Deeply Thrilling Telegrams
1.7K subscribers
134 photos
23 videos
46 links
Download Telegram
One of my biggest peeves is bad progress bars.

Progress bars should approximate, as best as they can, the amount of time remaining.

A progress bar at 50% after 2 minutes should mean there’s about 2 minutes remaining.

A progress bar at 75% after 3 minutes should mean that there’s about 1 minute remaining.

That same progress bar should take just as long to go from 10% to 20%, as it does going from 90% to 100%.

Don’t even get me started on progress bars that go to 99% and then hang for an untold amount of time.

Programmers should make their progress bars be linearly coordinated with time remaining. Any other approach is unacceptable.
Merry Christmas!
I paid the $200/month for ChatGPT Pro this past month.

I’m sure it can solve advanced math problems really well…

But honestly there’s nothing I’ve used it for that the $20/month version couldn’t do.

It is cool how it iterates its own answer until it thinks it’s “most correct”, scaling inference compute instead of scaling training compute.

But I think OpenAI is optimizing for benchmarks like advanced math problems to show off to the community instead of solving an actual consumer needed use case.

I don’t think scaling is going to lead to AGI alone; we need creative paradigm shifts (like we had when the first ChatGPT came out) not just THROW MOAR COMPUTE MONEY AT THE PROBLEM to crack AGI.
🍾 Happy New Years 2025, everyone!

May your 2025 be blessed with excellence and love.

My New Year’s resolution is to keep my fridge fully stocked with healthy, fresh food.

What’s yours?

Hope you’ve enjoyed my musings for 2024. Comment below what you want to hear more of for 2025 in the Deeply Thrilling Telegrams 👇
Man is to God as a neuron is to a body?
Multiverse Theory with Von Neumann Quantum Interpretation implies that our consciousness creates free will by collapsing the probability waves that create our neurotransmitter quarks and direct them in the direction of our “choices”.
WHO ARCHIVES THE ARCHIVE?

One of the biggest issues of our time is: Who archives the archive?

So much knowledge and wisdom hinges on The Wayback Machine.

It’s a single point of failure and thus a major strategic risk for the future of humanity.

We need to decentralize the archive of the Internet, perhaps on a blockchain.

And before it gets hacked and modified. We need proper version control on The Archive.
You know, they say it’s impossible for a human to beat the machines at chess.

Maybe one day, we will birth someone so genius, so creative, so intelligent, so prophetic, that they’ll beat the machines at chess.
Humans first have a concept they want to express in their head, and use words to best express it.

LLM’s use relationships between words to flesh out probable concepts via generated tokens that are going to be coherent and rational.

Similar, but not equivalent.
When the news says “the rate of inflation has significantly reduced!” many (uneducated) people assume that means prices will start to go down.

No, it means the prices are still going up, just not as rapidly.

If you recall your high school calculus, inflation is the first derivative and the rate of inflation is the second derivative.

People assume “the rate of inflation has decreased!” means the first derivative is negative, but really it’s the second derivative that’s negative.

(Aside, if you never learned calculus, that’s a major failing of your education, your parents, and you. Get educated, anon.)

Saying the rate of inflation is going down is like saying the fat girl is still getting fat, but not as quickly as before; it doesn’t mean she’s losing weight.

Remember: Second derivative, not first derivative.
We need to bring back Windows 3.1 style macros of operating system clicks and actions. iykyk.
Most people need a target to aim their decisions towards in order to understand and embrace the meaning of their lives.

For me, I have both a personal and a professional target.

My professional life is aimed towards a super high tech, futuristic, video-game-esque, super intelligent AI hospitals that actually heal or prevent real diseases, and takes healthcare AI into the twenty fifth century. Star Trek Tricorder type medicine.

Find your purpose, and aim your decision-making spear through the quantum foam, towards that slice of the multiverse containing your vision!
My fear of being replaced by a machine is zero percent, plus or minus zero.

If a machine can do what you do, skill issue.
What are the future emperors of humanity doing now? Because you know they’re not in positions of power currently.