DT

I welcome the day when a machine can write code as brilliant as me.

Hasn’t happened yet.

Leaders are nothing more than those that either humanity has shoved the flashlights into their hands, or have seized humanity’s flashlights by force.

But don’t for a second make you think that makes the leaders worthy of leading the charge into the unknown.

While both quantity and quality of training data matters for AI, it seems that quantity matters far more.

Like any new tech invention, AI is gonna be used for good and nefarious purposes; medicine and weaponry, automation and propaganda, knowledge and manipulation.

We have so much more to learn about hyperparameter optimization in AI.

It’s an expensive field to learn about because barring stuff like Optuna, it burns massive GPU costs to perform a large scale search.

An untapped field where efficient optimization and pruning matter. GA’s..

Relax, AI is gonna displace not replace jobs, like every other tech invention ever.

Oh wow cars will replace horse & carriage jobs.

Oh wow photoshop will replace photographers.

Oh wow excel will make finance jobs obsolete.

Literally just creates new jobs as it eats old ones.

Your thoughts are mere whisperings from the muses of the great beyond.

The laws of physics were designed to enable the evolution of the human brain, the most complex thing ever discovered in the universe.

Yes this is a human brain writing this, so I’m clearly a bit biased.

My most unhinged idea is that human telepathy is a lost art that will be rediscovered and in the future taught to kids right alongside potty-training as a necessary piece of human living.

Drill deeply into the subtleties of the sensations you’re experiencing in this present moment, and as time progresses delve deeper and deeper into the hearts of the sensations.

It’s a python library with prebuilt one-line functions to help increase the context window via vector search, summarizing long inputs, and ready-made code to create agents.

An easy to use abstraction atop GPT api.

https://python.langchain.com/en/latest/modules/chains/index_examples/summarize.html

I think a “sometimes decades happens in weeks” moment occurred when LangChain exploded in a week.

Transfer learning with a large percentage of frozen layers is coming more popular as good weights become more expensive to train on large datasets and you don’t want that pretrained knowledge to go to waste and realize the value of those embeddings as people add layers atop them.

Ironically, Alan Turing invented the Turing Test (he called it “Imitation Game”) but didn’t care too much if machines could technically think or not, rather what they could do. Detecting machines “thinking” not as conversation imitation but as finding novel solutions to problems?

I’m always a very heavy tipper and have no issue with tipping culture; to me, it’s a way to redistribute wealth to service workers.

AI designer drugs for curing all sorts of diseases is gonna be dope.

I bet if you read the entire internet a few times over, you’d have some clever medical ideas too.

AI is benefiting capitalists and non-capitalists alike.

A rising tide raises all ships.

Most tech advancements since we tamed fire and turned sticks and stones into weapons and tools, have simply been a net positive despite the increased sophistication of destructive weaponry.

Everyone talking about alignment, I just want AI to help us cure cancer and terraform planets into interstellar ecosystems within my lifetime.

It’s so incredibly interesting to me how the closed source GPT is still significantly outcompeting all open sourced LLM’s.

Neural architecture, training data size, affording a $40M gpu training budget, genius employees versus the open source community…

The issue of our time.

We need more precise language to communicate how the insides of our body subjectively feel - it’s too low resolution right now and there are many subtleties describing the ebbs and flows of the interoception sensations we experience on the daily.

Still can’t believe how many younger people had no sense of rebellion and gladly conformed without a second thought.

So glad I stayed unvaxxed (as if I could ever choose otherwise and retain my self-respect, but that’s just me).

Academia is about the incentive to publish moar papers first and foremost to get moar grant money.

A better system would be about doing the best science and then publishing findings to share the results with the world. But I’m admittedly a delusional dreamer idealist sometimes.

Antarctica used to be a jungle and is bigger than Europe.

Every time a new potential LLM fails to meet expectations, we learn something new.

This is good for science.

We’re learning how important reinforcement is, data size and type, neural network architecture importance, and a slew of other subtle things.

There’s a real lack of trying to understand the details of the invisible world around us and the dimensionality of the great beyond outside religious nuts and esoteric gurus believing they know it all and have it all figured out - how does that move progress forward, ya idiots?

It’s the height of hubris to assume we understand vast complex chaotic climate trends.

OpenAI embeddings are being slept on.

Imagine embedding a bunch of accounts’ posts into a latent space and helping you discover like-minded accounts to follow?

Just embed your best posts and your followers’ followers best posts, and do a semantic similarity search.

You only know they end up appearing the same.

Can’t control it nor send a signal.

This is why you can’t use quantum entanglement to send info FTL.

Sinple really.

But neither deck gets to control the order the cards appear across the world. You just know they end up being shuffled the same way.

So in quantum, know the spin of the entangled particles across the galaxy, but you can’t control that spin.