Anyone who tells you they know where Al is headed is lying to you.

[Jason Voorhees]

If some tech influencer claims they have a roadmap for the next five years of AI and machine learning, here's how you can be prepared, they are just trying to sell you a course. I have 2 friends friend who are Al researcher at one of the top frontier labs and I'll tell you an open secret in the Silicon Valley. The very people building these massive models have absolutely no idea where we are actually headed.

One thing you need to realize about AI is AI isn’t a closed engineering problem lilike say building a car. we know everything there is to know about a car but for AI it's this strange living learning developing thing that we still don't fully understand how or why it works at scale. It’s not closed or open in the normal sense.

AI has advanced so rapidly with so much money being poured into it that the timeline for major breakthroughs has collapsed so quickly that the engineering has entirely outpaced the science. Like just 2 years ago we learned about emergent properties. AI learning stuff it wasn't trained for explicitly. AI being fed coding corpus somehow learning French in the process etc. Why this happens still not properly understood. Just theories about information hierarchies.

Much of the tech industry rn is stuck in massive hype cycle. Everyone is moving where ever the investor money blowing, in whichever way the wind turns. There is no master plan or milestones these companies operate on. They are just reacting to what they've built rather than executing a predictable master plan. Like oh wow AI did this. Interesting. Not validation of theory but discovery. No body can tell with 100% certainty that this or this is going to happen. If he does he is either some ultra genius who can see the future or someone who has no idea what he is talking about. We will just have to wait and watch how all this will play out in this decade

 

[R@m@]

 

[Joeseminate]

Either: small boost to gdp, expontential growth or complete extinction of humanity

 

[Jason Voorhees]

@imontheloose @lnceIs @Divineincel @Chadeep

 

[tansel]

i’ll give my thoughts on this when i get tagged

 

[Waffe]

like building a car like say building a car

oops

 

[mcmentalonthemic]

Wherever it's headed it'll be a huge under performance of what people think it'll manage to do

 

[MediterraneanMoid]

 

[inceliusndius]

jason do you think a company will pay a group of 3rd worlders with no experience in a shit college 30k a year to do their compliance (soc2)

asking for a friend

 

[Jason Voorhees]

@tansel

i’ll give my thoughts on this when i get tagged

 

[EvilSatanArseRapist]

View attachment 5076936

What a shitty ai reaction image

Wherever it's headed it'll be a huge under performance of what people think it'll manage to do

depends a lot on the person thinking..

I'm not sure abt that tbh, idk if it can keep getting crazier nd crazier exponentially as it has been doing for THAT much longer, but i wouldn't exclude it eather..

 

[Jason Voorhees]

@Petsmart @Embryo @CollioureViews @AgentAngularity

 

[krieg]

Headed to sex robots

 

[tansel]

@tansel

i don’t have any thoughts on this tbh lol

 

[Jason Voorhees]

jason do you think a company will pay a group of 3rd worlders with no experience in a shit college 30k a year to do their compliance (soc2)

asking for a friend

No. Don't get into compliance without creds especially b2b.

 

[AgentAngularity]

Ai will never live up to the hype.

 

[Jason Voorhees]

What a shitty ai reaction image

depends a lot on the person thinking..

I'm not sure abt that tbh, idk if it can keep getting crazier nd crazier exponentially as it has been doing for THAT much longer, but i wouldn't exclude it eather..

Give reps then replies

 

[gooner23]

I think LLM generalization is its weakness i think phyiscal model labs are trying to cover that

Yann LeCun's AMI Labs raises $1.03B to build world models | TechCrunch

“My prediction is that ‘world models’ will be the next buzzword,” AMI Labs CEO Alexandre LeBrun told TechCrunch. “In six months, every company will call itself a world model to raise funding.”

techcrunch.com

 

[inceliusndius]

No. Don't get into compliance without creds especially b2b.

but my friend said it was a good idea

 

[1966Ford]

This is one of those "big new thing"

 

[Divineincel]

If some tech influencer claims they have a roadmap for the next five years of AI and machine learning, here's how you can be prepared, they are just trying to sell you a course. I have 2 friends friend who are Al researcher at one of the top frontier labs and I'll tell you an open secret in the Silicon Valley. The very people building these massive models have absolutely no idea where we are actually headed.

One thing you need to realize about AI is AI isn’t a closed engineering problem lilike say building a car. we know everything there is to know about a car but for AI it's this strange living learning developing thing that we still don't fully understand how or why it works at scale. It’s not closed or open in the normal sense.

AI has advanced so rapidly with so much money being poured into it that the timeline for major breakthroughs has collapsed so quickly that the engineering has entirely outpaced the science. Like just 2 years ago we learned about emergent properties. AI learning stuff it wasn't trained for explicitly. AI being fed coding corpus somehow learning French in the process etc. Why this happens still not properly understood. Just theories about information hierarchies.

