It's weird to see some folks acting like LLMs are the path to actual artificial intelligence.

Call me when an LLM can make deductions and have original thoughts without being fed the entirety of the Internet first. That's not how actually intelligent humans work. Of course compressing the entirety of human knowledge into a fuzzy completion machine will make it sound intelligent. You're basically making a lossy encoded compressed version of things actually intelligent people have said, with a modicum of ability to generalize by simple association and statistics.

The Turing test breaks down when you throw enough Big Data at it. Congratulations, you've beat the test. You haven't created real intelligence though, you used big data to build a parrot with a huge vocabulary.

can’t believe so many of you have you have stuck around for a year since this all started taking off. you the real ones. y’all are awesome 🫂

After using #Nostr for several weeks, one thing that has taken me by surprise is how there are very liberal and very conservative people on here, and yet they both share a sense of community and don't spend all day posting content against each other. Despite differences in political persuasion, they find common ground on here.

Now that I've seen this, I better understand how dangerous centralized control of social media is -- where algorithms continuously drive inflammatory content to users to keep them addicted to the platform. I find it ironic that a censorship-resistant platform is largely devoid of much of the inflammatory content that's present on censorable social media. One may have thought the opposite would be the case.

👀

Written during my initial awakening to technology and culture, there's a melancholic beauty in realizing that living through the intervening years is essential to fully appreciate the longevity of his ideas.

Neil Postman published Technopoly in 1992 and it's still one of the most prescient books i've read

right up there with The Sovereign Individual (1997)

Yes, if you think about it… words themselves are extremely low bandwidth. So it initially seems human communication is extremely slow.

But combinations of just a fews words is actually an incredibly vast storage space (12 words prove this phenomena).

Sentences (not even paragraphs) have an enormous Shannon number, and therefore sentences are essentially an insanely powerful compression function for communication of the sophisticated abstractions that we each hold in our heads.

LLM’s have undeniably cracked this compression function and also now build their own internal abstractions from these decipherings. LLM’s are also able to communicate their abstractions the other way via the same compression function.

Those 3 capabilities are really impressive and were actually quite simple to reach through mere scaling.

But LLM’s are yet to emulate abstractions of our direct jungle experience. I think ML can do this, as we see with self guided rocket boosters, drones and driverless cars. I think the multimodal LLM’s are now beginning to bridge the two worlds of cyber and jungle. Once they do, the next step is that they can then navigate the jungle by being embodied as robots.

I expect we will see all that in around 10 years. The race is real, still not spoken about much is the fact that it solves demographic taxbase erosion without the need for inward migration for all the major economies and we are already seeing Western politics turning away from the immigration strategy.

You can walk through where this is all heading and the societal choices that populations and powers will face along the way.

But again, I think we will land on highly specialised forests and not monoliths. I think we will see declouding and embodiment and I think we will own local embodied private agents, with the state owning large numbers of their own public agents.

A social protocol of disseminated power will emerge and it will be enforced by the tyranny of majority.

I think society is going to go through a series of radical changes, migration U-turn will be the first, as we are collectively persuaded to collaborate with the interests of our local power centre (political capital).

"Pathological anxiety about fate and death impels toward a security which is comparable to the security of a prison. He who lives in this prison is unable to leave the security given to him by his self-imposed limitations.

But these limitations are not based on a full awareness of reality. Therefore the security of the neurotic is unrealistic. He fears what is not to be feared and he feels to be safe what is not safe.

The anxiety which he is not able to take upon himself produces images having no basis in reality, but it recedes in the face of things which should be feared. That is, one avoids particular dangers, although they are hardly real, and suppresses the awareness of having to die although this is an ever-present reality.

*Misplaced* fear is a consequence of the pathological form of the anxiety of fate and death."

—Paul Tillich

From my own experience scrubbing up the data quality is the secret sauce in all of this, I would take halving the anomalies v’s 25x volume in any training dataset.

But my point above is really about the application of AI. I still think generalisation is a mirage, you want 10^6 specialist models each deployed on 10^3 instances rather than 1 general model deployed 10^9 instances.

You get those specialists via automated data quality ops.

Automate the folks doing the automating.

Build a specialist forest instead of a monolith.

Then you have to workflow the specialists.

Then make it dynamic / self organising.

This is the path.

OK, this is the most profound forward looking thing that’s occurred to me in the past 6 months.

For 30 years big tech is built on B2C models that leverage the enormous surface of the internet. Build 1 good monetisation channel and just scale it to 10^9 people.

But AI is not like this. Not at all.

AI is a solution to the sprawling complexity of discrete connections that network throughout the entire economy. It is 10^6 monetisation channels each occurring 10^3 instances.

To win this space you need a B2B biz model and you need legendary workflow management.

This makes it much easier to filter out winners and losers over the next 10 years.

I actually think ChatGPT will disappear as it’s a loss leader product and only the OpenAI API will survive, but it will have to compete with a bunch of other API’s and local models.

I think it’s obvious that the end state is local AI. There will be a day when training peaks and then terminally declines as economics is the real law of the jungle.

Chasing diminishing returns is simply not worth the cost. There is no free lunch here.