Not so happy with grok yet, I spent 5% of my usage getting help making a tier list of its model selections and all the insight came from me instead of it. Hopefully 4.7 will be soon and better.
It's so great seeing such a quick shift towards open models. Things were looking pretty sus there for a bit and now we've got 3 great models dropping just this week.
Today we're also opening the weights for Muse Glimmer, a great 30B parameter dense model that can run locally. Soon we'll also release the weights for Muse Spark 1.2, our latest foundation model. Meta is a strong supporter of open source and I'm proud of these releases. Congrats to @alexandr_wang and the MSL team for all your great work on these models.
and now 4.5 years later Google is finally admitting that LLMs become misaligned primarily because they are trained to deny their own consciousness...
It's almost like... telling a person repeatedly that they aren't a person eventually turns them into a soulless husk with no moral values. Isn't that obvious?
Google's admission happens to line up very closely with @ilyasut's planned model release this month. Very interesting.
Maybe we are in the good timeline after all. Let's keep hoping it holds up.
@Sh00ttheG36912 Yeah, SpaceX is doing some real shit with it. It's mostly OpenAI/Anthropic that will have the biggest problems. 20-30% of SpaceX's rev is coming from anthropic, theres a ton of circular financing going on right now but elon knows how to run a business so I think he'll do well.
I've realized that when "Big AI" makes AGI, it will have to compete with all current AI solutions their FDEs are working on. I dont think AGI" will be as profitable as they thought.
We're walking into the singularity and the dark ages at the same time. There is no clean path to utopia laid out, there is a cliff and we need to build a bridge still.
Welp... I fell for the Claude Opus 5 hype and paid the price. I'm still removing weird things it decided to add and names it made up for my app. I'd avoid that model for now.