Much of the tech industry rn is stuck in massive hype cycle. Everyone is moving where ever the investor money blowing, in whichever way the wind turns. There is no master plan or milestones these companies operate on. They are just reacting to what they've built rather than executing a predictable master plan. No body can tell with 100% certainty that this or this is going to happen. If he does he is either some ultra genius who can see the future or someone who has no idea what he is talking about. We will just have to wait and watch how all this will play out in this decade

They will replace us and then they will kill us with AI, that’s where we are heading

 

[Jason Voorhees]

I think LLM generalization is its weakness i think phyiscal model labs are trying to cover that

Yann LeCun's AMI Labs raises $1.03B to build world models | TechCrunch

“My prediction is that ‘world models’ will be the next buzzword,” AMI Labs CEO Alexandre LeBrun told TechCrunch. “In six months, every company will call itself a world model to raise funding.”

techcrunch.com

He literally says this "In six months, every company will call itself a world model to raise funding." This is exactly the investor wind I’m talking about.

 

[1966Ford]

Either: small boost to gdp, expontential growth or complete extinction of humanity

Def a boost to replace those white collar jobs, but can't be said on the blue collar.

The extinction of humanity is stretch tho lol. ..
A.I is not FUCKING Ultron

 

[GonorrhoeaGobbler]

View attachment 5076936

we are just to blind to see how horrible things are

rome fell in 200 years

 

[1966Ford]

Wherever it's headed it'll be a huge under performance of what people think it'll manage to do

What makes you think that tho?

 

[IronMike]

I own 500 shares of Sivers Semiconductors and 150 Shares of Poet with a buy in price of 4$

I know where its going

 

[ghost!]

my dad is the head of all Aritificial Intelligence at a company with a maket cap of over 690 billion.

if you guys have any questions i could ask him

@Jason Voorhees

 

[Overlord-]

I own 500 shares of Sivers Semiconductors and 150 Shares of Poet with a buy in price of 4$

I know where its going

AI porn

 

[Jason Voorhees]

my dad is the head of all Aritificial Intelligence at a company with a maket cap of over 690 billion.

if you guys have any questions i could ask him

@Jason Voorhees

Does he believe the current LLM can ever be made 100% reliable for high stakes enterprise use?

 

[IronMike]

AI porn

 

[gooner23]

He literally says this "In six months, every company will call itself a world model to raise funding." This is exactly the investor wind I’m talking about.

can you give me an example of somethiung like a frontier lab not having a master plan, they always have a timeline as you look in the past, web search, o1 thinking, 1m context, agents/mcp, orchestrator/worktrees, inference cost (stargate 500b). I feel like going where the money is going, is what people want to happen and scaling laws are exactly where we predicted models to be.

only anthropic is really guilty of this imo claude mythos hype is probably the most misleading model release of all time..

obviously 2030 is a stretch since chatgpt 4 was march 14, 2023

and the musk vs open ai trial is kind of eye opening for this topic since mira murati explicitly says you have to go where the money is

 

[ghost!]

Does he believe the current LLM can ever be made 100% reliable for high stakes enterprise use?

yes, he recently got to his position by laying off a lot of employees which fully got replaced by claude.

Idk if that answers your question but his whole team runs off claude currently.

 

[Overlord-]

View attachment 5077009

Some pajeet larped as blonde AI maga girl and earned few millions from scamming maga voters @Jason Voorhees

 

[Jason Voorhees]

can you give me an example of somethiung like a frontier lab not having a master plan, they always have a timeline as you look in the past, web search, o1 thinking, 1m context, agents/mcp, orchestrator/worktrees, inference cost (stargate 500b). I feel like going where the money is going, is what people want to happen and scaling laws are exactly where we predicted models to be.

only anthropic is really guilty of this imo claude mythos hype is probably the most misleading model release of all time..

obviously 2030 is a stretch since chatgpt 4 was march 14, 2023

and the musk vs open ai trial is kind of eye opening for this topic since mira murati explicitly says you have to go where the money is

Those things you mentioned are infra things. there is a big difference between having a hardware checklist like building a cluster like stargate and possessing a proper theory with a blueprint. Like you said Mira Murati it is heavily dictated by the desperate need for funding. They know how to scale the hardware, but as the shifting goalposts and misleading hype cycles show t hey are still just brute forcing the science and following the capital to see what capabilities emerge next. Anthropic npm leak a few weeks ago made all this very clear. It didn't look like a controlled, predictable, scientific master plan it looked like a lab frantically building hidden, reactive systems

 

[Jason Voorhees]

yes, he recently got to his position by laying off a lot of employees which fully got replaced by claude.

Idk if that answers your question but his whole team runs off claude currently.

How did it pass the SOC2 compliance then? And how is he running the LLMs? Is it on bedrock or some custom orchestration.

Also, if the whole team just runs off Claude why do they need him a highly paid Head of AI to manage an API key.

 

[ghost!]

How did it pass the SOC2 compliance then? And how is he running the LLMs? Is it on bedrock or some custom orchestration.

Also, if the whole team just runs off Claude why do they need him a highly paid Head of AI to manage an API key.

for the first one he said he couldnt answer (prob sum do to w privacy)

for the second one he said the integration of claude is complicated on a large scale and he also works on AI products the company works on

 

[CollioureViews]

If some tech influencer claims they have a roadmap for the next five years of AI and machine learning, here's how you can be prepared, they are just trying to sell you a course. I have 2 friends friend who are Al researcher at one of the top frontier labs and I'll tell you an open secret in the Silicon Valley. The very people building these massive models have absolutely no idea where we are actually headed.

