Tweet épinglé
Please help me become the best version of myself by leaving me anonymous feedback here:
forms.gle/B4ejWSsRvZtm86…
Thank you! 🫶🫶
English
Tim Farkas
110 posts
Tim Farkas
@FarkasTim
visiting student @ Synthetic Neurobiology Lab MIT
Boston Inscrit le Ekim 2017
444 Abonnements141 Abonnés
Tim Farkas retweeté
copyright law seems actively anti-human at this point. I still can't have full-text search over the world's books, but because the LLM providers are de facto allowed to train on them, the incentives point me to read bastardized LLM-paraphrases vs direct quotes from human authors
English
Tim Farkas retweeté
@gleb_razgar if we define techno-humanism as technologically accelerating that what current/future/evolved humans would see as valuable we agree that it's the only good thing?
English
@FarkasTim TH might be the best way to preserve the human value function, but it’s not the only way. Don’t be Eudoxus. Allow for multiple possibilities.
English
@byersblake cool that you are building this and good luck but boooh for not giving credit to @isaakfreeman @h_rosengarten2 who you got the idea from
integrity pays off in the long run!
English
A few friends and I are going to fund a $400k research study on whether we can get a taste of this superpower for everyone. We have some friends that run a gene editing lab at Harvard who are willing to run the experiments. Key points to validate:
(1) Determine at what stage of organism development you need to intervene to get the superpower. We know having these mutations from birth gives you sleep superpowers, but if you edited someone as an adult, would they get equal benefit or is there something about the wiring of your brain during development that is required? And when in development? embryonic, toddler, teenage? Need to run some studies in animals to figure this out.
(2) Some of these short sleeper genetic variants my be easier to drug than others (with RNAi, epi editors, base editors, etc). So we need to run some quick experiments to select which are optimal.
Blake Byers@byersblake
There are people who genetically only need 4-6 hrs of sleep per night. They also "tend to be more optimistic, more energetic and better multitaskers. They also have a higher pain threshold, don’t suffer from jet lag and some researchers believe they may even live longer." We know this can be driven by a single mutation in the gene ADRB1. I need someone to hurry up and make a gene editor so I can get a taste of this freedom. To frame the upside here: in 80 years of life, you sleep for ~20 years. Short sleepers only sleep for ~10 years. Thats +10 years of awake life! Insane life alpha.
English
@euacchq They have known this for many years. Why change now?
(I am super optimistic to hear this though, definitive change in tune).
English
Oops I got addicted to Twitter again. What are the state-of-the-art tips on how to disengage?
English
just became a lot more bearish on BCI - even at the recently accelerated pace, electrode interfaces with just 1% of the human cortex still seems wildly out of reach...
how can we reach actual cortex-scale interfacing 50 years sooner?
English
@IvanVendrov (directed) graphs are great for mapping to hierarchies, networks, causal chains
English
i really believe a lot of social problems come down to mathematical properties of numbers
numbers are incredibly powerful coordination tech but their overuse leads to a measure-predict-control society
what are some alternative mathematical objects we could organize around?
English
It's -17 Celsius outside right now
People will tell me that it makes my city a bad city to live in. Quite the contrary. It's a feature, not a bug
The cold kills. If you are not a civilized person, you cannot live here. It keeps the riff raff out.
You'll actually die
English
Tim Farkas retweeté
text dominik if you are looking for an amazing flat in Berlin
Dom@dominik_scherm
We have two open spots in our hacker house @AureaBerlin for January. DM me if interested.
English
@louis030195 damn respect, that's 1 book read every 3 days!
what would you say is the optimal frequency to optimize not only *no. of books read* but also content learned/absorbed / pondered on a deeper level? is 3 days optimal?
English
read 136 books this year
what's your goodreads profile?
English
IMO, GPT-type world modeling is necessary but most likely not sufficient for AGI. LLMs excel in pattern completion but they lack a coherent self-representation and they can only *simulate* agency. (which may never converge onto *having* human-like agency in the current paradigm)
English
@ChShersh Like the 80:20 of each of the five individually. If you say that's covered by learning CS, how much CS?
English
Tech changes every day.
Fundamentals don't change.
Invest in learning evergreen knowledge. Specifically,
1. Data Structures and Algorithms
2. Computer Science
3. Math
4. Category Theory
5. Relational Algebra
Do not invest (especially don't overinvest) only in learning tech like:
1. Latest framework
2. Latest cool programming language
3. New AI prompting techniques
4. New shiny database
5. New SaaS tool that promises to solve your problems better
6. Shiny DevOps technology
English
@dankuntz X should generally allow individualised tweaking of the algorithm
English
X should have “semantic muting” so I can stop seeing shit like this
iris@irisivanyi
every designer can learn to code very few engineers can learn design
English
