Pinned
new blog post
"There Are No New Ideas In AI.... Only New Datasets"
in which i summarize LLMs in exactly four breakthroughs and explain why it was really *data* all along that mattered... not algorithms
Jack Morris
5,622 posts
Jack Morris
@jxmnop
research // language models, information theory, science of AI // formerly @cornell
- *taps the sign*
the task length an ai can reliably finish (conservatively) doubles every 7 months when i'm the age my mom was when she watched me graduate, ai will be able to do tasks that would take someone ~1000 millennia - i'm about to be so rich
00:29 - i wonder if this is what it will be like when we get 10-trillion parameter LLMs
- my grandmother’s lifetime - color film - talking movies - television - cars - airplanes - plastics - penicillin - transistors - computers - the Internet - the Concorde my lifetime - gpt-1 - gpt-2 - gpt-3 - gpt-4 - gpt-5 - the iPad - ozempic
- me when i come across an exceptionally well-crafted personal website
- this guy invented VLLM. he's basically the john wick of CUDA kernels
- observations from my first two weeks as a Meta research intern - research jobs are the same everywhere: no one ever asks me what I’m doing or how I’m spending my time; there’s an implicit expectation to be interested and work hard - biking to work in the sunshine has noticeably
- heard a funny story about a friend who worked at a French LLM startup a couple yrs ago > their plan. to be the first to market with a certain type of multilingual model > early in year: incorporate. start building > hired an awesome team by march > scraped / acquired all the
- openAI: we will build AGI and use it to rewrite the social contract between computer and man DeepSeek: we will build AGI for 3% the cost. and give it away for free xAI: we have more GPUs than anyone. and we train Grok to say the R word
- my understanding is that both AMD and qualcomm make chips that have ~equivalent performance to nvidia but neither can write the software tooling that N provides, like CUDA i get that it's complicated, but which part of the stack could possibly be so hard to replicate? are
- the first evidence i ever saw of superintelligence was in medicine deep learning can tell male from female eyeballs from just a picture with 70–90% accuracy doctors still can't do this, and don't understand how it's possible
- the great irony here is that we aren't even close to having the tech required to build this, so their customers are actually the ones being cheatedCluely is out. cheat on everything.
00:00