One thing you need to realize about AI is AI isn’t a closed engineering problem lilike say building a car. we know everything there is to know about a car but for AI it's this strange living learning developing thing that we still don't fully understand how or why it works at scale. It’s not closed or open in the normal sense.

AI has advanced so rapidly with so much money being poured into it that the timeline for major breakthroughs has collapsed so quickly that the engineering has entirely outpaced the science. Like just 2 years ago we learned about emergent properties. AI learning stuff it wasn't trained for explicitly. AI being fed coding corpus somehow learning French in the process etc. Why this happens still not properly understood. Just theories about information hierarchies.

Much of the tech industry rn is stuck in massive hype cycle. Everyone is moving where ever the investor money blowing, in whichever way the wind turns. There is no master plan or milestones these companies operate on. They are just reacting to what they've built rather than executing a predictable master plan. Like oh wow AI did this. Interesting. Not validation of theory but discovery. No body can tell with 100% certainty that this or this is going to happen. If he does he is either some ultra genius who can see the future or someone who has no idea what he is talking about. We will just have to wait and watch how all this will play out in this decade

Regardless of the pace of development, the adoption curve is its own beast.

Shit, we could have transatlantic flights in half the time like we used to, but supersonic travel is a nightmare.

Similarly, digital record keeping was already viable in the 60s and 70s, but many institutions lagged behind. I remember HMRC and my doctors "going digital".

 

[Jason Voorhees]

Bjml

 

[Jason Voorhees]

@Kara @mohito

 

[mohito]

The so-called renowned analysts are always so good at underestimating emerging trends and making terrible predictions lol

 

[Chadeep]

Thoughts on Citadel Ceo saying this.

 

[Jason Voorhees]

@Glorious King

 

[Glorious King]

@Glorious King

Such a good thread

And I get tagged dead last

 

[Glorious King]

Tbh all I can see is in the next 6 months

We'll see more " worlds first muh model " chud startups raise huge funding

 

[Alienmaxxer]

I'm doing machine learning for my masters thesis so I can enter this space. Very exciting next few years.

 

[Jason Voorhees]

I'm doing machine learning for my masters thesis so I can enter this space. Very exciting next few years.

Is your thesis leaning more toward building/optimizing models or toward interpreting it. The black box problem understanding how these massive models actually arrive at their answers in the holy grail rn

 

[Jason Voorhees]

@BigBallsLarry

 

[BigBallsLarry]

If some tech influencer claims they have a roadmap for the next five years of AI and machine learning, here's how you can be prepared, they are just trying to sell you a course. I have 2 friends friend who are Al researcher at one of the top frontier labs and I'll tell you an open secret in the Silicon Valley. The very people building these massive models have absolutely no idea where we are actually headed.

One thing you need to realize about AI is AI isn’t a closed engineering problem lilike say building a car. we know everything there is to know about a car but for AI it's this strange living learning developing thing that we still don't fully understand how or why it works at scale. It’s not closed or open in the normal sense.

AI has advanced so rapidly with so much money being poured into it that the timeline for major breakthroughs has collapsed so quickly that the engineering has entirely outpaced the science. Like just 2 years ago we learned about emergent properties. AI learning stuff it wasn't trained for explicitly. AI being fed coding corpus somehow learning French in the process etc. Why this happens still not properly understood. Just theories about information hierarchies.

Much of the tech industry rn is stuck in massive hype cycle. Everyone is moving where ever the investor money blowing, in whichever way the wind turns. There is no master plan or milestones these companies operate on. They are just reacting to what they've built rather than executing a predictable master plan. Like oh wow AI did this. Interesting. Not validation of theory but discovery. No body can tell with 100% certainty that this or this is going to happen. If he does he is either some ultra genius who can see the future or someone who has no idea what he is talking about. We will just have to wait and watch how all this will play out in this decade

This is offtopic for AI discussion but i always have a rule in my head that if someone needs to try hard to convince me to buy something (aka. broo i know where ai is headed you should buy my course brooo) its ALWAYS bullshit. If something is good i wouldn't need to be repeatedly convinced to buy it.

This is ESPECIALLY the case for such unpredictable things like crypto or AI or any other scheme things like those

 

[AstroSky]

The mass amount of data centers near me they are building and hiring pip fitters and all these laborers tells you all you need to know. They know something we dont. The money is massive. The land they are distorying in my state is huge and no one here is happy about. The water and resource these places take to run is insane. This is not a bubble at least right now. They are throwing trillions at this shit. They are trying to race to get there first. At least its creating jobs for people for now ...to build it.

 

[AstroSky]

I own 500 shares of Sivers Semiconductors and 150 Shares of Poet with a buy in price of 4$

I know where its going

Make sure to buy data center stock. They sure will grow in the future....the sheer mass of them being built in all states is a telling sign. I might buy a few shares since money is not an issue. Only problem is some are private and you gotta know someone to know someone to get in. Rich get richer I guess