AI @Nominal_io • “The choice is between constructing Babel or rebuilding Jerusalem.” - Pope Leo XIV on AI
NYC
Joined March 2014
- Tweets526
- Following1.1K
- Followers678
- Likes14K
Laya 210M (on the typed decisions checkpoint) is super aggressive as well, and weirdly it always decides to drink all of its Estus right at the beginning of the fight!
Kev 0.8B is hyper aggressive and often spams attacks out of range for no reason. It wins occasionally by just being fast enough to string together a lucky sequence of dodges.
Perplexity's pplx-decider-v1.1-27b is impressively strategic and measured in its attacks, and mostly loses due to a long tail of latency.
Jev is a bit greedier, and the extra latency really costs it. Often times it will just miss the boat at a critical moment and take a hit before it can decide to dodge.
Observing their different play styles is a fun way to characterize the models too. GPT-6 Luna seems to be the most strategic, timing attacks and dodges to line up with Gundyr's attack sequences and position itself well to hit back.
A Dark Souls boss fight is an interesting eval because latency and intelligence are both critical. Small local models aren’t skilled enough to win, despite their much faster latency.
Decision models can play Dark Souls in real time!
I’ve been trying to better understand the spatial/temporal reasoning and latency tradeoffs of different models, so I had them all fight Gundyr.
GPT-6 Luna Decisions came out on top, with multiple no hit runs:
Just filtering on a label in google scholar and screenshotting it as your team slide is a terrachad move
We created a pitch deck to tell a handful of VC firms about us and what we were up to (a fun experience!). Here’s a few slides about our background and some of the things we’ve worked on from the pitch deck (it was fun putting together the list of people in our teams who have gone on to found a whole range of exciting companies). We are delighted to have selected @radicalvcfund
and @khoslaventures to lead our initial funding round, along with participation from @lightspeedvp, @kleinerperkins, Doerr Capital (@johndoerr), and Alphabet (@Google). We’ll be working with them to close our seed round over the next few weeks.
Late stage attention economy
