feat: Add agi-compute

config.toml

@@ -1,10 +1,10 @@
 baseURL = 'http://example.org/'
 languageCode = 'en-us'
-title = "Amaury's Blog"
+title = "Storm's Blog"
 theme = 'xmin'
 
 [menu]
 main = [{ name = "Home", url = "/", weight = 1 }]
 
 [params]
-footer = "© Amaury 2023"
+footer = "© Storm 2023"

content/_index.md

@@ -2,8 +2,8 @@
 title: Home
 ---
 
-# Amaury's Blog
+# Storm's Blog
 
-A blog where I try to find what I'm interested in.
+I'm Storm, and you're reading a blog where I try to find what I'm interested in.
 
 ## Posts

content/post/01_write_as_you_are.md

@@ -1,6 +1,7 @@
 ---
 title: "Write as you are"
 date: 2023-11-23T09:28:59+01:00
+slug: write-as-you-are
 draft: false
 ---
 

content/post/02-agi-compute/index.md

@@ -0,0 +1,31 @@
+---
+title: "AGI is just a matter of more compute"
+date: 2023-12-19T09:28:59+01:00
+slug: agi-compute
+draft: false
+---
+
+Human can write poems, erect buildings and go to space. Animals can't. I've always believed that brain size was not the only reason for this, and that humans had something else: a higher cognitive ability, different not only quantitatively but also qualitatively. An intelligence that was different in nature.
+
+With the recent developments of LLMs, I would like to challenge that.
+
+Let's consider primates. We learned at school that there's some positive correlation between brain size and behaviors that indicate intelligence. Research also tends to confirm this, for example [this one](https://www.pnas.org/doi/10.1073/pnas.082105099) by Reader and Laland which "found significant, positive correlations between brain size and all three behaviors", referring to innovation, social learning, and tool use.
+
+![Human brains (top left) are bigger](./02_1-s2.0-S1053811920311708-gr2_lrg.jpg)
+_Human brains (top left) are bigger. [Source](https://www.researchgate.net/publication/342720361_Imaging_the_primate_brain_evolution_the_next_frontier)._
+
+There are obviously counter-examples. A study from the German Primate Center (DPZ) found that cognitive abilities of lemurs, which have smaller brains, were on par with those of primates with larger brains ([source](https://www.sciencedaily.com/releases/2020/09/200925113353.htm)). However, in general, I would argue that this positive correlation exists.
+
+Other primates still don't write poems, erect buildings and go to space (at least not willingly). So we can argue that human brains, via evolution and genetic mutations, simply surpassed that threshold necessary for general intelligence, and then conquered the world. Maybe "threshold" is not a good metaphor either, it's probably more linear and progressive, but you get the idea. Anyways, after this threshold, "la mayonnaise prend" [^1], and cognitive abilities grow exponentially.
+
+My hypothesis is: it can be the same with artificial intelligence. Today, LLMs display some kind of intelligence, definitely not enough to be qualified as AGI. But what if we throw 10x, 100x more computation power into LLMs? There's a small but non-negligeable chance that this will be achieve AGI, just like with apes. I guess this is the belief of people claiming we'll have [AGI soon](https://twitter.com/robbensinger/status/1537724894678380544). It's defendable.
+
+And the example of lemurs having smaller brains yet being on par with larger primates reminds me how smaller LLMs (in terms of parameters) like Mistral can give results almost as good as larger ones like GPT4. Maybe there are more parallels between primate intelligences and artifical intelligences.
+
+I read [somewhere on the internet](https://publicfirst.co.uk/ai/) that people think AGI will arrive somewhere between 2030-2039 (median value). We'll see by then if the hypothesis is true or false.
+
+And at that time, let's also circle back to what that means for human intelligence. If we do achieve AGI by throwing more computation power, than it supports the idea that human intelligence is achieved by surpassing a critical brain size. If not, then it obviously doesn't disprove this thesis, but makes it less promising for sure.
+
+See you in 2030-2039.
+
+[^1]: You know when you make mayonnaise yourself, your stir and stir the liquid mix of egg yolks, mustard and salt. Then at some point, this liquid becomes thick and creamy. That's when "la mayonnaise prend" (French expression).