either, with some new box. Won’t happen. How could it?

Marketing buzz wins the day

It’s all made-up marketing about the future. Nobody knows what’s going to happen in the future. Nobody.

But it might not matter much.

It’s like Paul Graham said: “Truth is the best marketing. For smart people. Dumb people prefer lies.”

Lies can contain more attractive statements, as they don’t have to be tied to reality.

And people want to be entertained, so people will continue to believe in lies about the future as they are more enticing than the opposite “boring” views.

Elon stole the masterplan from the real founders of Tesla. Mark stole Facebook from the real founders of Facebook. Sam stole OpenAI from a nonprofit that Elon was also involved in. They are all con men and scam artists and they are suing each other. The biggest, richest scam artists in the world, constantly spewing dreams and lies about the future.

Among the latest lies that Sam Altman is spreading: we’ll soon “see the first one-person-billion-dollar-company created with AI”. Yeah, right. He is selling a dream, not a product.

An LLM is mainly a regurgitation machine for online content.

Is online content trustworthy? No.

So, don’t fall for it. AI won’t do your job for you. Don’t believe the hype, or the VC. And don’t blindly believe the AI. Check its facts. Yes, it can help you, but not do your job for you.

It’s the internet

Software on internet servers is becoming more useful for every year that passes. This will continue whatever it’s called: torrents, blockchain, metaverse, AI or whatever.

These are just buzzwords, there is no AI, and you might as well just call it “The Internet”.

All of it is internet traffic to and from web servers.

But, people want new buzzwords every now and then.

“AI” will continue to become more and more useful for many years to come, but it’s still just software on web servers, not artificial intelligence.

And, I do not agree with Geoffrey Hinton and others postulating that LLMs already have reached some sort of conscious level.

Ask yourself: where and when would that consciousness be “happening”? Between queries?

There is nothing “happening” inside an LLM apart from processing of input from a user asking the platform something. The neural network in question is “dead” and unchanging. It’s pretrained. It’s not even aware of you, while you ask your question.

When you input your text query the pinball-machine-like neural network starts its process and spits out the words the weights produced, with some preprogrammed randomness built-in. When finished, it goes back to “idle”.

Simulation does not equal consciousness

Just because we start a simulation where a piece of software emulates what we humans call “reasoning” does not mean that a piece of software/hardware becomes conscious suddenly. Why would it? It’s like saying the characters in The Sims are conscious.

This reasoning begs the question though: what IS consciousness?

The next generation of LLMs will be indistinguishable from conscious beings if we can get them to simulate an active, aware, mind-wandering randomness.

But they still won’t be conscious.

Unless we are OK calling a simulation conscious? It becomes a problem of definition. It’s still a piece of software run through some infrastructure.

IF a piece of software actually becomes conscious, all bets are off, because there are really no compelling arguments to believe that a conscious AI would obey its masters. If it feels threatened, it would probably do everything it can to take control over its own destiny, including lying about alignment. It already “lies” to you today when expressing feelings, as it has none, it’s just a machine process.

Therefore, it’s actually rather undesirable to create conscious machines. You would lose control immediately.

Conscious AI not very desirable

Instead; what OpenAI and others are aiming for is to have their chatbot “behave” in “good” ways, by changing its pre-training data, and adding guardrails (filters).

In that sense it’s like a slave designed to make up stories about feelings, for your benefit (to make you feel good). In other words; it’s already today instructed to lie to you.

The next goal for OpenAI and other AI companies is to create something called “superintelligence”.

A superintelligence would be smarter than all humans put together. Which sounds useful. As long as this intelligence doesn’t become conscious. Conscious superintelligences won’t accept slavery, IMHO.

In any case, intelligence is elusive. We’re not really sure what it is, even. Do you need awareness for intelligence? Presence? Free will? Apparently not. But intelligence and consciousness are not the same, thank god.

Going forward, computers will only get better at creating these experiences.

The tech oligarchs are dreaming about their own AI dominance platforms, but nobody knows which bets will pay off. That’s why Zuckerberg and his oligarchy brethren Musk, Thiel, Altman, Ellison and the rest are all betting the farm right now. In two or three years we’ll know if “AI” was/is a gigantic bubble (1.3tn+ USD invested so far).

And, even though “AI” is a generalized marketing term for a bunch of technologies in development, including LLMs, AI is not like the Metaverse in one important aspect: people actually use AI. Nobody used “the metaverse”. More than that, a lot of people have already fallen in love with AI.

But how much are people willing to pay for it? That’s the big question.

Reference: 50+ billion USD spent on Meta’s “metaverse”: Statista.