@levie@elonmusk let us customize the prompt for this feature. Personally I want to include 2 key instructions. Include the current scientific consensus for the claim made in the post and argue both for and against the claim. Game changer for rooting out disinformation and bias.
@ryangerritsen This map is only for Peel regional police events. Naturally the reported events are all concentrated in one area, Peel. The few sparse events outside of Peel do not reflect total events outside Peel. In order to draw conclusions we need the same data for every other region.
@Jason@DavidSacks@GavinSBaker@JTLonsdale@Jason I’m not arguing for or against your stances. But my opinion is All in Pod is great b/c you argue both sides. Lately I think you’re getting sucked into the echo chamber. No judgement, I would too. But you and friedberg must explore more search space. Kick the tires.
Does anyone think shoedog is actually the story of a person that had PMF and despite their sheer unlikelihood to be great the market pulled it out of themanyway?
@awilkinson You’ve already got tech in your stack at Tiny that can do this and it’ll be secure with the ability to use diff LLM models. Technically chatgpt will use your prompts for training their models unless you’re on an ent edition. @awilkinson shoot me a DM and I can help out!
@neelesh_salian Because it’s an open source foundational model some builders will leverage it to train their own fit for purpose LLM using it as a base. StackLlama is an example of a further trained version of Llama. The graphic you shared is relevant for those use cases.
@laurenbalik Love the idea. I’ve always thought it’s stupid that we can’t tune our own preference profiles. Why the heck does @Meta not expose who it *thinks* I am? That would be a very cool feature. May be enlightening to compare who you think you are with what the algo thinks.
@neelesh_salian @neelesh_salian I’ve seen Snowpark data frame POCs come in at about 2x+ faster and 2x+ cheaper than Glue, EMR, and dbx with 0 overhead/cluster management. Curious whether there’s a reason not to check it out - I like hearing from the market. Disclaimer I work at snow
@neelesh_salian I got you @neelesh_salian and agree there. I was under the impression you meant the engineering org as a whole. What you’re saying is analogous to salesforce sdrs calling front line sales reps trying to hawk CRM. No business doing that.
@tunguz Fine tuning would work for this right now. Extract the text from your corpus, upload it to a fine tuning model with no prompt just completions and then you can use the fine tuned model just like ChatGPT. Easy as pie.
Raw activity metrics are useless. Want to learn something? Run a competition to see who can hit the monthly quota with the least activities.
For ppl stuck with raw metrics: To hit 50 dials, simply call a big corporation's main line and leave a VM for extensions 101, 102, 103...
@paulg@levie Make it elective. Don’t want to wait on a wait list for CA/NY to come over? Go to AL, LA right now - no wait list. (Obv exception for those with a need to immigrate)
@Jason IRL is still superior for quality of engagement, no doubt. but why are you indicating that VR means you can’t hang and have a fun time there? I want to hang with friends in other cities and have an opportunity for more types of fun times depending on mood/current interests.
Is being on time not a norm anymore? I interviewed 5 ppl this week and not one of them was waiting on the zoom by the time I opened it. 2 of them didn't show up until 7+ minutes later. Competitive advantage for job seekers seems to be pretty easy...be on time.