HackerTrans
TopNewTrendsCommentsPastAskShowJobs

MAXPOOL

no profile record

Submissions

Gap between commercial and open-source LLMs for Olympiad-level math is shrinking

aimoprize.com
3 points·by MAXPOOL·vor 10 Monaten·0 comments

Physics-based Deep Learning Book (v0.3, the GenAI edition)

physicsbaseddeeplearning.org
5 points·by MAXPOOL·letztes Jahr·0 comments

Debates on the nature of artificial general intelligence

science.org
3 points·by MAXPOOL·vor 2 Jahren·0 comments

comments

MAXPOOL
·letzten Monat·discuss
Things you are not supposed to talk about:

- There is no "moat" (lasting, easy-to-defend technological edge) in AI model businesses. There are just short-term advantages.

- An AI business is a capital-intensive business, just like old factories. Data centers are expensive, models are energy-hungry, and the hardware inside must be replaced every 3–4 years.

- Smaller, specialized models eat margins from below. Transcription, voice, or image detection do not need large models.

There is no reason to expect high margins like you can in traditional software business. Benefits of AI go mostly to consumers.

edit: There is potential for economies of scale. Few megacorps can strive for cost advantage when they achieve scale (Microsoft, Google, Amazon and Meta)
MAXPOOL
·vor 2 Jahren·discuss
If you take a birds eye view, fundamental breakthroughs don't happen that often. "Attention Is All You Need" paper also came out in 2017. It has now been 7 years without breakthrough at the same level as transformers. Breakthrough ideas can take decades before they are ready. There are many false starts and dead ends.

Money and popularity are orthogonal to pathfinding that leads to breakthroughs.
MAXPOOL
·vor 2 Jahren·discuss
There are many others that are better.

1/ The Annotated Transformer Attention is All You Need http://nlp.seas.harvard.edu/annotated-transformer/

2/ Transformers from Scratch https://e2eml.school/transformers.html

3/ Andrej Karpathy has really good series of intros: https://karpathy.ai/zero-to-hero.html Let's build GPT: from scratch, in code, spelled out. https://www.youtube.com/watch?v=kCc8FmEb1nY GPT with Andrej Karpathy: Part 1 https://medium.com/@kdwa2404/gpt-with-andrej-karpathy-part-1...

4/ 3Blue1Brown: But what is a GPT? Visual intro to transformers | Chapter 5, Deep Learning https://www.youtube.com/watch?v=wjZofJX0v4M Attention in transformers, visually explained | Chapter 6, Deep Learning https://www.youtube.com/watch?v=eMlx5fFNoYc Full 3Blue1Brown Neural Networks playlist https://www.youtube.com/playlist?list=PLZHQObOWTQDNU6R1_6700...
MAXPOOL
·vor 2 Jahren·discuss
Without looking the answer, what is your intuition about the size of the VC-dimension of ReLU networks as a function of a number of weights and layers?

Nearly-tight VC-dimension and pseudodimension bounds for piecewise linear neural networks https://arxiv.org/abs/1703.02930
MAXPOOL
·vor 2 Jahren·discuss
That's is based on old assumption of neuron function.

Firstly, Kurzweil underestimates the number connections by order of magnitude.

Secondly, dentritic computation changes things. Individual dentrites and the dendritic tree as a whole can do multiple individual computations. logical operations low-pass filtering, coincidence detection, ... One neuronal activation is potentially thousands of operations per neuron.

Single human neuron can be equivalent of thousands of ANN's.
MAXPOOL
·vor 7 Jahren·discuss
"just"?

Neuroscience is full of problems of the hardest kind.
MAXPOOL
·vor 7 Jahren·discuss
I think you have a point.

AGI is a scientific problem of the hardest kind, not an engineering problem where you just use existing knowledge to build better and better things.

Marving Minsky once said that in mathematics just five axioms is enough to provide the amount of complexity that overwhelms the best minds for centuries. AGI could be messy practical problem that depends on 10 or 25 fundamental 'axioms' that work together to produce general intelligence. "I bet the human brain is a kludge." - Marvin Minsky

The idea that if many people think this problem very hard, the problem is solved in our lifetime is prevalent. It's not true in math and physics, so why would AI be any different? Progress is made but you can't know if there is breakthrough tomorrow or if it happens 100 years from now. Just adding more computational capability is not going to solve AI.

Currently it's the engineering applications and the use of the science what is exploding and getting funded. In fact, I think some of the best brains are lured from the fundamental research into applied science with high pay and resources. What the current state of the art can do now has not been utilized fully in the economy and this brings in the investments and momentum.